Download here - AgentLink
Download here - AgentLink
Download here - AgentLink
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
European Coordination Action for Agent-based Computing<br />
News<br />
Features<br />
Responsive and Generative Music for Public Installations<br />
Semantic Web Services with the Web Service Modeling Ontology (WSMO)<br />
Trust: Challenges and Opportunities<br />
Why Argue Dialogue Types and Argumentation in Agent Interaction<br />
Answer Set Programming and Agents<br />
AgentCities.ES: Spanish network of agent-based computing (2003-2005)<br />
19Agent Research Overview<br />
ISSUE 19 | November 2005 | ISSN 1465-3842<br />
Agentlink Technical Forum III<br />
Budapest, Hungary, September 15-17, 2005<br />
Towards a Standard Agent to Agent Argumentation Interchange Format (AIF)<br />
Agent-Oriented Software Engineering (AOSE)<br />
Coordinating Agent Standardisation Activities (CASA)<br />
Environments For Multiagent Systems (ENV)<br />
Multiagent Resource Allocation (MARA)<br />
Programming Multi-Agent Systems (PROMAS)<br />
Self-Organisation in Multi-Agent Systems (SELFORG)
November 2005<br />
<strong>AgentLink</strong> News CONTENTS<br />
Terry R. Payne<br />
University of Southampton<br />
<strong>AgentLink</strong> Co-coordinator<br />
Editor in Chief<br />
Rebecca Earl<br />
University of Southampton<br />
<strong>AgentLink</strong> Publications<br />
Administrator<br />
Serena Raffin<br />
University of Southampton<br />
<strong>AgentLink</strong> Publications<br />
and Web Coordinator<br />
Julian Padget<br />
University of Bath<br />
Event Reports Editor<br />
Agentlink Report<br />
Challenges for Agent-Based Computing, Michael Luck,<br />
Peter McBurney, Onn Shehory and Steve Willmott<br />
Features<br />
Responsive and Generative Music for Public Installations,<br />
Mark d’Inverno, John Eacott and Fredrik Olofsson<br />
Semantic Web Services with the Web Service Modeling<br />
Ontology (WSMO), John Domingue, Dieter Fensel<br />
and Dumitru Roman<br />
Trust: Challenges and Opportunities, Nathan Griffiths<br />
Agent Research Overview<br />
Why Argue Dialogue Types and Argumentation in<br />
Agent Interaction, Katie Atkinson and Trevor Bench-Capon<br />
Answer Set Programming and Agents, Thomas Either<br />
and Jürgen Dix<br />
AgentCities.ES: Spanish network of agent-based<br />
computing (2003-2005), Antonio Moreno<br />
Standards Reports<br />
Latest News from the Standardisation World, Monique<br />
Calisti<br />
Agentlink Technical Forum III<br />
Towards a Standard Agent to Agent Argumentation<br />
Interchange Format (AIF), Steven Willmott, John Fox,<br />
Dan Grecu, Simon Parsons, Iyad Rahwan, Chris Reed,<br />
Dave Robertson, Nicolas Maudet<br />
Agent-Oriented Software Engineering (AOSE), Massimo<br />
Cossentino and Juan Pavón<br />
Coordinating Agent Standardization Activities (CASA),<br />
Monique Calisti, Giovanni Rimassa, Omer Rana, Stefan<br />
Poslad, Terry Payne and James Odell<br />
Environments For Multiagent Systems (ENV),<br />
Danny Weyns and Tom Holvoet<br />
Multiagent Resource Allocation (MARA), Ulle Endriss<br />
Programming Multi-Agent Systems (PROMAS)<br />
Editorial<br />
3<br />
5<br />
12<br />
18<br />
20<br />
Mehdi Dastani and Jorge Gómez Sanz<br />
Self-Organisation in Multi-Agent Systems (SELFORG),<br />
Giovanna Di Marzo Serugendo, Marie-Pierre Gleizes,<br />
Anthony Karageorgos<br />
Event Reports<br />
Semantic Web Days Munich 6-7 October, 2005,<br />
Andrea Kulas and Alain Léger<br />
The 1 st International Workshop on Context for Web<br />
Services (CWS’05), Djamal Benslimane, Chirine Ghedira<br />
and Zakaria Maamar<br />
The 8 th Biennial Israeli Symposium on Foundations of<br />
Artificial Intelligence (BISFAI), Claudia V. Goldman<br />
E4MAS 2005: Environments for Multiagent Systems,<br />
Danny Weyns, Eric Platon and Fabien Michel<br />
International Workshop on Agent-Based Models for<br />
Economic Policy Design (ACEPOL05), Herbert Dawid<br />
The 2005 IEEE/WIC/ACM International Joint Conference<br />
on Web Intelligence and Intelligent Agent Technology<br />
(WI’05 and IAT’05), Pierre Morizet-Mahoudeaux<br />
3 rd Workshop on Agents Applied in Health Care, at<br />
the 19th International Joint Conference on Artificial<br />
Intelligence (IJCAI 2005), Antonio Moreno<br />
Agents in Space, Michael Fisher<br />
1 st International Workshop on Security and Trust<br />
Management (STM 2005), Sjouke Rauw<br />
Project Report<br />
Site Report<br />
Business Decision Making using Multi-Agent<br />
Systems, Nadia Yakounina<br />
Books<br />
Agent Events Calendar<br />
26<br />
34<br />
ATHENA: Advanced Technologies for interoperability<br />
of Heterogeneous Enterprise Networks and their<br />
Applications, Klaus Fischer<br />
36<br />
38<br />
39<br />
Chris van Aart<br />
Acklin B.V.<br />
Book Reviews Editor<br />
Dominic Greenwood<br />
Whitestein Technologies AG<br />
Features Editor<br />
Stefan Poslad<br />
Queen Mary University<br />
Industry Editor<br />
The development of Agent Technology has found a place in a variety of application domains, from the Semantic<br />
Web to The Grid, from Service-Oriented to Peer-to-Peer Computing. Given the advances and improvements in both<br />
hardware and software in recent years, the objectives of developers has evolved, from minimising the complexity<br />
and size of code to maximising the modularity and re-usability of code. As systems scale, the number of networked<br />
hosts continues to rise and the number of intelligent consumer devices increases, issues such as autonomicity<br />
and semantic interoperation are becoming ever more signifi cant factors in the way software systems should be<br />
developed.<br />
In this, the 2005 Christmas issue of the <strong>AgentLink</strong> Newsletter, we include an excerpt from the recently published<br />
<strong>AgentLink</strong> Roadmap, which examines the fi eld of Agent Technology and its current trends and key drivers. d’Inverno<br />
reports on a novel use of agents for interactive music spaces in public areas, whilst Griffi ths reviews many of the<br />
challenges in<strong>here</strong>nt in modelling trust amongst agent communities. The third feature introduces WSMO; an EUfunded<br />
effort for representing and modelling services within the Semantic Web.<br />
The Third <strong>AgentLink</strong> III Technical Forum Meeting, co-located with CEEMAS, attracted researchers from all over Europe<br />
to the beautiful city of Budapest to engage in three days of stimulating discussion on topics old and new. Five<br />
of the seven meetings followed on from earlier Technical Forum meetings held in Rome or Ljubjana, w<strong>here</strong>as one of<br />
the new meetings, CASA (Co-ordinating Agent Standardisation Activities), followed on from the 2004 Paris meeting<br />
on “Agent Technology and Standardisation Activities - The Role of <strong>AgentLink</strong>”, which was reported on in Issue 17.<br />
As the year closes, so will the doors of the <strong>AgentLink</strong> III offi ces. Whilst the future of <strong>AgentLink</strong> IV is still unknown,<br />
events such as the the Technical Forum meetings clearly indicate a promising and certain future for Agent-based<br />
research in Europe over the coming years. Its been a pleasure to work with the <strong>AgentLink</strong> team, and to collaborate<br />
with such a dynamic and insightful community, and so from the offi ces of <strong>AgentLink</strong> III, we wish you Seasons Greetings,<br />
and a fruitful 2006.<br />
Terry R. Payne<br />
<strong>AgentLink</strong> Coordinator<br />
University of Southampton
AGENTLINK REPORT<br />
Challenges for Agent-Based Computing<br />
Michael Luck<br />
University of Southampton<br />
United Kingdom<br />
mml@ecs.soton.ac.uk<br />
Peter McBurney<br />
University of Liverpool<br />
United Kingdom<br />
p.j.mcburney@csc.liv.ac.uk<br />
Onn Shehory<br />
IBM Research<br />
Israel<br />
onn@il.ibm.com<br />
Steven Willmott<br />
Universitat Politècnica de Catalunya (UPC)<br />
Spain<br />
steve@lsi.upc.edu<br />
Introduction<br />
In its brief history, computing has enjoyed<br />
several different metaphors for the notion of<br />
computation. From the time of Charles Babbage<br />
in the nineteenth century until the mid-<br />
1960s, most people thought of computation<br />
as calculation, or operations undertaken on<br />
numbers. With widespread digital storage and<br />
manipulation of non-numerical information<br />
from the 1960s onwards, computation was reconceptualised<br />
more generally as information<br />
processing, or operations on text, audio or<br />
video data. With the growth of the Internet<br />
and the World Wide Web over the last<br />
fifteen years, we have reached a position<br />
w<strong>here</strong> a new metaphor for computation<br />
is required: computation as interaction. In<br />
this metaphor, computing is something that<br />
happens by and through communication<br />
between computational entities. In the<br />
current radical re-conceptualisation of<br />
computing, the network is the computer, to<br />
coin a phrase.<br />
In this new metaphor, computing is an<br />
activity that is in<strong>here</strong>ntly social, rather than<br />
solitary, leading to new ways of conceiving,<br />
designing, developing and managing<br />
computational systems. One example<br />
of the influence of this viewpoint is the<br />
emerging model of software as a service, for<br />
example in service-oriented architectures.<br />
In this model, applications are no longer<br />
monolithic, functioning on one machine<br />
(for single user applications), or distributed<br />
applications managed by a single organisation<br />
(such as today’s intranet applications), but instead<br />
are societies of components.<br />
How should we exploit this new metaphor<br />
of computing as social activity, as interaction<br />
between independent and sometimes intelligent<br />
entities, adapting and co-evolving with one<br />
another The answer, many people believe, lies<br />
with agent technologies. An agent is a computer<br />
program capable of flexible and autonomous<br />
action in a dynamic environment, usually an<br />
environment containing other agents. In this<br />
abstraction, we have encapsulated autonomous<br />
and intelligent software entities, called agents,<br />
and we have demarcated the society in which<br />
they operate, a multi-agent system. Agent-based<br />
computing concerns the theoretical and practical<br />
working through of the details of this simple twolevel<br />
abstraction.<br />
Agent Technology: Computing as Interaction<br />
Technologies<br />
Trends and Drivers<br />
Related Disciplines<br />
Related Techniques<br />
Mathematical<br />
Modelling<br />
Logic<br />
Programming<br />
User<br />
Interaction<br />
Design<br />
ormal<br />
Methods<br />
Grid<br />
Computing<br />
Self<br />
Systems<br />
Ambient<br />
Intelligence<br />
Uncertainty<br />
in AI<br />
Robotics<br />
Artificial Life<br />
Programming<br />
Languages<br />
Simulation<br />
Organisations<br />
Comple Systems<br />
PeertoPeer<br />
Computing<br />
A Roadmap for Agent Based Computing<br />
Biology<br />
Economics<br />
Game Theory<br />
Philosophy<br />
Compiled, written and edited by<br />
Michael Luck, Peter McBurney, Onn Shehory, Steve Willmott and the <strong>AgentLink</strong> Community<br />
In the sense that it is a new paradigm, agentbased<br />
computing is disruptive. As outlined<br />
above, it causes a re-evaluation of the very nature<br />
of computing, computation and computational<br />
systems, through concepts such as autonomy,<br />
coalitions and ecosystems, which make no sense<br />
to earlier paradigms. Economic historians have<br />
witnessed such disruption with new technologies<br />
repeatedly, as new technologies are created, are<br />
adopted, and then mature.<br />
Logic<br />
Anthropology<br />
Sociology<br />
Service Oriented<br />
Computing<br />
Organisation Design<br />
Political Science<br />
Decision<br />
Theory<br />
Trust and Reputation<br />
Coordination egotiation Communication<br />
Softare<br />
Engineering<br />
Interoperability<br />
Infrastructure<br />
Marketing<br />
Reasoning<br />
and Learning<br />
Semantic Web<br />
The concept of an agent has found currency in<br />
a diverse range of sub-disciplines of information<br />
technology, including computer networks,<br />
software engineering, artificial intelligence,<br />
human-computer interaction, distributed and<br />
concurrent systems, mobile systems, telematics,<br />
computer-supported cooperative work, control<br />
systems, decision support, information retrieval<br />
and management, and electronic commerce.<br />
In practical developments, web services, for<br />
example, now offer fundamentally new ways of<br />
doing business through a set of standardised<br />
tools, and support a service-oriented view<br />
of distinct and independent software<br />
components interacting to provide valuable<br />
functionality.<br />
In the context of such developments, agent<br />
technologies have increasingly come to the<br />
foreground. Because of its horizontal nature,<br />
it is likely that the successful adoption of<br />
agent technology will have a profound, longterm<br />
impact both on the competitiveness<br />
and viability of IT industries, and on the<br />
way in which future computer systems will<br />
be conceptualised and implemented.<br />
Challenges<br />
Hardware and software have improved<br />
significantly in performance and availability<br />
over the six decades of modern computing.<br />
As these changes have occurred, the<br />
objectives of programmers have also<br />
changed. Initially, most programmers<br />
sought to minimise memory usage and to<br />
maximise throughput or processing speeds in<br />
their applications. With increasing availability<br />
and lower costs of memory, and increasing microprocessor<br />
speeds, these objectives became far less<br />
important. Instead, by the 1970s and 1980s, the<br />
object-oriented paradigm sought to maximise<br />
the modularity and re-usability of code, and to<br />
minimise post-deployment system maintenance.<br />
However, these objectives too have become dated.<br />
Partly, this is because the development of proven<br />
November 2005|<strong>AgentLink</strong> News 19<br />
3
AGENTLINK REPORT<br />
OO methods and support tools has enabled the<br />
objectives to be readily achieved, and indeed,<br />
taken for granted, over the last two decades. More<br />
importantly, however, the rise to prominence of<br />
the Internet has led to a new understanding of the<br />
nature of computation, an understanding that<br />
puts interaction at its centre. In this context, the<br />
agent-oriented paradigm has sought to maximise<br />
adaptability and robustness of systems in open<br />
environments.<br />
It is <strong>here</strong> that one can see how a new technology<br />
may be a disruptive force. By tackling a different<br />
set of objectives, agent technologies address<br />
different problems and different applications than<br />
do object technologies. It is not simply that the<br />
rules of the game have changed, but rather that<br />
a different game is being played. In a world of<br />
millions of independent processors interconnected<br />
via the Internet and, through it, engaged in<br />
distributed cognition, a software design team can<br />
no longer assume that software components will<br />
share the same goals or motivations, or that the<br />
system objectives will remain static over time.<br />
Systems t<strong>here</strong>fore need to be able to adapt to<br />
dynamic environments, to be able to configure,<br />
manage and maintain themselves, and to cope<br />
with malicious, whimsical or just plain buggy<br />
components. The power of the agent paradigm is<br />
that it provides the means, at the appropriate level<br />
of abstraction, to conceive, design and manage<br />
such systems.<br />
Broad Challenges<br />
For agent-based computing to support the visions<br />
of the Semantic Web, ambient intelligence, the<br />
Grid, autonomic systems, etc, considerable<br />
challenges remain, both broad, over-arching<br />
challenges across the entire domain of agent<br />
technologies, and challenges specific to particular<br />
aspects. The broad challenges are as follows.<br />
• Creating tools, techniques and methodologies<br />
to support agent systems developers.<br />
Compared to more mature technologies<br />
such as object-oriented programming, agent<br />
developers lack sophisticated software tools,<br />
techniques and methodologies to support the<br />
specification, development and management<br />
of agent systems.<br />
• Automating the specification, development<br />
and management of agent systems. Agent<br />
systems and many of their features are still<br />
mostly hand-crafted. For example, the design<br />
of auction mechanisms awaits automation, as<br />
does the creation and management of agent<br />
coalitions and virtual organisations. These<br />
challenges are probably several decades from<br />
achievement, and will draw on domainspecific<br />
expertise (for example, economics,<br />
social psychology and artificial intelligence).<br />
• Integrating components and features. Many<br />
different theories, technologies and infrastructures<br />
are required to specify, design, implement and<br />
manage agent systems. Integrating these pieces<br />
co<strong>here</strong>ntly and cost-effectively is usually a major<br />
undertaking in any system development activity,<br />
a task made more challenging by the absence of<br />
mature integration tools and methodologies.<br />
• Establishing appropriate trade-offs between<br />
adaptability and predictability. Creating<br />
systems able to adapt themselves to changing<br />
environments, and to cope with autonomous<br />
components, may well lead to systems<br />
exhibiting properties that were not predicted<br />
or desired. Striking a balance, appropriate<br />
to the specific application domain, between<br />
adaptability and predictability is a major<br />
challenge, as yet unresolved either theoretically<br />
or practically. Associated with predictability<br />
is the requirement for practical methods and<br />
tools for verification of system properties,<br />
particularly in multi-agent systems that are<br />
likely to exhibit emergent behaviour.<br />
• Establishing appropriate linkage with other<br />
branches of computer science and with other<br />
disciplines, such as economics, sociology<br />
and biology. One task <strong>here</strong> is to draw<br />
appropriately on prior research from these<br />
other areas and disciplines. Another task is to<br />
avoid reinvention of existing techniques and<br />
methods, whether by agent researchers or by<br />
others. Awareness-building between areas<br />
and disciplines, and coordination of research<br />
and development activities, are essential if the<br />
appropriate linkages are to established and<br />
maintained.<br />
Conclusions<br />
Agent technologies can be distinguished from<br />
other programming technologies on the basis of<br />
their differing objectives. For agent technologies,<br />
the objectives are to create systems situated in<br />
dynamic and open environments, able to adapt to<br />
these environments and capable of incorporating<br />
autonomous and self-interested components.<br />
How quickly agent technology is adopted by<br />
software developers, t<strong>here</strong>fore, will depend at least<br />
partly on how many application domains require<br />
systems with these characteristics. Considering<br />
the domains receiving attention from agent<br />
software development companies such as Agentis,<br />
Magenta, Lost Wax or Whitestein (among<br />
others), the main areas are currently: logistics,<br />
transportation, utility management and defence.<br />
Common to many of these domains are multiple<br />
stakeholders or organisations linked in a network,<br />
such as a supply-chain, and with mission-critical,<br />
real-time processing requirements. In other<br />
words, t<strong>here</strong> are both functional and technical<br />
requirements for these applications, a divide that<br />
agent technologies are able to bridge.<br />
In application terms, we are already seeing the<br />
deployment of agent-like systems (in the areas<br />
of pervasive computing, the Semantic Web,<br />
P2P networks, and so on). In the longer term,<br />
we expect to see the industrial development<br />
of infrastructures for building highly scalable<br />
applications comprising pre-existing agents<br />
that must be organised or orchestrated.<br />
However, making the transition from research<br />
laboratory to deployed industrial applications<br />
is indeed a challenge, and it will be important<br />
to make scientifically sound business cases for<br />
implementations and descriptions that work as<br />
stimulators both for industry adoption and for<br />
further research.<br />
For commercial and industrial systems, agent<br />
technologies must emerge from the laboratory<br />
with a focus on business issues, on quality and<br />
on convergence with existing and emerging<br />
industrial technologies rather than innovation.<br />
Here, safety, reliability and traditional software<br />
quality measures are equally important, and<br />
must all be addressed to achieve wider adoption.<br />
In particular, we need agent solutions for<br />
distributed, enterprise-wide environments with<br />
exacting development requirements. This might<br />
be achieved through transition approaches by<br />
which existing systems can be upgraded with<br />
a successively increased agent presence in a<br />
seamless fashion. Wrapping legacy systems<br />
within autonomous agents situated in a larger<br />
multi-agent system is one approach that is being<br />
tried, for example, in connecting new and old<br />
telecommunications switches together seamlessly,<br />
allowing legacy switches to be gradually replaced<br />
without major disruption to the overall system.<br />
More generally, the adoption of agent<br />
technologies in business environments depends<br />
on how fast and how well agent technologies can<br />
be linked to existing and proven software and<br />
software methods. Agent technologies should be<br />
targeted at those application domains to which<br />
they are best suited, augmenting traditional<br />
techniques that should be used when agents<br />
are not applicable or appropriate. Ultimately,<br />
achieving this aim requires a commitment on the<br />
part of both business and research communities<br />
to collaborate effectively in support of more<br />
effective solutions for all. Such a dialogue is<br />
already underway.<br />
This article is based on extracts from the <strong>AgentLink</strong> III Roadmap, “Agent Technology: Computing as Interaction”,<br />
available from www.agentlink.org/roadmap<br />
4<br />
November 2005|<strong>AgentLink</strong> News 19
ALSO IN THIS SECTION...<br />
Semantic Web Services with the Web Service Modeling Ontology (WSMO)<br />
Trust: Challenges and Opportunities<br />
FEATURE<br />
Responsive and Generative Music for Public Installations<br />
Mark d’Inverno, John Eacott and Fredrik Olofsson<br />
University of Westminster<br />
UK<br />
M.dInverno@westminster.ac.uk<br />
john@informal.org<br />
f@fredrikolofsson.com<br />
Introduction<br />
In a recent poll conducted by Time Magazine,<br />
Muzak was voted one of the worst ideas of the<br />
last century. To give some indication of just how<br />
bad, it was only one place below the marriage<br />
of Lisa Marie Presley to Michael Jackson. In<br />
another poll conducted by Mori, 17% of those<br />
questioned voted Muzak as the thing they most<br />
detest about modern life. It was considered a<br />
raise pollution from which t<strong>here</strong> is no escape.<br />
However, with developments in algorithmic and<br />
generative music, and design techniques from<br />
intelligent agent technology we are investigating<br />
new possibilities for creating music generating<br />
devices that can respond flexibly to the requests<br />
of groups of users. And this is more than just an<br />
on/off button or graphic equalizer; it can provide<br />
users with the power to turn off tracks, change<br />
the density of the drums, transpose or invert the<br />
melody, re-harmonise,<br />
add effects, and more.<br />
In fact, the opportunity<br />
is t<strong>here</strong> to build a whole<br />
new musical sound<br />
system that can, in<br />
theory, respond to our<br />
every musical desire.<br />
In this article we will outline the design and<br />
implementation of an environmental sound<br />
installation that is able to respond to the requests<br />
of users interacting together. If you want to<br />
hear some of the music it generates and see it in<br />
action to help decide whether this article might<br />
be interesting enough to continue reading, then<br />
check out this movie: http://users.wmin.ac.uk/<br />
~dinverm/istreet/shortistreet.mov.<br />
The output from the installation is generated<br />
solely using algorithmic processes written by<br />
some of Europe’s leading composers of generative<br />
music, using the SuperCollider language. Because<br />
this music is represented solely (or very nearly<br />
solely) in terms of algorithmic structures, it<br />
allows for much more sophisticated and subtle<br />
operations to be performed on the musical<br />
performance in real time. In the installations we<br />
have implemented, musical output is controlled<br />
by the collective history of text commands from<br />
users’ mobile phones (by some distance the<br />
most common mode of communication at the<br />
University of Westminster). In this way, people in<br />
a shared environment can now directly, and more<br />
importantly, collectively influence their sonic<br />
environment and collaboratively create their<br />
mutually desired soundscape.<br />
The Installation<br />
Of course t<strong>here</strong> are countless mechanisms for<br />
providing environmental or ambient sound, but<br />
they are typically more closely entwined with the<br />
natural environment (consider the most obvious<br />
example of the wind chime, for example) and<br />
not under the direct, explicit control of situated<br />
users. It is now a real possibility to provide users<br />
with the ability to collectively or individually<br />
control their sound and music environment, and<br />
with sufficient flexibility to mitigate boredom. At<br />
the very least we wish for our work to highlight<br />
It is now a real possibility to provide users with<br />
the ability to collectively or individually control<br />
their sound and music environment.<br />
the possibilities users have to interact with their<br />
sonic environment such that Muzak will become<br />
simply unacceptable.<br />
In order to offer music which is deeply<br />
interactive, and that offers genuine scope for new<br />
and personalised forms that reflect the tastes of<br />
users situated within a common space, the sound<br />
must be generated in real time using algorithmic<br />
structures. Algorithmic music can be manipulated<br />
very explicitly. Not only is it a simple matter to<br />
reduce tempos or transpose particular parts the<br />
piece up or down a tone, it can be achieved much<br />
more effectively with algorithmic representations<br />
of music rather than sampled and recorded ones.<br />
Of course manipulating samples has made lots of<br />
people very rich, but samples are not as readily<br />
manipulated as code.<br />
Just to get a flavour, <strong>here</strong> is what a pretty decent<br />
sounding snare drum looks like.<br />
Instr([\istreet, \synthSN],<br />
{arg amp, pan, release, filter;<br />
var out, env;<br />
out= Resonz.ar(PinkNoise.ar,filter,0.5,2);<br />
env= Line.kr(1, 0, release, amp);<br />
Pan2.ar(out, pan, env); },<br />
#[\amp, \pan, \release, \filter] );<br />
Of course we wanted users to interact with the<br />
installation space in the most comfortable way<br />
and without inhibition. It didn’t take long to<br />
come up with the notion of mobile phone text<br />
messaging as an interaction metaphor. As long<br />
as the user commands used to interact with the<br />
software are kept simple and short, and problems<br />
with spelling, typos, case and so on are filtered out,<br />
it proves to be a successful means of interaction.<br />
This was also partly helped by Vodafone, who<br />
provided free handsets and SIM cards!<br />
The set of pre-defined user commands was visually<br />
projected into the space.<br />
In the future t<strong>here</strong> will<br />
be scope to build a<br />
mechanism that enables<br />
users to dynamically<br />
create a language for<br />
interaction with one<br />
another and with the installation, but this was<br />
simply beyond the scope of the first prototype.<br />
We also decided to use a mixture of words with<br />
different semantics, some with explicit semantics<br />
and some with ambiguous semantics, and also<br />
include both emotional and non-emotional<br />
words in order to investigate patterns of use. A<br />
team of evaluators was employed to report on the<br />
reaction and satisfaction of both participants and<br />
non-participants in the space.<br />
An additional motive for displaying user<br />
commands was that it led to more interaction<br />
between users and a collective sense of<br />
expectation. All text commands were queued<br />
in a buffer and executed simultaneously every<br />
16 bars. Once the commands were executed the<br />
buffer would be cleared and the process started<br />
again. Executed commands were displayed; the<br />
current output of the system being a function of<br />
that command history. Future commands were<br />
displayed separately with a timer that indicated<br />
the time left until execution, which helped build<br />
the sense of expectation.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
5
FEATURE<br />
The building blocks of the output came from<br />
original algorithmic and generative compositions<br />
commissioned from some of Europe’s leading<br />
composers in the field. In order to ‘fuse’ these<br />
styles in a meaningful way, the composers were<br />
asked to use certain constant parameters: that it<br />
be beat-based; structured into bars of 4-4 time;<br />
with 10 tracks each relating to either a type<br />
of instrument or a musical role; and that it be<br />
structured into 4 different 16 bar passages (that<br />
might, say, be mapped to intro, verse, bridge,<br />
chorus) with some nominal structure between<br />
the passages. These styles took the form of<br />
compositions written within a common style<br />
framework. It was absolutely necessary to have<br />
a common framework for writing pieces (styles)<br />
for this composition to ensure that intelligent,<br />
musically sensitive operations could be made at<br />
playback, especially when it came to combining<br />
these styles for performance.<br />
Global system parameters allow information<br />
concerning the current output to be stored<br />
such that decisions on how to alter the music<br />
according to some user-command can be made<br />
more effectively. These parameters affect the basic<br />
sound as defined by the styles. We can have as<br />
many global variables as we like in order to make<br />
intelligent decisions on how to interpret any new<br />
command. In the current system, these include<br />
tempo, key, reverb, delay, pitch effects (high and<br />
low), sampling rate, phaser effect, warp effect (a<br />
cut-up effect) and pan. Each of these variables has<br />
a predefined system default value.<br />
A track can be thought of in the very traditional<br />
sense of a track in a recording studio, or as in<br />
standard sequencing. Each style has exactly ten<br />
tracks. Moreover, we have pre-specified the role<br />
and function of each of these tracks and given<br />
them names as indicated below. We have also<br />
associated each track with a “level” paramenter<br />
which determines the “influence” a tack has on<br />
the overall sound. This information can be found<br />
in the table below.<br />
A blend is a listing of which tracks are currently<br />
associated with which styles, and t<strong>here</strong>by can be<br />
viewed a snapshot of what is heard. To gradually<br />
switch between different blends of styles we<br />
found a simple, effective and musically satisfying<br />
solution: Viewing the ten tracks as a stack, so that<br />
styles could be pushed around in discrete steps<br />
according to their levels. This retained a flavour<br />
of the current structure whilst introducing new<br />
elements from a different style.<br />
Examples of some user command call methods<br />
include simple operations such as muting or<br />
unmuting tracks, transposing tracks and changing<br />
tempo to adjust the complexity of the melodic or<br />
harmonic possibilities for a given tune or chord<br />
sequence as listed below:<br />
• Increase density un-mutes the highest-level<br />
muted tracks.<br />
• Decrease tempo reduces the beats per second<br />
by a given amount.<br />
• Transpose tracks transposes melodic tracks<br />
according to an array of order and amount.<br />
• Low pass applies a low-pass filter to rhythm<br />
tracks.<br />
• Melodic filter progressively reduces the set of<br />
allowable notes.<br />
In total, t<strong>here</strong> are twenty commands available<br />
with different functions. Eight of them<br />
correspond to styles composed in a certain genre,<br />
their use adding elements from these styles to<br />
the soundscape in the way described above. For<br />
example, the command “Dark” emphasizes the<br />
lower frequencies of the music, removing the<br />
higher frequencies and diminishing focus on the<br />
tempo to give the music a darker timbral quality.<br />
When commands such as Dark are repeated, the<br />
music is increasingly affected until an absolute<br />
pre-defined lower bound is reached.<br />
Final Thoughts<br />
We are interested in designing software<br />
architectures for sound that enables users<br />
to interact, create and compose their own<br />
sound landscapes. Our prototype architecture,<br />
implemented in several sites around the UK and<br />
Level Track Name Role<br />
5 1 Percussion specific genre effect<br />
4 2 Samplebeat sampled sound<br />
3 3 Bass drum typically every beat<br />
2 4 Snare off beat, patterns, fills<br />
1 5 High hat key in defining style<br />
1 6 Bass bass riff<br />
3 8 Pad ambient backing<br />
2 7 Chord harmonic sequence<br />
4 9 Melody2 generic melody<br />
5 10 Melody1 style related melody<br />
Sweden, has been extensively evaluated by a team<br />
of sociologists who found that the installations<br />
essentially became a playground for students to<br />
talk about music.<br />
T<strong>here</strong> are many potential avenues to explore new<br />
practical and theoretical possibilities including the<br />
user interface perspective (dynamically evolving<br />
language) and more sophisticated techniques<br />
drawn from AI, intelligent agents and generative<br />
music. We intend to investigate the mechanism<br />
by which users annotate their compositions in<br />
order that user-commands and system-methods<br />
can make more intelligent and musical use of the<br />
structures. This work also suggests new avenues<br />
of research exploring compositions consisting<br />
of not simply a single linear piece, but ideas<br />
and concepts that can be merged, tweaked and<br />
manipulated by users in real time to determine<br />
the actual performance.<br />
Some of the continuing project goals are as<br />
follows:<br />
• To investigate the relationship between<br />
algorithmic manipulation of music and sound<br />
and individual/social emotional response.<br />
• To map the space of possible mechanisms<br />
for users to exert more flexible control over a<br />
responsive sonic landscape.<br />
• To create social spaces w<strong>here</strong> users collectively<br />
compete, discuss and evaluate the aesthetics<br />
of the music and its responses.<br />
• To provide a forum for algorithmic<br />
composition and formulating challenges to<br />
existing notions of musical composition.<br />
• To develop the means for users to converse<br />
and interact with musical performances in<br />
real time.<br />
We have provided an infrastructure w<strong>here</strong> users<br />
can explore the possibilities of blending music,<br />
tracks and instruments, control the complexity of<br />
melodic or harmonic sequences, and in general,<br />
interact with the music for both aesthetic and<br />
educational purposes. Potential application<br />
environments for this project include nearly all<br />
shared spaces, from the shopping mall, to the<br />
office, to pubs and clubs, film and media. We are<br />
also planning an online version of this installation<br />
to allow users to “jam” with each other remotely<br />
using our system. This should be available in<br />
2006.<br />
Further Reading<br />
David Cope, Virtual Music: Computer Synthesis<br />
of Musical Style, MIT Press, 2001.<br />
John Eacott and Mark d’Inverno, ‘Embedded<br />
intelligent music or ihifi the intelligent hifi’,<br />
Digital Creativity, 14(2), 2003.<br />
James McCartney, SuperCollider audio synthesis<br />
environment, http://www.audiosynth.com, 2005.<br />
6<br />
November 2005|<strong>AgentLink</strong> News 19
FEATURE<br />
Semantic Web Services with the Web Service Modeling Ontology<br />
(WSMO)<br />
John Domingue, Open University, United Kingdom, j.b.domingue@open.ac.uk<br />
Dieter Fensel, DERI, Ireland and Austria, dieter.fensel@deri.org<br />
Dumitru Roman, DERI Innsbruck, Austria, dumitru.roman@deri.org<br />
In recent years, Web technologies have become<br />
very relevant for agent technologies. For the<br />
agent community, Web service and Semantic<br />
Web technologies have the potential to increase<br />
interoperability, to make use of Web standards,<br />
and provide a link between the agent paradigm<br />
and the newly emerging paradigm of Service<br />
Oriented Computing.<br />
In this context, the description of Web services in<br />
a machine-understandable fashion is expected to<br />
have a great impact in areas such as e-Commerce<br />
and Enterprise Application Integration through<br />
enabling dynamic and scalable cooperation<br />
between different systems and organizations. Web<br />
services provided by co-operating businesses or<br />
applications can be automatically located, based<br />
on another business or application needs. They<br />
can be composed to achieve more complex, valueadded<br />
functionalities, and cooperating businesses<br />
or applications can interoperate without prior<br />
agreements or custom codes. This leads to more<br />
flexible and cost-effective integration with the<br />
potential to create dynamic, scalable and costeffective<br />
marketplaces and eCommerce solutions.<br />
This has driven recent research efforts toward socalled<br />
Semantic Web services which enrich Web<br />
services with machine-processable semantics.<br />
The Semantic Web and Web services are<br />
envisioned as the enabling technologies for the<br />
next generation of Web applications. The former<br />
aims at enhancing the machine-readability of<br />
WSML WG<br />
A Rule-based Language for SWS<br />
WSMO WG<br />
A Formal Lenguage for WSMO<br />
Web content, with ontologies identified as the<br />
key technical building block. The objective of the<br />
latter is to enable distributed computation over<br />
the Internet by means of automated and dynamic<br />
discovery, composition, and execution of services,<br />
thus providing a new technology for Web-based<br />
systems engineering. The current Web service<br />
technology stack enables the exchange of messages<br />
between Web services (SOAP), describes the<br />
technical interface for consuming a Web service<br />
(WSDL), and supports the advertisement of Web<br />
services in registries (UDDI). However, these<br />
technologies do not include explicit descriptions<br />
of the functionality of a Web service. Moreover, the<br />
existing descriptions are represented syntactically<br />
and thus do not characterize the meaning of<br />
the information to be interchanged. Semantic<br />
Web services apply Semantic Web technology<br />
to Web services which raises the level of this<br />
discourse. More specifically, through the use of<br />
semantic description frameworks, Semantic Web<br />
services will support the provision of intelligent<br />
mechanisms for the discovery, composition,<br />
contracting, and execution of Web services.<br />
To this end, the Web Service Modeling Ontology<br />
(WSMO) provides the basis for Semantic Web<br />
services by specifying a fully fledged framework.<br />
This begins with a conceptual model, continues<br />
with a formal language to provide formal syntax<br />
and semantics for the conceptual model, and<br />
is completed with an execution environment<br />
that glues together the components that use<br />
WSMX WG<br />
A Conceptual Model for SWS<br />
An Execution Enviroment<br />
for WSMO<br />
Figure 1. The relationship between the WSMO, WSML and WSMX Working Groups, that form a framework for<br />
Semantic Web Services.<br />
the language to enable automation of services.<br />
The WSMO is developed in the context of<br />
the WSMO Working Group [1], as part of the<br />
SDK cluster [2]. Through alignment with key<br />
European research projects in the Semantic Web<br />
service area, the WSMO aims to further the<br />
development of Semantic Web services through<br />
work on a common architecture and platform,<br />
and further standardization in the area of Semantic<br />
Web service languages. As shown in Figure 1, the<br />
WSMO Working Group includes the WSML<br />
Working Group [3], which aims to develop the<br />
Web Service Modeling Language (WSML) that<br />
formalizes the WSMO, and the WSMX Working<br />
Group [4], which aims to provide an execution<br />
environment and reference implementation for<br />
the WSMO. Figure 1 depicts these relationships.<br />
WSMO Design Principles<br />
The WSMO provides ontological specifications<br />
for the core elements of Semantic Web services.<br />
In fact, Semantic Web services are an integrated<br />
technology for the next generation of the Web that<br />
combine Semantic Web technologies and Web<br />
services, t<strong>here</strong>by evolving the Internet from an<br />
information repository for human consumption<br />
into a world-wide system for distributed Webbased<br />
computing. T<strong>here</strong>fore, frameworks for<br />
Semantic Web services must integrate basic Web<br />
design principles, both those defined for the<br />
Semantic Web, and those defined for distributed,<br />
service-orientated computing over the Web. As<br />
a result, the WSMO is based on the following<br />
design principles:<br />
• Web Compliance<br />
The WSMO inherits the concept of URI<br />
(Universal Resource Identifier) for unique<br />
identification of resources as the essential<br />
design principle of the World Wide Web.<br />
Moreover, the WSMO adopts the concept<br />
of Namespaces for denoting consistent<br />
information spaces, supports XML and other<br />
W3C Web technology recommendations, and<br />
supports the decentralization of resources.<br />
• Ontology-Based<br />
Ontologies are used as the data model<br />
throughout the WSMO, implying that all<br />
resource descriptions and all data interchanged<br />
during service usage are based on ontologies.<br />
Ontologies are a widely accepted state-ofthe-art<br />
means of knowledge representation<br />
and have thus been identified as the central<br />
enabling technology of the Semantic Web.<br />
The extensive use of ontologies allows<br />
semantically enhanced information processing<br />
and support for interoperability. The WSMO<br />
November 2005|<strong>AgentLink</strong> News 19<br />
7
FEATURE<br />
supports the ontology languages defined for<br />
the Semantic Web.<br />
• Strict Decoupling<br />
This denotes that WSMO resources are<br />
defined in isolation, meaning that each<br />
resource is specified independently without<br />
regard to possible interactions with other<br />
resources. This complies with the open and<br />
distributed nature of the Web.<br />
• Centrality of Mediation<br />
As a complementary design principle to strict<br />
decoupling, centrality of mediation addresses<br />
the handling of heterogeneities that naturally<br />
arise in open environments. Heterogeneity<br />
can occur in terms of data, underlying<br />
ontology, protocol or process. The WSMO<br />
recognizes the importance of mediation for<br />
the successful deployment of Web services by<br />
making mediation a first class component of<br />
the framework.<br />
• Ontological Role Separation<br />
Users, or more generally clients, exist in<br />
specific contexts which will not always be an<br />
exact match with available Web services. For<br />
example, a user may wish to book a holiday<br />
according to preferences for weather, culture<br />
and childcare, w<strong>here</strong>as available Web services<br />
may only cover airline travel and hotel<br />
availability. The underlying epistemology of<br />
the WSMO differentiates between the desires<br />
of users or clients and the functionality of<br />
available services.<br />
• Description versus Implementation<br />
The WSMO differentiates between the<br />
descriptions of Semantic Web service elements<br />
(description) and executable technologies<br />
(implementation). While the former requires<br />
a concise and sound description framework<br />
based on appropriate formalisms to provide<br />
concise semantic descriptions, the latter is<br />
concerned with the support of existing and<br />
emerging execution technologies for the<br />
Semantic Web and Web services. The WSMO<br />
aims at providing an appropriate ontological<br />
description model, and complaince with<br />
ONTOLOGIES<br />
Figure 2. WSMO Top-level elements.<br />
GOALS<br />
MEDIATORS<br />
existing and emerging technologies.<br />
• Execution Semantics<br />
In order to verify the WSMO specification,<br />
the formal execution semantics of reference<br />
implementations such as WSMX, as well as<br />
other WSMO-enabled systems, provide the<br />
technical realization of the WSMO.<br />
WSMO Top-level Elements<br />
As depicted in Figure 2, the WSMO [1] consists<br />
of four main elements for semantically describing<br />
Semantic Web services: (1) ontologies that<br />
provide the concepts and relationships used<br />
by other elements, (2) goals that define the<br />
users’ objectives, i.e. the (potential) problems<br />
that should be solved by Web services, (3) Web<br />
service descriptions that define various aspects of<br />
a Web service, and (4) mediators which bypass<br />
interoperability problems.<br />
Ontologies provide the formal semantics for<br />
the terminology used within all other WSMO<br />
components. A set of non-functional properties<br />
are available for characterizing ontologies; they<br />
usually include the DC Metadata elements.<br />
Imported ontologies allow a modular approach<br />
to ontology design and can be used if no conflicts<br />
arise between the ontologies. When importing<br />
ontologies in realistic scenarios, some steps for<br />
aligning, merging and transforming them are<br />
needed in order to resolve ontology mismatches.<br />
For this reason, ontology mediators are used.<br />
Concepts constitute the basic elements of the<br />
agreed terminology for some problem domain.<br />
Relations are used to model interdependencies<br />
between several concepts (and respectively<br />
instances of these concepts); functions are special<br />
relations with a unary range and an n-ary domain<br />
(parameters inherited from relations), w<strong>here</strong> the<br />
range value is functionally dependent on the<br />
domain values, and instances are either defined<br />
explicitly or by a link to an instance store, i.e., an<br />
external storage of instances and their values.<br />
The WSMO Web service element provides the<br />
means to describe services through the following<br />
WEB SERVICES<br />
characteristics: The non-functional properties<br />
and imported ontologies for Web services play a<br />
role that is similar to that found in the ontology<br />
class with the minor addition of a quality-ofservice<br />
non-functional property. An extra type<br />
of mediator to deal with protocol and process<br />
related mismatches between Web services, is<br />
also included. The final two attributes of the<br />
definition are the two core WSMO notions for<br />
semantically describing Web services: a capability,<br />
which is a functional description of a Web service,<br />
describing constraints on the input and output<br />
of a service through preconditions, assumptions,<br />
postconditions, and effects; and service interfaces<br />
which specify how the service behaves in order<br />
to achieve its functionality. A service interface<br />
consists of a choreography which describes the<br />
interface for the client-service interaction required<br />
for service consumption, and an orchestration<br />
which describes how the functionality of a Web<br />
service is achieved by aggregating other Web<br />
services.<br />
A Goal as defined in the WSMO specifies the<br />
objective that a client has when consulting<br />
a Web service. In particular, aspects relating<br />
to user desires with respect to the requested<br />
functionality and behavior. Ontologies are used<br />
as the semantically defined terminology for goal<br />
specification. Goals model the user view in the<br />
Web service usage process and t<strong>here</strong>fore are a<br />
separate top level entity in the WSMO. They are<br />
mainly defined using the following properties:<br />
the requested capability in the definition of a<br />
goal represents the functionality of the services<br />
the user would like to have, and the requested<br />
interface represents the interface of the service<br />
the user would like to have and interact with.<br />
The concept of Mediator in the WSMO addresses<br />
the handling of heterogeneities occurring between<br />
elements that need to interoperate. This is achieved<br />
by resolving mismatches between different<br />
terminologies (data level), on the communicative<br />
behavior between services (protocol level), and on<br />
the business process level. A WSMO Mediator<br />
connects elements and provides mediation<br />
facilities for resolving such mismatches. The<br />
WSMO defines different types of mediators for<br />
connecting the distinct WSMO elements: OO<br />
Mediators connect and mediate heterogeneous<br />
ontologies, GG Mediators connect Goals, WG<br />
Mediators connect Web services to Goals, and<br />
WW Mediators connect interoperating Web<br />
services to resolve mismatches between them.<br />
Web Service Modeling Language<br />
(WSML)<br />
The Web service Modeling Language (WSML)<br />
[3] is a language for the description of ontologies,<br />
goals, Web services and mediators, based on the<br />
conceptual model of the WSMO. The WSML<br />
provides one co<strong>here</strong>nt framework which brings<br />
together Web technologies with different, wellknown<br />
logical language paradigms: Description<br />
8<br />
November 2005|<strong>AgentLink</strong> News 19
FEATURE<br />
Logics, Logic Programming, as well as F-Logic as<br />
starting points for the development of a number<br />
of WSML language variants: WSML-Core,<br />
WSML-DL, WSML-Flight, WSML-Rule, and<br />
WSML-Full. The WSML variants differ in logical<br />
expressiveness and in the underlying language<br />
paradigms. They allow users to make the tradeoff<br />
between the degree of expressiveness and the<br />
implied complexity on a per-application basis.<br />
The WSML has two alternative layerings:<br />
• WSML-Core → WSML-DL → WSML-<br />
Full, and<br />
• WSML-Core → WSML-Flight → WSML-<br />
Rule → WSML-Full.<br />
For both layerings, WSML-Core and WSML-<br />
Full denote the least and most expressive layers<br />
respectively. The two layerings are disjoint to a<br />
certain extent in the sense that inter-operation<br />
between the Description Logic variant (WSML-<br />
DL) and the Logic Programming variants<br />
(WSML-Flight and WSML-Rule) is only possible<br />
through a common core (WSML-Core) or<br />
through a very expressive superset (WSML-Full).<br />
The WSML, can be seen as a testing ground for<br />
the development of formal techniques for Web<br />
service description.<br />
Web Service Modeling Execution<br />
Environment (WSMX)<br />
The Web service Execution Environment<br />
(WSMX) [4] is an execution environment which<br />
enables discovery, selection, mediation, and<br />
invocation of Semantic Web services described<br />
according to the philosophy of WSMO. It is<br />
thus a WSMO reference implementation. The<br />
WSMX provides a tangible testbed for the<br />
WSMO in order to prove its viability as a means<br />
to achieve dynamic interoperability between<br />
Semantic Web services.<br />
In short, WSMX functionality can be summarized<br />
as performing discovery, mediation, selection<br />
and invocation of Web services on receiving a<br />
user goal specified in WSML. The user goal is<br />
first matched against the formal descriptions<br />
of Web services registered with the WSMX.<br />
If successful, one or more service descriptions<br />
(ranked according to user preference) can be<br />
returned. The most appropriate service as<br />
selected by the user is then invoked and the<br />
result returned to user. Prior to the invocation<br />
step, the WSMX ensures that the data provided<br />
for the service invocation is in the format that<br />
the Web service expects. If necessary, a data<br />
mediation process is performed to ensure interoperability<br />
between different entities. Presently,<br />
the WSMX architecture relies on a set of<br />
loosely-coupled main components that provide<br />
functionality for each step of the Web service<br />
usage process: discovery, selection, mediation<br />
and invocation.<br />
Final Remarks<br />
Semantic Web services, by combining recent<br />
Web-related trends, constitute one of the<br />
most promising research directions to improve<br />
the integration of applications within and<br />
across enterprise boundaries. In this context,<br />
the WSMO aims to provide the conceptual<br />
and technical means to realize Semantic Web<br />
services, improving the cost-effectiveness,<br />
scalability and robustness of current solutions.<br />
The WSML provides a formal syntax and<br />
semantics for the WSMO by offering variants<br />
based on different logics in order to provide<br />
several levels of logical expressiveness and thus<br />
allowing tradeoffs between expressivity and<br />
computability). Finally, the WSMX provides a<br />
reference implementation for the WSMO and<br />
the interoperation of Semantic Web services.<br />
References<br />
[1] http://www.wsmo.org/<br />
[2] http://www.sdk-cluster.org/<br />
[3] http://www.wsmo.org/wsml<br />
[4] http://www.wsmx.org/<br />
Trust: Challenges and Opportunities<br />
Nathan Griffiths<br />
University of Warwick<br />
United Kingdom<br />
nathan@dcs.warwick.ac.uk<br />
Introduction<br />
Trust is fundamental in distributed systems<br />
w<strong>here</strong> individual components interact to achieve<br />
some overall objective. In small-scale or closed<br />
systems this trust can be implicit, imbued to the<br />
individual components and the system overall by<br />
its designers and implementers. In open or largescale<br />
systems however, it is becoming increasingly<br />
common for trust to be explicitly represented and<br />
reasoned about by the components, or agents, in<br />
the system. In recent years trust has become a hot<br />
research topic, with numerous conferences and<br />
workshops attracting both academic researchers<br />
and industrial representatives from solutions<br />
providers in areas as diverse as telecoms,<br />
logistics and e-business. This article gives a brief<br />
overview of the alternative approaches to trust<br />
and attempts to identify some of the important<br />
research questions.<br />
The current interest in trust creates an<br />
active environment for trust researchers and<br />
practitioners. However, it also raises some<br />
challenges. Trust research has parallels with<br />
agent research a decade ago — it is an exciting<br />
area of clear value, but t<strong>here</strong> is a risk that debate<br />
about definitions and mechanisms might add<br />
confusion and delay widespread adoption. Just as<br />
t<strong>here</strong> was (is) no consensus definition of agents,<br />
t<strong>here</strong> is similar debate over trust. For example,<br />
how does trust relate to reputation Is trust an<br />
individual (experience-based) notion, or should it<br />
encompass others’ (potentially subjective) views<br />
Are trusted agents secure and reliable, or do they<br />
simply have “good” intentions These questions<br />
are important, but it is equally important to<br />
ensure that confusion is avoided, and that as a<br />
community we have a clear overall view.<br />
Psychological and Cognitive<br />
Approaches<br />
Many trust models take a cognitive view of<br />
agents and trust, typically relying on folk<br />
psychology notions such as belief and desire.<br />
Agents trust others with respect to some activity<br />
or the performance of some task, and consider<br />
trustworthiness according to beliefs about such<br />
aspects as competence, disposition, willingness,<br />
dependence, and fulfilment [1]. The level of<br />
trust is determined by these beliefs, along with<br />
past experiences and possible recommendations<br />
from others. Some cognitive approaches also<br />
consider modelling the desires/motivations<br />
of other agents, and incorporate this into<br />
assessing trustworthiness. These approaches<br />
give a powerful mechanism for reasoning about<br />
interactions and the trust, power and dependence<br />
relationships between agents. However, it can be<br />
difficult to translate their richness into a practical<br />
implemented system, since the data structures<br />
and the functions needed to manipulate them<br />
are expensive to maintain.<br />
Numerical Approaches<br />
Numerical approaches are perhaps the most<br />
November 2005|<strong>AgentLink</strong> News 19<br />
9
FEATURE<br />
commonly used with numerous proposed<br />
mechanisms, in which agents represent the<br />
trustworthiness of others in numeric intervals,<br />
typically [-1, +1] or [0, 1]. The lower bound<br />
corresponds to complete distrust and the upper<br />
bound to blind trust. Agents either keep track of a<br />
numerical value that is updated by some function<br />
after each interaction, or are equipped with a<br />
function to transform a history of interactions<br />
into a numerical value. Some approaches<br />
decompose trust into separate values for different<br />
situations, or according to<br />
the different dimensions<br />
of an interaction (such as<br />
cost, quality, timeliness).<br />
Numerical methods tend<br />
to use a trust threshold<br />
and only when trust is<br />
above the threshold will<br />
cooperation take place.<br />
The main advantage<br />
of numerical approaches is their simplicity,<br />
since it is relatively inexpensive to incorporate<br />
trust into an agent’s decision making. This<br />
simplicity, however, is also their disadvantage.<br />
Firstly, numerical approaches do not provide<br />
the richness of reasoning available with other<br />
techniques. Secondly, t<strong>here</strong> is limited meaning<br />
to the values themselves, which encumbers the<br />
sharing of information between agents and<br />
external reasoning about the system.<br />
Probabilistic Methods<br />
Probabilistic methods are a subset of numerical<br />
approaches in which trust is represented in the<br />
interval [0,1]. However, this number represents a<br />
probability and has a clearer semantics associated<br />
with it. T<strong>here</strong> are many probabilistic approaches<br />
ranging from those based on simple objective<br />
and subjective probabilities, to those using more<br />
complex Bayesian probability distributions.<br />
Decisions are made in a similar manner to<br />
numerical approaches by use of trust thresholds<br />
Socially-oriented<br />
and maximising the probability of success.<br />
Reputation Systems<br />
Many reputation-based approaches have been<br />
proposed, ranging from centralised systems<br />
that aggregate feedback (à la eBay), through<br />
decentralised feedback systems, to numerical and<br />
probabilistic approaches that are augmented with<br />
recommendations from other agents. Reputation<br />
systems generally use a combination of direct<br />
experience, recommendations, and knowledge<br />
Trust research has parallels with agent research a decade ago — it<br />
is an exciting area of clear value, but t<strong>here</strong> is a risk that debate<br />
about definitions and mechanisms might add confusion and delay<br />
widespread adoption. Just as t<strong>here</strong> was (is) no consensus definition<br />
of agents, t<strong>here</strong> is similar debate over trust.<br />
Service-oriented<br />
of the social structure of the system to represent<br />
and reason about trust. The incorporation of<br />
reputation greatly enhances the richness of a<br />
trust model, but typically increases complexity,<br />
and opens up questions about issues such as<br />
collusion.<br />
Certificates and Keys<br />
W<strong>here</strong> trust is viewed as a mechanism for<br />
ensuring security, it tends to be achieved via<br />
protocols, certificates or keys. Some approaches<br />
define detailed interaction protocols that ensure<br />
detection of any deviation from expected<br />
behaviour, and define the actions or sanctions<br />
that should be taken in such cases. Other<br />
approaches use trusted third parties to provide<br />
verification and authentication, and to act as<br />
intermediaries in interactions. However, the<br />
most common security-oriented approach is to<br />
use certificates and keys: a certification authority<br />
issues a certificate verifying that an agent’s public<br />
key is owned by that agent. These public keys can<br />
Security-oriented<br />
then be used to sign and encrypt data to ensure<br />
authentication and privacy. Certificates and keys<br />
provide a powerful mechanism for achieving<br />
security, but it can be difficult to combine them<br />
with more general social- or service-oriented<br />
approaches.<br />
Challenges for Trust Researchers<br />
Computational trust is a young and active<br />
research area and numerous techniques have<br />
been proposed, as introduced above. However,<br />
t<strong>here</strong> are a number of open<br />
research challenges. Since<br />
t<strong>here</strong> is no overarching<br />
view of trust it is difficult<br />
for implementers to select<br />
a trust mechanism for a<br />
given environment. T<strong>here</strong><br />
is a need for researchers to<br />
frame their work within<br />
the context of the general<br />
trust landscape to enable simple comparisons<br />
between approaches. A trust taxonomy would<br />
begin to address this problem. Furthermore, trust<br />
should be seen as a fundamental component of<br />
any multi-agent system. Again, a taxonomy of<br />
trust would assist in enabling this, especially<br />
if aided by the provision of suitable software<br />
tools to assist in incorporating trust into<br />
agents. Finally, much trust research takes place<br />
in simulated semi-closed environments, and a<br />
number of issues must be addressed to enable<br />
trust mechanisms to be effective in real-world<br />
domains. We t<strong>here</strong>fore propose three primary<br />
challenges for trust researchers.<br />
A Trust Taxonomy A taxonomy of trust would be<br />
valuable, firstly to aid implementers in selecting<br />
appropriate trust techniques for a particular<br />
context and, secondly, to assist researchers in<br />
positioning their work, and comparing it to<br />
other approaches. The division of trust literature<br />
into socially-, service- and security-oriented<br />
trust might a starting point, but perhaps what<br />
should be aimed for is a pattern library of trust<br />
techniques. Ideally, an implementer should be<br />
able to select a trust approach easily, based on<br />
the characteristics of the domain.<br />
logic economic certification<br />
numerical/threshold-based<br />
PKI<br />
probabilistic<br />
philosophical<br />
protocols/policies<br />
norms<br />
psychological<br />
Figure 1. Selected techniques and mechanisms for trust.<br />
Agent Platforms Trust is fundamental to multiagent<br />
systems. However, many existing agent<br />
platforms do not incorporate trust by default. Trust<br />
should be viewed as a fundamental component<br />
of any agent platform, and implementers should<br />
be able to select an appropriate “pattern” for a<br />
given domain. Existing platforms need to be<br />
augmented with “trust wrappers”, while new<br />
platforms should include placeholders for trust<br />
mechanisms by default.<br />
Coping with the “Real-World” The real-world<br />
is a complex place, with interactions potentially<br />
failing in numerous ways for a variety of reasons,<br />
including malicious motivations of cooperative<br />
partners, competition between agents, and<br />
10<br />
November 2005|<strong>AgentLink</strong> News 19
FEATURE<br />
unavoidable environmental change. Current<br />
trust models are typically relatively simplistic<br />
in updating trust after interactions. If trust is to<br />
be effectively applied in real-world applications,<br />
then t<strong>here</strong> must be a mechanism to distinguish<br />
between intentional and unintentional failure.<br />
Furthermore, as Marsh proposes, t<strong>here</strong> must<br />
be a distinction between trust, mistrust and<br />
distrust, i.e. trust, misplaced and incorrect<br />
trust, and explicit distrust (c.f. information,<br />
misinformation and disinformation) [2].<br />
Summary<br />
Trust is a rich notion and an area of active<br />
research. In this article we have tried to give a<br />
flavour of the breadth of trust research. As in any<br />
young research area t<strong>here</strong> are a number of open<br />
challenges; this article has presented a personal<br />
view in identifying three areas that the author<br />
sees as meriting immediate attention. Clearly,<br />
t<strong>here</strong> are numerous other open questions, but<br />
the author hopes that a focus on these challenges<br />
will help facilitate more widespread use of trust<br />
in agent-based systems.<br />
[1] C. Castelfranchi. Trust mediation in<br />
knowledge management and sharing. In C.<br />
Jensen, S. Poslad, and T. Dimitrakos, editors,<br />
Proceedings of the Second International<br />
Conference on Trust Management (iTrust<br />
2004), pages 304–318, 2004.<br />
[2] S. Marsh and M. R. Dibben. Trust, untrust,<br />
distrust and mistrust — an exploration of<br />
the dark(er) side. In P. Herrman, V. Issarny,<br />
and S. Shiu, editors, Proceedings of the<br />
Third International Conference on Trust<br />
Management (iTrust 2005), pages 17–33.<br />
Springer-Verlag, 2005.<br />
[3] S. D. Ramchurn, D. Huynh, and N. R.<br />
Jennings. Trust in multi-agent systems.<br />
Knowledge Engineering Review, 19(1):1–<br />
25, 2004.<br />
Diversity of Trust Research<br />
T<strong>here</strong> is a growing corpus of literature on trust, a detailed overview of which can be found in<br />
the further reading identified below. T<strong>here</strong> is no overarching taxonomy of trust research, and<br />
the wide applicability of trust gives rise to a wide range of approaches. However, we can<br />
broadly divide these approaches into three areas.<br />
Socially-Oriented Trust - Typically influenced by social science, psychology or philosophy,<br />
socially-oriented trust is viewed as a social notion for modelling and reasoning about the<br />
relationships between agents. Socially-oriented trust often considers issues such as the<br />
motivations of agents, and the power and dependence relationships between them.<br />
Service-Oriented Trust - Taking a pragmatic view, service-oriented trust is a mechanism<br />
for achieving, maintaining, and reasoning about quality of services and interactions. Agents<br />
typically maintain their own trust information about others, possibly incorporating the<br />
recommendations of others, and use this to inform their decision making processes.<br />
Security-Oriented Trust - Taking the view of trust as a mechanism for ensuring security,<br />
encompassing issues of authentication, authorisation, access control, privacy, etc. Securityoriented<br />
trust also includes work on “trusted computing”, i.e. building trusted platforms to<br />
ensure privacy and security.<br />
In each of these areas a range of techniques and mechanisms have been proposed, drawing<br />
on work in areas as diverse as logic and the social sciences. For any given situation t<strong>here</strong><br />
are generally several alternative candidate trust models/techniques for an implementer to<br />
choose from, as illustrated in Figure 1. (Note that this figure is not intended to be exhaustive,<br />
and many other techniques exist.)<br />
Trust literature can be further divided according to whether it is concerned with individual- or<br />
system-level trust. In the former, individual agents model and reason about others, while in<br />
the latter agents are forced to be trustworthy by externally imposed regulatory protocols,<br />
and mechanisms [3]. Additionally, some trust models are centralised and have a single<br />
repository of information, while others are decentralised with individuals maintaining their<br />
own information. Finally, we can distinguish between models concerned with direct trust<br />
w<strong>here</strong> agents trust others directly based on their experiences, and recommendation trust<br />
w<strong>here</strong> trust is based on the recommendations of others. For each of these categories of trust<br />
t<strong>here</strong> are a number of alternative approaches discussed in the literature, the most common<br />
of which are also briefly introduced.<br />
Benchmark simulations for Multi-Agent Learning<br />
Maarten van Someren (University of Amsterdam)<br />
maarten@science.uva.nl<br />
Evaluating methods and systems on publicly available datasets has proved to be a succesful methodology<br />
in Machine Learning and in Information Retrieval. Evaluating multi-agent learning systems requires<br />
simulation environments, even though this can be more difficult to achieve than for the other areas (e.g.<br />
the traffic light simulation, see http://sourceforge.net/projects/stoplicht).<br />
This effort may become part of the new network of excellence KDUbiq which will start in December or<br />
January. If you are interested then please contact me by email (maarten@science.uva.nl).<br />
November 2005|<strong>AgentLink</strong> News 19<br />
11
AGENT RESEARCH OVERVIEW<br />
ALSO IN THIS SECTION...<br />
Answer Set Programming and Agents<br />
AgentCities.ES:<br />
Spanish network of agent-based computing<br />
Why Argue Dialogue Types and<br />
Argumentation in Agent Interaction<br />
Katie Atkinson and Trevor Bench-Capon<br />
University of Liverpool<br />
Liverpool<br />
UK<br />
{katie,tbc}@csc.liv.ac.uk<br />
Agents interact not only with their environment,<br />
but also with one another. Like human agents,<br />
software agents effect this interaction through<br />
conversation and dialogue. Inter-agent dialogues<br />
are typically founded on the theory of speech<br />
acts, as developed by Searle [3]. Different speech<br />
acts give different force to the same content.<br />
Thus ask(“Paris is in France”) and tell(“Paris is<br />
in France”) are two speech acts with the same<br />
content, but with different forces, one seeking<br />
and the other providing information. Agent<br />
interactions, however, require more than one-shot<br />
utterances, and may require the exchange of a<br />
complex series of speech acts. Such dialogues need<br />
to be sustained and co<strong>here</strong>nt. This can be effected<br />
through the use of conversation classes (see [1] for<br />
an early example), or protocols which determine<br />
which speech acts can be used to initiate a<br />
particular kind of conversation, and which speech<br />
acts can be<br />
used at various<br />
points in<br />
conversations<br />
of this sort.<br />
This technique<br />
certainly provides a means of co-ordinating<br />
and structuring dialogues between agents. It is,<br />
however, important to recognise that the speech<br />
acts change their meanings according to the type<br />
of dialogue in which they appear. T<strong>here</strong> is a strong<br />
pragmatic element to the meaning of speech acts,<br />
and it is the type of dialogue which provides<br />
the context which determines this pragmatic<br />
element. Thus if agent A poses a question to<br />
agent B, we usually expect the following to be<br />
true: Agent A does not know the answer, but<br />
wants to know and furthermore expects agent B<br />
to know and provide the answer. But in a quiz<br />
or an oral examination, the questioner will know<br />
the answer and wishes to discover whether the<br />
agent asked also knows it. Similarly if an agent<br />
states information that it does not know, and<br />
perhaps does not even believe, whether it be a lie<br />
or a legitimate guess will depend on the dialogue<br />
of which the assertion is a part. For this reason<br />
discourse theorists e.g. [2] and informal logicians,<br />
such as Walton and Krabbe [3], have seen the<br />
importance of identifying and characterising<br />
types of dialogue so that the pragmatic elements<br />
of speech acts can be understood.<br />
Walton and Krabbe’s typology of dialogues<br />
in Commitment in Dialogue, is probably the<br />
most influential classification. They identify six<br />
dialogue types: persuasion, negotiation, inquiry,<br />
information seeking, deliberation and eristic.<br />
The last of these involves venting grievances<br />
or quarrelling such that dialogue substitutes<br />
for physical confrontation. This type seems of<br />
little application to agent systems so we will not<br />
consider it further <strong>here</strong>. The remaining five are all<br />
plausibly useful in agent interactions.<br />
Dialogue types are characterised by the initial<br />
positions of the participants, the main goal of the<br />
dialogue and the individual aims of the participants.<br />
Persuasion starts from a disagreement with the<br />
aim of bringing both parties into agreement, with<br />
each party trying to encourage the other to change<br />
their point of view. Negotiation starts from an<br />
initial allocation of resources or tasks, and aims<br />
to move to a better allocation, with each party<br />
trying to achieve an allocation more favourable<br />
to itself. Inquiry, which requires agents to pool<br />
their information to solve a problem, begins<br />
with neither party having information t<strong>here</strong>by<br />
requiring cooperation via information exchange<br />
to solve the problem. In information seeking<br />
dialogues the parties start in an asymmetric<br />
position, one with some information which the<br />
other lacks. The aim of this dialogue type is the<br />
spread of information. One participant has the<br />
goal of acquiring this information. Another may<br />
have the goal of sharing its information, but in<br />
some circumstances may have the goal of hiding<br />
it. Finally, deliberation dialogues concern actions:<br />
the starting point is that some action must be<br />
taken and the aim of the dialogue is to reach a<br />
decision acceptable to the parties involved. The<br />
goals of the participants are to influence the<br />
decision such that it is acceptable to them.<br />
These dialogues types are intended to be rather<br />
broad, and will have a number of flavours depending,<br />
for example, on the relationship between the<br />
agents. Agents<br />
six dialogue types: persuasion, negotiation, inquiry,<br />
information seeking, deliberation and eristic<br />
may have a<br />
common cause<br />
and be disposed<br />
to co-operate,<br />
or they may be<br />
indifferent or even hostile towards one another.<br />
Thus in a negotiation, an agent may be concerned<br />
only with achieving the allocation which is best<br />
for itself, or it may constrain its choices to ensure<br />
that another agent improves its position to some<br />
extent, or it may even be prepared to sacrifice itself<br />
so as to achieve the best outcome with respect to<br />
joint utility. Similarly, in a deliberation, agents<br />
may have more or less concern for whether the<br />
decision is acceptable to the other parties: contrast<br />
a couple planning a holiday with a democratic<br />
debate between political parties. Perhaps the<br />
most striking contrast is the information seeking<br />
dialogue w<strong>here</strong> the agent with the information<br />
may either be willing or reluctant to share its<br />
knowledge.<br />
12<br />
November 2005|<strong>AgentLink</strong> News 19
AGENT RESEARCH OVERVIEW<br />
An important fact to recognise is that an<br />
interaction will typically involve several of these<br />
dialogue types. Thus a deliberation dialogue<br />
may well move through information seeking,<br />
inquiry, persuasion and negotiation phases before<br />
a decision can be reached. Walton and Krabbe<br />
stress the importance of these dialectical shifts,<br />
which may be sharp or gradual. In particular,<br />
misunderstandings often occur when a shift has<br />
taken place but one of the parties to the dialogue<br />
has not noticed. Moreover shifts may be illicit,<br />
and it is these illicit shifts that give rise to informal<br />
fallacies. The important lesson from this is that we<br />
cannot study particular dialogue types in isolation:<br />
for example if we focus on negotiation dialogues<br />
we will still need to consider the other types in<br />
order to model the complete interaction.<br />
Another point to note is that these dialogues can<br />
be about beliefs or about actions. Deliberation is<br />
tied to action and inquiry to belief, but persuasion<br />
may involve convincing the other party that<br />
something is the case, or that a particular action<br />
is wrong. Negotiation may be about the division<br />
of resources, but also about division of tasks.<br />
Information may be sought about facts or about<br />
options for action. This distinction will become<br />
important as we come to discuss arguments.<br />
The prevailing model, based on speech acts,<br />
sees the content of utterances to be essentially<br />
propositional, with the different speech acts<br />
imparting different forces, such as assertive or<br />
interrogative. To participate in the above dialogue<br />
types, however, the agents will often need to<br />
exchange information with more structure so as to<br />
back up their assertions and choices with reasons.<br />
In particular they will need to exchange arguments,<br />
to justify themselves and to persuade other agents.<br />
Proofs are the limiting case of arguments: w<strong>here</strong><br />
information is complete, concepts are sharp and<br />
conditionals are strict, utterances may be justified<br />
by proof. Such conditions can be met in formal<br />
domains such as mathematics, but in the domains<br />
in which agents typically operate information is<br />
incomplete, concepts are vague and conditionals<br />
are defeasible. Thus justification has to take the<br />
form of an argument, and as such is open to<br />
challenge and defeat.<br />
It has become accepted that an argument, often<br />
expressed in the form “since P then Q” can be<br />
challenged in three ways. Since the underlying<br />
conditional is not strict, the argument can<br />
be rebutted by an argument for the opposite<br />
conclusion, say not Q. Alternatively the premise<br />
may be rejected. Finally it may be argued that the<br />
rule is not applicable in the particular case, perhaps<br />
because the circumstances are exceptional, or the<br />
offered rule is based on incomplete information.<br />
Such criticisms are normally called undercutters.<br />
These forms of attack do not, however, cover<br />
all the ways in which an argument justifying an<br />
action can be challenged. Philosophers have, since<br />
the time of Aristotle, seen practical reasoning<br />
– reasoning about action – as a separate subject<br />
of study, and these additional attacks motivate<br />
this. Arguments for actions can be challenged<br />
not only in the above three ways but also by<br />
proposing alternatives, other actions that will<br />
achieve the same goal, or some more desirable<br />
goal, and by pointing to side effects, inadvertent<br />
and undesirable consequences of the action<br />
which will occur in addition to realising the<br />
goal. Note that these attacks do not, unlike the<br />
attack on an argument for a belief, render that<br />
argument invalid or inapplicable. The reasons<br />
advanced for performing the action remain valid<br />
reasons, but they do nevertheless provide reasons<br />
for rejecting the conclusion. The resolution of a<br />
discussion also differs when it concerns actions<br />
rather than beliefs. Because beliefs are validated<br />
against the objective measure of what is in reality<br />
the case, ending the discussion with disagreement<br />
cannot be considered successful. Because actions,<br />
in contrast, relate to what we want to be the case,<br />
what we shall attempt to achieve, the discussion<br />
can rationally end in disagreement since the<br />
interest and the aspirations of the parties may<br />
differ.<br />
Arguments play different roles in the various<br />
dialogue types. Argument is obviously<br />
fundamental to persuasion dialogues: it is<br />
through argument that the parties can be brought<br />
to change their positions. An important point to<br />
recognise, however, is that the argument needs<br />
to be persuasive not to the party advancing it,<br />
but to the audience. An understanding of the<br />
background assumptions and prejudices of the<br />
audience is t<strong>here</strong>fore essential for successful<br />
persuasion. In negotiation, arguments are used to<br />
indicate limits to flexibility, or other problems with<br />
the proposal, and so to indicate what subsequent<br />
offers may be successful. If I am buying a car,<br />
and refuse the asking price, my arguments may<br />
show that I cannot afford it, or that I find the car<br />
unsatisfactory, and the seller’s response will differ<br />
in the two cases. Argument is useful in guiding<br />
the negotiation so as to improve efficiency and<br />
perhaps also the outcome. In inquiry dialogues,<br />
the participants are seeking a joint solution,<br />
and so are in effect seeking to co-operate in the<br />
construction of an argument to justify their<br />
conclusion, and in the course of the dialogue<br />
partial arguments leading towards the solution<br />
may be proposed. Similarly an argument to justify<br />
a line of action is jointly sought in deliberation,<br />
and in deliberation a variety of arguments for<br />
alternatives will typically be advanced to allow<br />
the parties to determine the options from<br />
which the best course of action can be chosen.<br />
In information seeking dialogues the role of the<br />
argument is essentially to justify the information<br />
provided, to round out the understanding of<br />
the information seeker. Arguments can play a<br />
variety of roles, and these determine who needs<br />
to accept them. They can justify, in which case<br />
it is enough that the agent making the argument<br />
accepts it. This predominates in negotiation and<br />
information providing. Or they can convince,<br />
when the agent to whom the argument is directed<br />
must accept it, as in persuasion. Or they can be<br />
used as part of a problem solving process, as in<br />
inquiry and deliberation, in which case all parties<br />
involved need to find them acceptable.<br />
In this article we have drawn attention to some<br />
important insights from discourse theory and<br />
informal logic which we believe need to be absorbed<br />
by those working on agent communication if the<br />
aspiration of agents capable of interacting in a<br />
rich and fruitful manner is to be achieved. T<strong>here</strong><br />
are a number of important lessons:<br />
• The pragmatic content of speech acts: it is<br />
vain to seek a single meaning of a speech act,<br />
such as tell, independent of the discourse<br />
context in which it is used.<br />
• The range of dialogue types, and the need to<br />
recognise that interactions will shift between<br />
them. It is not possible either to treat all<br />
dialogues as alike without reference to the<br />
initial starting point, the aims of the dialogue,<br />
and the goals of the participant. Nor can any<br />
single type be studied in isolation.<br />
• The distinction between dialogues about<br />
belief and dialogues about action. In<br />
particular, arguments relating to action can<br />
be challenged in additional ways, and rational<br />
disagreement is always possible.<br />
• The importance of argument and the<br />
recognition that arguments play different<br />
roles in different dialogue types has important<br />
implications for what counts as a successful<br />
argument in the various contexts.<br />
These issues indicate a number of lines for<br />
interesting research, and it is questions like these<br />
that make agents and argumentation such an<br />
exciting field of study.<br />
References<br />
[1] Barbuceanu, M., and Fox, M.S., (1995),<br />
COOL: A Language for Communication<br />
in Multi Agent Systems in Lesser, V., (ed),<br />
Proceedings of the First International<br />
Conference on Multiagent Systems, MIT<br />
Press, Cambridge, Mass.<br />
[2] Carlson, L., (1983), Dialogue games: an<br />
approach to discourse analysis, Reidel,<br />
Dordrecht.<br />
[3] Searle, J.R. (1969). Speech Acts: An Essay<br />
in Philosophy of Language. Cambridge<br />
University Press.<br />
[4] Walton, D.N., and Krabbe, E.C.W., (1995)<br />
Commitment in Dialogue, State Univ of<br />
New York Press.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
13
AGENT RESEARCH OVERVIEW<br />
Answer Set Programming and Agents<br />
Jürgen Dix<br />
Clausthal University of Technology<br />
Germany<br />
dix@tu-clausthal.de<br />
Thomas Either<br />
Vienna University of Technology<br />
Austria<br />
eiter@kr.tuwien.ac.at<br />
Introduction<br />
Agents need to make decisions based on what<br />
they perceive and their model of the world.<br />
Both sources are often incomplete and subject to<br />
change.<br />
Representing knowledge and reasoning under<br />
incomplete information about the world is indeed<br />
a very difficult task. Formal approaches extending<br />
classical reasoning (which was mostly based on<br />
classical logic) were first considered in the 80ies<br />
with the seminal works of Reiter, McCarthy, and<br />
McDermott/Doyle (see [1]) and led to the area of<br />
nonmonotonic logics. In parallel, in databases, an<br />
important milestone was the rise of the relational<br />
model introduced by Codd back in 1974, which<br />
revolutionised the area and found its way into<br />
industry and became a standard. At about the<br />
same time, Colmerauer and Kowalski introduced<br />
PROLOG as a programming language (we refer<br />
to [1] and the references t<strong>here</strong>in).<br />
Answer set programming (due to Gelfond<br />
and Lifschitz [2]) can be seen as a hybrid of<br />
these three approaches above: it combines a<br />
true declarativeness (which PROLOG never<br />
really achieved) with the efficiency of database<br />
technology and the expressive power of<br />
nonmonotonic reasoning.<br />
ASP in a nutshell<br />
What exactly is answer set programming (ASP for<br />
short) Let us consider a simple example and try<br />
to show the essential features of ASP. Consider<br />
the following simple rules:<br />
(1) col(X,red) v col(X,blue) v col(X,green) ←node (X)<br />
(2) ← edge(X,Y), col(X,C), col(Y,C)<br />
The first rule expresses that an object that is a<br />
node, has to have at least one of three colours<br />
red, blue or green. The second rule is a constraint<br />
stating that if two objects are related (t<strong>here</strong> is an<br />
edge between them), then they can’t take on the<br />
same colour.<br />
Suppose now t<strong>here</strong> is a database consisting of<br />
facts of the form node(a), node(b), node(c), ...<br />
and edge(a,b), edge(b,c), ... Such a database<br />
encodes a graph with nodes a,b,c,... and edges<br />
a→b, b→c, ... The question we would like to<br />
ask is: what are all possibilities to satisfy the rules<br />
(1) and (2) Note that X, C are uppercase letters<br />
that are considered as variables (as in PROLOG):<br />
all terms can be instances of them. It should<br />
not come as a surprise that all such possibilities<br />
(intuitively) constitute all 3-colourings of the<br />
underlying graph.<br />
The semantics of ASP (which defines the notion<br />
of an answer set of a program) makes sure that<br />
exactly the answer sets are 3-colourings. If we<br />
used classical logic and considered classical<br />
models, this would not hold. The reason is that<br />
in classical models, nodes might have assigned<br />
multiple colors; in answer sets, this does not<br />
happen since rules are applied parsimoniously<br />
and only one colour per node is included.<br />
An answer set program t<strong>here</strong>fore consists of<br />
• rules of the form head ← body<br />
• that can contain variables,<br />
• the head can be a disjunction (or empty)<br />
• the body is a conjunction (or empty).<br />
An important feature is that the body of a rule<br />
can also contain negation (a kind of negation-asfailure).<br />
This creates a close relationship between<br />
answer sets and extensions in default logic (and<br />
thus allows to use methods from nonmonotonic<br />
reasoning).<br />
We are not presenting all the different features of<br />
the language like disjunction, negation, strong<br />
negation, in the head and in the body, cardinality<br />
constraints, arithmetic, built-in predicates,<br />
preference handling etc. These and other<br />
extensions (surveyed in [17]) are very important<br />
from a user perspective. However the main<br />
features of ASP are as follows:<br />
• the definition of “if-then” rules,<br />
Figure 1. Schema for problem encodings with a separation between specification and data.<br />
• which contain variables but no functions,<br />
• an ASP program consists of finitely many such<br />
rules,<br />
• the semantics associates to each program a<br />
number of answer sets (0, 1, or several),<br />
• the semantics does not depend on the ordering<br />
of the rules or subgoals in them,<br />
• t<strong>here</strong> exist nowadays several high performance<br />
ASP engines to compute answer sets, including<br />
DLV, Smodels, Cmodels, ASSAT (see [9]).<br />
Programming with ASP then means the following.<br />
Suppose you are given a problem which has<br />
several solutions (e.g. a planning problem with<br />
several valid plans). Write down an ASP program<br />
such that the answer sets of it correspond to the<br />
solutions of the original problem. Then compute<br />
the solutions with an ASP engine.<br />
ASP has been proven to be successful in a variety<br />
of areas such as planning, diagnosis, configuration<br />
and space shuttle control, which will be considered<br />
more in detail below.<br />
Extensions of the basic ASP-rule language<br />
permit the use of constituents which are calls<br />
to arbitrary legacy code (e.g. functions giving<br />
back perceptions of the world (but encapsulated<br />
in logic)). This makes it possible to use ASP as<br />
the underlying program of an agent: the agent is<br />
constantly computing the potential answer sets<br />
and acting according to them. These answer sets<br />
change as new perceptions are coming in. This<br />
is the approach adopted in IMPACT, which is a<br />
genuine agent platform project briefly discussed<br />
below. Other generalisations of ASP to interface<br />
legacy code, like DLVEX or DLVHEX, are<br />
currently under development and will be released<br />
soon.<br />
ASP and Agent Programming<br />
Thanks to its capability of dealing with incomplete<br />
and default knowledge, ASP lends itself as a simple<br />
yet expressive rule-based language for knowledge<br />
representation, which may be utilised to develop<br />
special plug-in reasoning modules for agent<br />
architectures. For example, modules for providing<br />
decision support, for reasoning about actions, or<br />
for reasoning about security and trust may be<br />
crafted and interfaced with other components in<br />
the agent architecture. The modules may then be<br />
activated in the thinking part of the prototypical<br />
Observe-Think-Act cycle that has been proposed<br />
14<br />
November 2005|<strong>AgentLink</strong> News 19
AGENT RESEARCH OVERVIEW<br />
for logic-programming based agents by Kowalski<br />
and Sadri [10]. As mentioned above, extensions<br />
of ASP which facilitate access of sensory input<br />
and other agent data structures as first class<br />
citizens in the language will be soon available.<br />
Moreover, an extension accounting for possible<br />
perception failure has also been recently devised<br />
[18].<br />
In the context of multi-agent systems, ASP has<br />
been utilised in different projects already. In the<br />
DALI project [11], ASP serves as a backbone for<br />
an Active Logic Programming Language, which<br />
is designed for executable specification of logical<br />
agents, without committing to any specific agent<br />
architecture, and provides reactive and proactive<br />
features. An important aim in its definition<br />
was introducing in a declarative fashion all the<br />
essential features, keeping the language as close<br />
as possible to the syntax and semantics of the<br />
plain Horn clause language.<br />
The work [12] deals with specifying and<br />
verifying socially acceptable patterns of agent<br />
interaction in multi-agent systems. Protocol<br />
specifications which indicate exactly what kind<br />
of messages agents should exchange and when<br />
for achieving certain outcomes might be viewed<br />
as compromising agent autonomy. A relaxed<br />
approach thus uses a set of regulations or norms<br />
which describe what sorts of agent behaviour are<br />
acceptable and forbidden according to the society,<br />
and how agents who violate these regulations<br />
should be treated and sanctioned, with the<br />
overall goal of discouraging but not necessarily<br />
preventing socially unacceptable behaviour from<br />
occurring. ASP is used for the formalisation and<br />
Fig 2. IMPACT Agent architecture.<br />
implementation of the model, as well as query<br />
and verification language. A major advantage<br />
<strong>here</strong> is the possibility to combine specification<br />
and implementation into one formalism, and<br />
flexibility. Using the same ASP program, one<br />
can plan for a certain state, diagnose a state or<br />
generate all acceptable states. Interactions among<br />
logic-based agents using ASP are also respected<br />
in the framework of Social Logic Programming<br />
[19].<br />
A full-fledged extension of ASP for agent<br />
programming has been developed in the<br />
IMPACT (Intelligent Maryland Platform for<br />
Agent Collaborating Together) project, which<br />
has been led by VS Subrahmanian. IMPACT<br />
agents have a rule-based agent program which<br />
governs the agent behaviour, in terms of actions<br />
that the agent takes, on top of the agent’s internal<br />
data structures, which include its message box<br />
for communication (see Figure 2).<br />
Access of the internal data structures is facilitated<br />
through special code calls and atom that allows<br />
checking whether a given value belongs to the<br />
result of a code call or not. Action execution<br />
is based on rules in language which features<br />
deontic modalities such permittance, obligation,<br />
or forbiddance of action execution, which might<br />
depend on other actions. For example, the rule<br />
Do(dial(N)) ← IN(N,phone(P)), O(call(P))<br />
intuitively says that the agent should dial the<br />
phone number N and is obliged to call P. The<br />
general form of an agent rule in IMPACT is<br />
Op_0 act_0 ← Cond, (~) Op_1 act_1, ... , (~)<br />
Op_m act_m,<br />
w<strong>here</strong> each Op_i is a deontic modality, act_i<br />
are action atoms, and Cond is a conjunction<br />
of constraint atoms and atoms of the form<br />
IN(X,f(Y)) resp. NOTIN(X,f(Y)), which access<br />
the data structures of the internal agent state<br />
through API functions f(Y) and test whether<br />
X is in resp. not in the result. The semantics of<br />
such an agent program, which consists of a set of<br />
such rules is given in terms of a feasible status set,<br />
which can be viewed as a set of status assumptions<br />
for the action atoms compliant with the rules of<br />
the agent program, deontic laws (e.g., an action<br />
can not be both permitted and forbidden at the<br />
same time), integrity constraints on the agent<br />
state, and possible constraints on parallel action<br />
execution. A suite of refinements of feasible status<br />
set semantics has been defined, of which the so<br />
called reasonable status set semantics implements<br />
and faithfully generalises the semantics of answer<br />
set programs to agent programs. For more details,<br />
we refer to the chapter about IMPACT in [3].<br />
Applications<br />
ASP has been successfully deployed to a number<br />
of applications in various areas including classic<br />
AI fields such as planning, diagnosis, or decision<br />
support, but also to other fields which gained a<br />
lot of interest in the recent past such as security<br />
management, information integration, and web<br />
service composition. A comprehensive survey<br />
of applications of ASP in these and further<br />
areas is available at [7], and an ASP showcase<br />
featuring difference applications in the areas of<br />
configuration, information integration, security<br />
analysis, agent systems, semantic web, and<br />
planning can be found at [8].<br />
Here, we just briefly touch three applications,<br />
reporting the gist and the main benefits of the<br />
ASP usage. An interesting application was<br />
a knowledge-based decision support system<br />
for controlling some functions of the Space<br />
Shuttle [5]. Here, the task was to find plans for<br />
the operation of the Reaction Control System<br />
(RCS) of the Shuttle, which has the primary<br />
responsibility for manoeuvring the aircraft while<br />
it is in space. When failures in the system become<br />
apparent, astronauts need to execute sequences of<br />
actions which must assembled on the fly in order<br />
to complete the mission. For that, a system based<br />
on ASP has been conceived, which implements<br />
a simplified model of the RCS (without loss of<br />
necessary detail) and incorporates agent actions<br />
and planning. The main benefits of ASP in this<br />
application are a seamless integration of reasoning<br />
about actions, change, and planning using a<br />
natural declarative language; in fact, the new<br />
system worked much better than its predecessor.<br />
Another task in which ASP turned out to be<br />
very useful is advanced information integration,<br />
which is an important issue for information<br />
November 2005|<strong>AgentLink</strong> News 19<br />
15
AGENT RESEARCH OVERVIEW<br />
agents. Here, the task is providing a uniform<br />
interface to various pre-existing data sources, so<br />
as to enable users to focus on specifying what<br />
they want, rather than on thinking of how to<br />
obtain the answer. Problems <strong>here</strong> are that the<br />
integration of data from different sources to<br />
a global user view, w<strong>here</strong> source data is usually<br />
heterogeneous, incomplete, and inconsistent. In<br />
the recent EU project INFOMIX [6], ASP was<br />
used as a computational backbone to implement<br />
query answering from inconsistent information<br />
integration systems. Here, an acknowledged<br />
formal model for query answering has been<br />
cast into an ASP program, which serves as an<br />
executable logical specification of the semantics,<br />
which aims at “repairing” inconsistencies. The<br />
benefits of ASP in this application are that it<br />
has expressiveness required for this application,<br />
and that it provides a language for expressing<br />
repair policies, which may be customised to a<br />
particular scenario, in a fully declarative manner<br />
rather than in a procedural way. This facilitates<br />
reasoning about specifications, their properties<br />
and behaviour much better since reasoning<br />
about programs is a principal issue in logic<br />
programming.<br />
A third application which we briefly touch<br />
upon <strong>here</strong> is monitoring of agent collaboration<br />
in a multi-agent system. Agent behaviour has<br />
to be judged on the basis of message exchange<br />
between agents. The task is to find out whether<br />
the collaboration is compliant with message<br />
exchanges, collected in a message log, which<br />
correspond to agent action sequences that will<br />
accomplish an overall goal. To this end, agent<br />
collaboration is modelled as an action theory, and<br />
feasible action sequences reaching the goal are<br />
determined by a planner, whose compliance with<br />
the actual behaviour of agents allows to detect<br />
possible collaboration failures, which may be due<br />
to implementation errors (the code of some agent<br />
is buggy) or due to design errors (the collaboration<br />
is improper). In this way, the approach can be<br />
applied to aid offline testing as well as online<br />
monitoring. It has been implemented within<br />
IMPACT and DLVK [13], which is a planning<br />
system based on a declarative action language and<br />
ASP. Thus, ASP comes into play <strong>here</strong> at different<br />
levels. Figure 3 shows the message protocol for<br />
the agents in the Gofish Postal Service system,<br />
which collaborate for servicing package delivery.<br />
Even in this small example, communication has<br />
many steps and errors are (especially in the system<br />
implementation phase) almost unavoidable;<br />
monitoring helped a lot in debugging different<br />
implementations of the system.<br />
WASP<br />
A working group on Answer Set Programming<br />
(WASP) has been created in 2001 as a European<br />
project within the 5th Framework Programme<br />
(IST project IST-FET-2001-37004). It consist of<br />
18 nodes spread all over Europe: they constitute<br />
the main body of research about ASP done in<br />
Europe.<br />
WASP has greatly helped to keep European<br />
leadership in ASP and to identify and realise a<br />
set of target industrial applications. It has also<br />
helped to develop a variety of language extensions<br />
providing constructs and features which ease<br />
problem solving in practice.<br />
The web-page contains a showcase collection<br />
which provides information about the use of<br />
ASP in configuration, information integration,<br />
security analysis, agent systems, semantic web,<br />
and planning.<br />
The interested reader is referred to the homepage<br />
of WASP - http://wasp.unime.it -.<br />
Conclusion<br />
Answer Set Programming started as a particular<br />
semantics for logic programs with negation in<br />
the late 80s. It is now considered a declarative<br />
paradigm for representing and reasoning about<br />
knowledgein a rule-based way. Methodologies<br />
have been (or are still being) developed to make<br />
ASP useable also for the non-specialist.<br />
ASP is useful in agent programming for several<br />
reasons:<br />
(1) to seamless integrate reasoning about<br />
actions, change, and planning using a natural<br />
declarative language;<br />
(2) to combine specification and implementation<br />
into one formalism (executable<br />
specifications);<br />
(3) to build decision support systems based on<br />
clear policies;<br />
(4) to get easy access to efficient database<br />
technology (the DLV solver, e.g., offers an<br />
ODBC interface to relational databases);<br />
and<br />
(5) to get access to a vast theory and methods<br />
about knowledge representation developed in<br />
the last decades.<br />
[15] and [16] point to some recent workshop<br />
series devoted to programming agent systems<br />
(w<strong>here</strong> ASP techniques might come into play).<br />
Fig.3: Protocol of message exchange in the Gofish system (message types).<br />
References:<br />
[1] G. Brewka, J. Dix, K. Konolige: Nonmonotonic<br />
Reasoning: An Overview, CSLI Lecture Notes<br />
73, Stanford (California), 1997.<br />
[2] M. Gelfond, V. Lifschitz: Classical negation<br />
in logic programs and disjunctive databases.<br />
New Generation Computing, 9:365--385,<br />
1991.<br />
[3] C. Baral: Knowledge Representation, Reasoning<br />
and Declarative Problem Solving with Answer<br />
Sets. MIT Press, 2002<br />
[4] R. Bordini, M. Dastani, J. Dix, A. El<br />
Fallah Segrouchni: Programming Multi<br />
Agent Systems: Languages, Platforms and<br />
Applications, Springer Series MASA, No. 15,<br />
2005<br />
[5] M. Nogueira, M. Balduccini, M. Gelfond,<br />
R. Watson, M. Barry: An A-Prolog decision<br />
support system for the space shuttle, Proc.<br />
3rd International Symposium on Practical<br />
Aspects of Declarative Languages (PADL<br />
2001), LNCS 1990, pp. 169-183. Springer,<br />
2001.<br />
[6] N. Leone et al.: The INFOMIX system for<br />
16<br />
November 2005|<strong>AgentLink</strong> News 19
AGENT RESEARCH OVERVIEW<br />
advanced integration of incomplete and<br />
inconsistent data, Proc. ACM SIGMOD<br />
2005 Conference, pp 915--917. Project<br />
homepage http://www.mat.unical.it/infomix/.<br />
[7] S. Woltran (ed), ASP Model Applications<br />
and Proofs-of-Concept: http://www.<br />
kr.tuwien.ac.at/research/projects/wasp/<br />
report.html.<br />
[8] ASP showcase: http://www.kr.tuwien.ac.at/<br />
research/projects/wasp/showcase.<br />
[9] ASP solvers: http://dit.unitn.it/~wasp/Solvers/.<br />
[10] R.A. Kowalski, F. Sadri: From logic<br />
programming towards multi-agent<br />
systems, Ann. Math. Artif. Intell. 25(3-<br />
4): 391-419 (1999)<br />
[11] S. Costantini, A. Tocchio: The DALI Logic<br />
Programming Agent-Oriented Language,<br />
Proc. 9th European Conference on Logics<br />
in Artificial Intelligence (Jelia 2004),<br />
LNCS/LNAI 3229, pp. 685-688. Springer,<br />
2004.<br />
[12] O. Cliffe, M. DeVos and J. Padget:<br />
Specifying and Analysing Agent-based<br />
Social Institutions using Answer Set<br />
Programming. Proc. Workshop on Agents,<br />
Norms and Institutions for Regulated<br />
Multiagent Systems (ANIREM) 2005,<br />
AAMAS.<br />
[13] DLVK Planning System: http://www.dbai.<br />
tuwien.ac.at/proj/dlv/K/<br />
[14] A European Working Group on ASP (http://<br />
wasp.unime.it/) coordinates and represents<br />
most of the work on ASP done in Europe.<br />
[15] ProMAS Workshop series: http://www.cs.uu.<br />
nl/ProMAS/<br />
[16] CLIMA Workshop series: http://centria.di.fct.<br />
unl.pt/~clima<br />
[17] I. Niemelä (ed): Language Extensions and<br />
Software Engineering for ASP: http://www.<br />
tcs.hut.fi/Research/Logic/wasp/wp3/wasp-wp3-<br />
web/wasp-wp3-web.html<br />
[18] F. Buccafurri, G. Caminiti, D. Rosaci.<br />
Perception-dependent reasoning and<br />
answer sets. Working Notes AI*IA-RCRA,<br />
pp. 119-126, 2005. http://www.ing.unife.it/<br />
eventi/rcra05/articoli/BuccafurriEtAl.pdf<br />
[19] F. Buccafurri, G. Caminiti: A Social<br />
Semantics for Multi-Agent Systems. Proc.<br />
8th Conference on Logic Programming and<br />
Nonmonotonic Reasoning (LPNMR 2005),<br />
LNCS 3662, 317-329. Springer, 2005.<br />
AgentCities.ES:<br />
Spanish network of agent-based computing<br />
(2003-2005)<br />
Antonio Moreno<br />
University Rovira i Virgili<br />
Spain<br />
antonio.moreno@urv.net<br />
AgentCities.ES 1 was a research network founded<br />
by the Spanish former Science and Technology<br />
Ministry from March 2003 to March 2005. It was<br />
coordinated by Antonio Moreno, from University<br />
Rovira i Virgili (URV) in Tarragona, Spain. The<br />
network had 18 members (17 University research<br />
groups on agent technology and the Artificial<br />
Intelligence Research Institute-IIIA). Its main<br />
aim was to join and coordinate the efforts of<br />
the Spanish research groups involved in the very<br />
succesful AgentCities European project 2 , which<br />
intended to build a worldwide network of agentbased<br />
platforms that provided interesting services<br />
to citizens. It is worth mentioning that Spain<br />
was one of the leading countries in AgentCities<br />
activities, and t<strong>here</strong> are also many Spanish research<br />
groups which have been actively involved in<br />
<strong>AgentLink</strong> III activities (e.g. promoting TFGs<br />
such as the ones on Agents Applied in Health<br />
Care, Networked Agents, Agent Oriented<br />
Software Engineering and Programming MAS).<br />
Some of the activities that have been carried out in<br />
the network in its 2-years duration have produced<br />
results that may be interesting to the <strong>AgentLink</strong><br />
III community. These activities have been the<br />
following:<br />
• A workshop called “Intelligent Agents in the<br />
Third Millenium” was organised within the<br />
Spanish Conference on Artificial Intelligence<br />
(CAEPIA-2003) in San Sebastián. It included<br />
the presentation of 9 papers, a demo session<br />
with 7 implemented multi-agent systems, and<br />
a panel on the use of formal methodologies in<br />
the design of agents and multi-agent systems<br />
(coordinated by Juan Pavón and Jorge Gómez,<br />
from the Univ. Complutense of Madrid).<br />
Michael Luck, the <strong>AgentLink</strong>-II coordinator,<br />
gave a very interesting invited talk on the<br />
status of agent technology in Europe at that<br />
moment. The workshop proceedings, with<br />
most of the papers in English, is freely available<br />
through the network web pages 3 .<br />
• The network organised the “First Spring School<br />
on Agents. Development of MAS. Concepts,<br />
methods and tools” at the University of Sevilla<br />
in March 2005. 35 PhD students attended this<br />
3-days school, in which the most prestigious<br />
Spanish researchers on agent technology<br />
lectured on fundamentals of multi-agent<br />
systems, FIPA standards, JADE, principles of<br />
complexity in MAS, electronic institutions, the<br />
INGENIAS methodology and development<br />
of agents in real-time environments. Some<br />
of the material is available in English in the<br />
school web pages 4 .<br />
• The members of the network also identified<br />
research areas of common interest and defined<br />
Working Groups associated with them. The<br />
most active WGs have been the following:<br />
Methodologies for Designing MAS, Learning<br />
and Planning, Trust and Security, and<br />
Application of Agents in Tourism. It is worth<br />
mentioning that the WG on Methodologies<br />
has built a very detailed set of web pages<br />
(in English) on Agent-Oriented Software<br />
Engineering (maintained by Pedro Cuesta at<br />
the Univ. of Vigo) 5 .<br />
• The network has funded the development of<br />
a document (in English) that describe the 18<br />
Spanish research groups that are its members.<br />
For each group, it provides information<br />
such as the main research lines, software<br />
products developed by the group, services<br />
that the group may provide, experience in<br />
technology transfer to industry, and current<br />
funded research projects. This document is<br />
freely available in the network web pages,<br />
and it may be useful to those searching for<br />
collaboration with Spanish partners in future<br />
projects 6 .<br />
1 http://grusma2.etse.urv.es/AgCitES/<br />
2 http://www.agentcities.org<br />
3 http://grusma2.etse.urv.es/AgCitES/subpagines/documents/f1.pdf<br />
4 http://imaginatica.eii.us.es/2005/agentes/index.phppag=schedule<br />
5 http://ma.ei.uvigo.es/aose/<br />
6 http://grusma2.etse.urv.es/AgCitES/subpagines/AgentCitiES.pdf<br />
A 1-year extension of the network, also coordinated<br />
by Antonio Moreno, has already been aproved by<br />
the new Education and Science Ministry, although<br />
it has not yet officially started. The new network<br />
has 20 research groups and involves around 120<br />
researchers.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
17
STANDARDS REPORT<br />
Latest News from the Standardisation World<br />
Monique Calisti<br />
Whitestein Technologies AG<br />
Switzerland<br />
mca@whitestein.com<br />
The “Latest News from the Standardisation<br />
World” column reports on relevant agent-driven<br />
or related standardisation activities and events.<br />
The central idea is to promote awareness of the<br />
important activities that several international<br />
standardisation bodies are undertaking in the<br />
perspective of consolidating the industrial uptake<br />
of agent technology.<br />
T<strong>here</strong> has been a flurry of activity regarding<br />
standardisation activities this summer. We<br />
report from the FIPA IEEE, which held its First<br />
Technical European Meeting in Budapest just<br />
before the AL3-TF3, and report on the latest news<br />
from the Global Grid Forum meeting, held in<br />
Boston (October 3-6, 2005). As ever, a plethora<br />
of recommendations and results from W3C<br />
working groups have appeared in recent months;<br />
we report on a selection that are of particular<br />
relevance to the <strong>AgentLink</strong> community.<br />
Finally, a Technical Forum Group for<br />
Coordinating Agent Standardization<br />
Activities, TFG-CASA, was formed, and met in<br />
Budapest, Hungary, with three main objectives:<br />
1) to assess the importance of ongoing agentdedicated<br />
and agent-related standardization<br />
activities; 2) to clarify how existing standardization<br />
bodies and groups relate and possibly contribute<br />
to establishing stronger liaisons; and 3) to<br />
understand and position the role of <strong>AgentLink</strong><br />
for ongoing and future agent-dedicated and<br />
agent-related standardization work. Full details<br />
of this meeting are reported in the TFG-CASA<br />
report on page 22.<br />
As usual, don’t forget, if you are directly involved<br />
in any standardisation activity, or you are aware<br />
of important standardisation work and events<br />
that you would like to publicise, you can directly<br />
contribute by sending your material to:<br />
standardisation-activity@agentlink.org<br />
Many thanks to all contributors for their<br />
support!<br />
FIPA – IEEE<br />
Contribution by:<br />
James Odell<br />
email: email@jamesodell.com<br />
The IEEE FIPA Standards Committee held<br />
its first meeting in Budapest, Hungary, 13-14<br />
September 2005, with 27 attendees. Membership<br />
signup will commence in October, with elections<br />
announced by the end of 2005.<br />
The following groups have been formed and<br />
approved:<br />
• Agents and Web Services Interoperability<br />
Working Group - AWSI WG. The primary<br />
objective of the AWSI WG is to fill the<br />
interaction gap between agents and web services.<br />
Agents should be able to locate and interact<br />
with web services seamlessly and vice versa.<br />
Chair: Hiroki Suguri (Comtec)<br />
Email: suguri@comtec.co.jp<br />
• Human-Agent Communications Working<br />
Group - HAC WG. The primary goal of<br />
the HAC WG is to produce one or more<br />
IEEE standards that extend the current<br />
FIPA performatives for human-agent<br />
communications. Even though the current<br />
FIPA agent communication language can be<br />
applied to human-agent communications,<br />
they were not designed for communications<br />
in which agents assist human in various<br />
contexts (e.g., decision makings). An<br />
extended set of performatives for humanagent<br />
communications can facilitate<br />
the application of agent technologies<br />
to domains that involve agents assisting<br />
humans regarding decision-making.<br />
Chair: John Yen (Pennsylvania State University)<br />
Email: jyen@ist.psu.edu<br />
• Mobile Agents Working Group - MA WG.<br />
The primary goal of this working group is to<br />
resume work on standards for mobile agents.<br />
Existing specifications will be improved and<br />
extended, and the latest research results and<br />
experiences from existing implementations<br />
will be incorporated. Along with specification<br />
proposals, the working group will develop<br />
reference implementations of protocols in the<br />
form of software components for agent toolkits.<br />
Chair: Peter Braun (Swinburne University of Technology)<br />
Email: pbraun@it.swin.edu.au<br />
• P2P Nomadic Agents Working Group<br />
- P2PNA WG. The objective is to define a<br />
specification for P2P Nomadic Agents, capable<br />
of running on small or embedded devices,<br />
and to support distributed implementations<br />
of applications for consumer devices, cellular<br />
communications and robots, etc. over a pure<br />
P2P network. This specification will leverage<br />
presence and search mechanisms of underlying<br />
P2P infrastructures such as JXTA, Chord,<br />
Bluetooth, etc. In addition, this working<br />
group will propose the minimal required<br />
modifications of existing FIPA specifications<br />
to extend their reach to P2P Nomadic Agents.<br />
Potential application fields for P2P Nomadic<br />
Agents are healthcare, industry, offices,<br />
home, entertainment, and transport/traffic.<br />
Chair: Bernard Burg (Panasonic)<br />
Email: burgb@research.panasonic.com<br />
• FIPA Specification Review Study Group<br />
- FSR SG. The primary goal is to review the<br />
existing pool of 25 Standards specifications<br />
and the related critiques of each of these<br />
specifications. Within its remit, it will also<br />
cover the specifications that did not make it<br />
to the Standards level. This review, reflecting<br />
also on the context of the development of<br />
the FIPA specifications, should help FIPA<br />
to decide whether or not to upgrade and<br />
revise it’s published FIPA specifications.<br />
Chair: Stefan Poslad (University of London)<br />
Email: stefan.poslad@elec.qmul.ac.uk<br />
The FIPA Board of Directors welcomes<br />
comments, suggestions and expressions of interest<br />
to join the IEEE FIPA Standards Committee. For<br />
any further information, please contact the FIPA<br />
Board of Directors at: board@fipa.org<br />
GGF - Global Grid Forum<br />
Contribution by:<br />
Omer Rana<br />
email: o.f.rana@cs.cardiff.ac.uk<br />
Cardiff University and Welsh eScience Centre<br />
The 15th GGF meeting was held in Boston<br />
(October 3-6, 2005), and brought together<br />
researchers and implementers involved in Grid<br />
computing. The event featured a keynote address<br />
by Tony Hey (Microsoft Corp) on the challenges<br />
of building a secure and robust middleware<br />
infrastructure to support Grid computing. Tony<br />
highlighted key challenges in the security area<br />
18<br />
November 2005|<strong>AgentLink</strong> News 19
STANDARDS REPORT<br />
that still need to be addressed, along with the<br />
need for open standards to support interoperation<br />
between different Grid systems.<br />
Security issues also formed the basis of the<br />
workshop on “Leveraging Site Infrastructute for<br />
Multi-Site Grids” (organised by Von Welch from<br />
NCSA). A key theme in this workshop was rolebased<br />
access control support within Grids, which<br />
extends current efforts based on X509 certificates<br />
and MyProxy; systems discussed included<br />
Shibboleth and “GridShib”. Security issues were<br />
also explored in the HealthGrid session.<br />
The International Grid Trust Federation (IGTF)<br />
was established at GGF 15 to enable a single<br />
sign-on facility for access to Grid resources across<br />
the world. This would form the basis for a first<br />
step authentication (identity-management) and<br />
access control mechanism that could be shared<br />
by scientific users. The IGTF is a federation<br />
of certification authorities or Grid policy<br />
management authorities (Grid PMAs), and<br />
comprises three regional Grid policy management<br />
bodies, the Asia Pacific Grid PMA (APGridPMA),<br />
the European Policy Management Authority for<br />
Grid Authentication in e-Science (EUGridPMA)<br />
and the Americas GridPMA (TAGPMA). It<br />
has 61 members and covers 50 countries and<br />
regions. Many Grid projects, which require largescale<br />
deployment (such as the EU EGEE and<br />
DEISA projects), have already started to interact<br />
with IGTF. The availability of such security<br />
infrastructure can be of significant benefit when<br />
deploying agents through multiple platforms.<br />
The use of Web Service standards continues to<br />
play an important role in Grid computing. The<br />
GGF15 workshop on “Web Services Performance”<br />
focused on addressing the performance issue that<br />
still restrict the full adoption of Web Services<br />
in Grid projects, such as advanced parsing<br />
techniques, alternative binary encodings of XML,<br />
and streaming XML to address memory usage.<br />
GGF16 is scheduled to take place in Athens,<br />
Greece (February 13-16, 2006).<br />
Details about the IGTF can be found at: http://<br />
www.gridpma.org/<br />
W3C - Semantic Web Activity<br />
Contribution by:<br />
Terry Payne<br />
email: trp@ecs.soton.ac.uk<br />
University of Southampton<br />
Mechanisms for accessing and indexing assertions<br />
or documents are central to both the World Wide<br />
Web and Semantic Web frameworks. URLs<br />
became ubiquitous during the rise of the WWW<br />
as a means of linking and addressing documents.<br />
On September 9th, 2005, the World Wide Web<br />
Consortium released “xml:id Version 1.0” as<br />
a W3C Recommendation. The specification<br />
defines an attribute name, xml:id, that can always<br />
be treated as an identifier and hence can always be<br />
recognized, without fetching external resources,<br />
and without relying on an internal subset [1].<br />
The problem of identifying and addressing services<br />
is necessary when communicating between<br />
services, and communication mechanisms such<br />
as SOAP have become well established for Web<br />
Service communication. In August, we saw<br />
the release of “Web Services Addressing (1.0 -<br />
Core)” as a W3C Candidate Recommendation.<br />
Web Services Addressing provides transportneutral<br />
mechanisms to address Web services<br />
and messages. The candidate recommendation<br />
document defines a set of abstract properties<br />
and an XML Infoset representation t<strong>here</strong>of to<br />
reference Web services and to facilitate endto-end<br />
addressing of endpoints in messages.<br />
This specification enables messaging systems to<br />
support message transmission through networks<br />
that include processing nodes such as endpoint<br />
managers, firewalls, and gateways in a transportneutral<br />
manner. More details are available at [2].<br />
The advent of the Semantic Web has witnessed<br />
a plethora of different knowledge-bases and<br />
languages to access them. SPARQL has<br />
recently emerged as a uniform query language<br />
(pronounced “sparkle”), and is also designed<br />
to convey queries defined in other RDF query<br />
languages. The Data Access Working Group<br />
(DAWG) has released a Last Call Working<br />
Draft of the SPARQL Protocol for RDF [3],<br />
which uses WSDL 2.0 to describe a means for<br />
conveying SPARQL queries to an SPARQL<br />
query processing service and returning the<br />
query results to the entity that requested them.<br />
Comments on this draft, and on the DAWG<br />
activities can be found at [4].<br />
[1] http://www.w3.org/TR/2005/REC-xml-id-<br />
20050909/<br />
[2] http://www.w3.org/TR/2005/CR-ws-addr-core-<br />
20050817/<br />
[3] http://www.w3.org/TR/2005/WD-rdf-sparqlprotocol-20050914/<br />
[4] http://lists.w3.org/Archives/Public/public-rdf-dawgcomments/<br />
The <strong>AgentLink</strong> Roadmap<br />
has now been published<br />
Michael Luck, Peter McBurney, Onn Shehory, Steve Willmott<br />
and the <strong>AgentLink</strong> Community<br />
Agent Technology: Computing as Interaction<br />
Technologies<br />
Trends and Drivers<br />
Related Disciplines<br />
Related Techniques<br />
Grid<br />
Computing<br />
A Roadmap for Agent Based Computing<br />
Economics<br />
Agent Technology: Computing as Interaction.<br />
A Roadmap for Agent Based Computing<br />
<strong>AgentLink</strong> 2005<br />
ISBN 085432<br />
Mathematical<br />
Modelling<br />
Logic<br />
Programming<br />
Self<br />
Systems<br />
Ambient<br />
Intelligence<br />
Organisations<br />
Trust and Reputation<br />
Game Theory<br />
Comple Systems<br />
Logic<br />
Philosophy<br />
Biology<br />
Coordination egotiation Communication<br />
Hard copies of the roadmap will be distributed to all members in due course. If<br />
you would like extra hard copies of the roadmap to distribute at conferences or, in<br />
particular, to industrial or other contacts, mail admin@agenlink.org with details of the<br />
distribution, and we will endeavour to provide them.<br />
User<br />
Interaction<br />
Design<br />
Uncertainty<br />
in AI<br />
Robotics<br />
PeertoPeer<br />
Computing<br />
Anthropology<br />
Sociology<br />
Service Oriented<br />
Computing<br />
Reasoning<br />
and Learning<br />
The roadmap aims to provide an assessment of the current state-of-art of agentbased<br />
computing, and to suggest important areas for the furure research and<br />
development, both for academia and industry.<br />
ormal<br />
Methods<br />
Programming<br />
Languages<br />
Artificial Life<br />
Simulation<br />
Softare<br />
Engineering<br />
Interoperability Infrastructure<br />
Organisation Design<br />
Political Science<br />
Marketing<br />
Decision<br />
Theory<br />
Semantic Web<br />
The Roadmap may be downloaded at<br />
http://www.agentlink.org/roadmap/<br />
Compiled, written and edited by<br />
Michael Luck, Peter McBurney, Onn Shehory, Steve Willmott and the <strong>AgentLink</strong> Community<br />
November 2005|<strong>AgentLink</strong> News 19<br />
19
AGENTLINK TECHNICAL FORUM 3<br />
The Third <strong>AgentLink</strong> III Technical Forum<br />
Budapest, Hungary,<br />
September 15-17, 2005<br />
Andrea Omicini, Università di Bologna, Italy, andrea.omicini@unibo.it<br />
Paolo Petta, Medical University of Vienna and Austrian Research Institute for Artificial Intelligence, Austria, paolo.petta@ofai.at<br />
Lazlo Z. Varga, MTA SZTAKI, Hungary, laszlo.varga@sztaki.hu<br />
The third and final Technical Forum meeting<br />
(AL3-TF3) to be organised by <strong>AgentLink</strong> III took<br />
place in Budapest, Hungary, co-located with the<br />
4th Central and Eastern European Conference<br />
on Multi-Agent Systems (CEEMAS 2005),<br />
which generated a positive synergy throughout<br />
the week, resulting from several stimulating and<br />
vibrant discussions that emerged during both<br />
breaks and in the evenings.<br />
Thanks to the deft and professional care of the<br />
<strong>AgentLink</strong> III staff and the local organisers,<br />
AL3-TF3 experienced a flawless unfolding of<br />
the choreography of the meetings of seven TF<br />
Groups, some of which have continued from<br />
previous meetings:<br />
• Agent-Oriented Software Engineering<br />
(AOSE)<br />
• Towards a Standard Agent-to-Agent<br />
Argumentation Interchange Format (AIF)<br />
• Coordinating Agent Standardisation<br />
Activities (CASA)<br />
• Environments for Multi-Agent Systems<br />
(ENV)<br />
• Multi-Agent Resource Allocation (MARA)<br />
• Programming Multi-Agent Systems<br />
(PROMAS)<br />
• Self-Organisation in Multi-Agent Systems<br />
(SELFORG)<br />
As usual, each day concluded with a plenary<br />
session w<strong>here</strong> the TFG Chairs could report<br />
back to the community on the emerging issues<br />
from their meetings, as well as an update on the<br />
recently released <strong>AgentLink</strong> Roadmap.<br />
For this final issue of the <strong>AgentLink</strong> III Newsletter,<br />
all the AL3-TF3 groups have summarised their<br />
meeting activities for the <strong>AgentLink</strong> audience.<br />
By reading the reports, one should get some<br />
clues and appreciation of the vibrant and intense<br />
discussions, the stimulating atmosp<strong>here</strong>, and the<br />
achievements of the TFGs, and long reports will<br />
appear shortly on the <strong>AgentLink</strong> III web site.<br />
We t<strong>here</strong>fore wish to thank Gusztav Hencsey<br />
and Magdolna Zsivnovszki for their assistance<br />
as the local liaisons, as well as Peter McBurney,<br />
Catherine Atherton, Adele Maggs, Serena Raffin,<br />
Rebecca Earl, Michael Luck, and Terry Payne for<br />
their invaluable help in pro-actively addressing<br />
issues well before they could be even perceived<br />
as problems by the scientific organisation. It<br />
is thanks to both the <strong>AgentLink</strong> team, and the<br />
countless hours of effort provided by TF chairs<br />
and participants over the last two years that the<br />
<strong>AgentLink</strong> III Technical Forum Meetings have<br />
been such a great success: indeed, all TFGs are<br />
continuing their work, exploiting the substantial<br />
impetus gained!<br />
We thank you all.<br />
Towards a Standard Agent<br />
to Agent Argumentation<br />
Interchange Format (AIF)<br />
Steven Willmott 1 (Universitat Politecnica de Catalunya, Spain),<br />
John Fox (Cancer Research UK, UK), Dan Grecu (Cancer<br />
Research UK, UK), Simon Parsons (New York State University,<br />
United States), Iyad Rahwan (The British University of Dubai,<br />
United Arab Emirates), Chris Reed (Dundee University, UK),<br />
Dave Robertson (University of Edinburgh, UK), Nicolas Maudet<br />
(Universite Paris 9 Dauphine, France).<br />
http://x-opennet.org/aif/<br />
Argumentation theories provide a powerful<br />
framework that facilitates decision making<br />
by interacting agents, assessing the validity of<br />
a fact, or otherwise resolving differences of<br />
opinion. Argumentation, in general, focuses<br />
on interactions w<strong>here</strong> parties plead for and<br />
against some conclusion [Argmas2005], and is<br />
an essential ingredient of inter-agent dialogue,<br />
negotiation, persuasion and collaborative<br />
decision-making. The field of argumentation<br />
draws on a wide range of areas including<br />
philosophy, natural language, rhetoric, and logic,<br />
and multi-agent system frameworks provide a<br />
natural home for implementing, studying and<br />
testing Argumentation models and scenarios.<br />
In this context, the Argumentation Interchange<br />
Format (AIF) TFG met in Budapest, Hungary,<br />
1 Contact person: steve@lsi.upc.edu<br />
as part of the TFIII meeting in order to open<br />
discussion on the development of a first cut<br />
notation or interchange format for argumentation<br />
and arguments which could be used as a<br />
convergence point for theoretical and practical<br />
work on argumentation technology. Concretely<br />
the TFG aimed to generate:<br />
· A collection of inputs and opinions on the<br />
nature of such an interchange format (now<br />
on-line on the TFG homepage given below);<br />
· A document capturing a draft first-cut format<br />
proposal based on inputs and discussion (due<br />
for release in October 2005 on the TFG web<br />
site);<br />
· A mini-roadmap for future development of<br />
the format.<br />
The event was attended by approximately 20<br />
researchers and received 12 input documents<br />
covering a broad range of perspectives from<br />
philosophical foundations to low level encoding<br />
issues; thus spanning an excellent cross section<br />
of interested communities. Discussion at the<br />
event yielded concrete decisions on the potential<br />
structure of the AIF and made progress on the<br />
definition of a top-level model for core concepts.<br />
Amongst other things, it was decided that:<br />
· Since operational concepts vary widely,<br />
work would focus on the definition of a set<br />
of argumentation concepts and subsequent<br />
reifications to different concrete syntax<br />
instantiations.<br />
· That the set of concepts would be structured<br />
as a core – i.e. common to many applications<br />
– and extensions, which provide functionality<br />
for more specific cases (and which could be<br />
extended by users).<br />
20<br />
November 2005|<strong>AgentLink</strong> News 19
AGENTLINK TECHNICAL FORUM 3<br />
Communication<br />
(Locutions/<br />
Protocols)<br />
Refers-to / (Manipulates)<br />
Drives<br />
Argument<br />
Networks<br />
(Arguments/<br />
Relations)<br />
set of concepts which would be defined in a core<br />
argumentation ontology, and the links between<br />
the circles indicate top level relationships which<br />
would subsequently be replaced by ontological<br />
relationships between concepts in a complete<br />
ontology.<br />
Manipulates<br />
Constrains<br />
Influences<br />
Constrains<br />
The general principles, top-level ontology and<br />
other discussion results are now being compiled to<br />
form the first draft document (with an anticipated<br />
completion date of the end of October, 2005).<br />
Other future plans include ongoing discussion via<br />
the associated AIF Google-Groups forum 2 , as well<br />
as potential future events collocated with relevant<br />
conferences/workshops.<br />
· That no separation would initially be made<br />
between the use of the format between static<br />
tools (e.g. import / export of argumentation<br />
schemas) and use of the format in agent-agent<br />
communication; however, it may be necessary<br />
to make this distinction in the future.<br />
After discussion and analysis of the inputs, core<br />
concepts in argumentation were grouped into three<br />
top level types: concepts to do with arguments<br />
themselves and relationships between arguments<br />
Context<br />
(Participants/<br />
Theory<br />
Fig 1: Grouping of top concepts for Argumentation Formats resulting from the TFIII Budapest Meeting.<br />
(labelled argument networks); concepts to do<br />
with communication of arguments – locutions<br />
and protocols for argumentation (labelled<br />
communication); and finally concepts belonging<br />
to the context of argument such as participants,<br />
rules/theory applying, groundings of terms and<br />
so forth (labelled context). While other concepts<br />
relevant to argumentation may fall outside of<br />
this scope, these broad categories appeared to<br />
capture the majority of relevant items. Figure 1<br />
shows the concept groups: each circle represents a<br />
The TFG organisers would like to greatly thank all<br />
those that participated – both in providing input, in<br />
attending the event and for post-event discussions.<br />
The organisers and participants would also like to<br />
heartily thank the European Commission-funded<br />
project - <strong>AgentLink</strong> III, the Agentlink III board,<br />
and the TFG organisers for providing this unique<br />
opportunity for joint work. Lastly, the event would<br />
not have been possible without the support of the<br />
European Commission funded ASPIC project<br />
(http://www.argumentation.org/ which provided<br />
technical support/coordination and impetus for<br />
the activity.<br />
References<br />
[ArgMAS05] Workshop on Argumentation for<br />
Multi-Agent Systems, organised in conjunction<br />
with the International Conference on Autonomous<br />
Agents and Multi-Agent Systems (AAMAS) July,<br />
2005, Utrecht, The Netherlands.<br />
Agent-Oriented Software Engineering (AOSE)<br />
Massimo Cossentino, ICAR - Consiglio Nazionale delle Ricerche, Italy, cossentino@pa.icar.cnr.it<br />
Juan Pavón, Universidad Complutense Madrid, Spain, jpavon@sip.ucm.es<br />
The third AOSE TFG meeting was aimed at<br />
completing open activities from the previous two<br />
meetings, and was enriched by the participation<br />
of invited speakers from Australia and USA, who<br />
introduced new arguments and perspectives to<br />
the discussion. The main topics of this meeting<br />
(apart from the talks) were the refinement of the<br />
MAS meta-model defined in Ljubljana, at the<br />
AL3-TF2 meeting, and the evaluation of several<br />
AOSE methodologies through a questionnairebased<br />
approach.<br />
2 http://groups.google.com/group/Agent-Argumentation<br />
The meeting was organised in three sessions:<br />
1. Presentations regarding agent-related topics<br />
from outstanding researchers in the area.<br />
2. Refinement of the Ljubljana MAS metamodel.<br />
3. Methodologies evaluation: results of the<br />
questionnaires, comparison and debate.<br />
The presentations were given during the first<br />
session, commencing with a talk from G. Fortino<br />
on an agent-based approach for the management<br />
of distributed workflows. This was followed<br />
by talks by L. Padgham on the Prometheus-<br />
ROADMAP methodology, and B. Henderson-<br />
Sellers, who described the FAME project that<br />
deals with a method engineering framework<br />
founded on a powertype-based metamodel.<br />
In the second part of the session, J. Odell reported<br />
about his most recent industrial experiences<br />
on the use of AUML. Other talks included:<br />
A. Garcia, who presented his work on the<br />
introduction of aspects in MAS development and<br />
followed this with a discussion on both lessons<br />
learned and open issues; A. Molesini, who talked<br />
about the importance and role of artefacts in<br />
Agent-Oriented Software Engineering in general<br />
and more specifically in the SODA methodology;<br />
and M. Cossentino, who reported about the<br />
experiences gained in the FIPA Methodology<br />
Technical Committee, w<strong>here</strong> the method<br />
engineering paradigm has been adopted in order<br />
to converge towards an agent-oriented design<br />
methodology standardisation proposal.<br />
The second day was divided into two sessions.<br />
The Ljubljana MAS meta-model was considered<br />
during the first session; w<strong>here</strong> the introduction of<br />
new elements to this model was discussed. The<br />
November 2005|<strong>AgentLink</strong> News 19<br />
21
AGENTLINK TECHNICAL FORUM 3<br />
work started from a proposal by P. Giorgini and<br />
M. Cossentino about the introduction of the Goal<br />
Element, and the resulting outcomes included:<br />
• A refinement of the definition of the Task<br />
Element - initially introduced at the Ljubljana<br />
meeting, and now defined as “the activity (set<br />
of activities) motivated by some co<strong>here</strong>nt<br />
reason”;<br />
• A definition of Goal as “a specification of a<br />
state of the world that may be achieved or<br />
maintained by the agent”.<br />
• An initial definition was drafted for Plan as<br />
follows: “A plan specifies how a goal can be<br />
achieved”.<br />
The session concluded with the agreement that<br />
the MAS meta-model deserves further work,<br />
which will proceed through the mailing list.<br />
The second session included an active and<br />
lively debate on Methodologies Evaluation,<br />
with emphasis on a comparison of seven<br />
different methodologies that emerged from<br />
the results of a questionnaire presented by<br />
Juan Pavón (the questionnaire was made<br />
available through the AOSE TFG website). It<br />
was adapted from a paper by Khanh Hoa Dam<br />
and Michael Winikoff, which was presented<br />
at the AOIS03 workshop held in Melbourne,<br />
2003. The main comments regarded the<br />
impact of the relationship between people<br />
(who compiled the questionnaire) with the<br />
methodology they evaluated, on the evaluation<br />
results. Judgements were typically more critical<br />
from users of methodologies (as would be<br />
expected), and as a consequence, it was agreed<br />
that the lack of a user-filled questionnaire<br />
for some methodologies partially limited the<br />
quality of the presented results. However, the<br />
overall experience was useful in improving the<br />
questionnaire for future editions. The type of<br />
questionnaire used will be one of the main<br />
tools used as part of the AOSE evaluation<br />
framework, which would also include case<br />
studies and specific metrics.<br />
Coordinating<br />
Agent Standardisation<br />
Activities (CASA)<br />
Monique Calisti, Giovanni Rimassa, Whitestein Technologies AG, Switzerland, {mca;gri@whitestein.com}<br />
Omer Rana, Cardiff University and Welsh eScience Centre, United Kingdom, o.f.rana@cs.cardiff.ac.uk<br />
Stefan Poslad, Queen Mary University, United Kingdom, stefan.poslad@elec.qmul.ac.uk<br />
Terry Payne, University of Southampton, United Kingdom, trp@ecs.soton.ac.uk<br />
James Odell, Agentis Software, Inc, France, omg@jamesodell.com<br />
The main objectives of the <strong>AgentLink</strong> TFG-<br />
CASA are:<br />
1. To assess the importance of ongoing agentdedicated<br />
and agent-related standardisation<br />
activities.<br />
2. To clarify how existing standardisation bodies<br />
and groups relate and possibly contribute to<br />
establishing stronger liaisons.<br />
3. To understand and position the role of<br />
<strong>AgentLink</strong> for ongoing and future agentdedicated<br />
and agent-related standardisation<br />
work.<br />
The key aim is to clarify and strengthen the<br />
importance of standardisation activities for<br />
software agent technology, an issue of central<br />
importance to many academic and industrial<br />
organizations involved in the production of agentoriented<br />
software and associated techniques.<br />
Such standardisation activities are critical to the<br />
commercialisation of agent-based research, and<br />
towards strengthening interaction within the<br />
agent community.<br />
Towards this end, the TFG-CASA took a highlevel<br />
perspective on standardisation activities at<br />
the AL3-TF3 meeting in Budapest, by bringing<br />
together researchers and organizations directly<br />
involved in a number of standardisation bodies,<br />
including IEEE FIPA, OMG, GGF, and W3C.<br />
Participants met at the AL3-TF3 to exchange<br />
information and discuss the importance of creating<br />
synergies between the different standardisation<br />
bodies. This included members already active<br />
within standardisation bodies, and individuals<br />
involved in identifying application requirements,<br />
addressing possible infrastructure themes, and<br />
promoting best practice and experience towards<br />
agent-oriented computing. From this perspective,<br />
the <strong>AgentLink</strong> TFG-CASA activity was an<br />
essential component in helping to guarantee and<br />
promote such meta-level and standardisationagnostic<br />
coordination across various groups.<br />
Additionally, the TFG-CASA hosted several<br />
coordination sessions involving individuals from<br />
other TF groups, including Agent Argumentation<br />
Interchange Format (AIF), Agent-Oriented<br />
Software Engineering (AOSE), and Programming<br />
Multi-Agent Systems (PROMAS). The purpose<br />
of these sessions was to gain an understanding<br />
of w<strong>here</strong> standards are required for the various<br />
TFGs. The main issues that were identified are<br />
listed below, and are expected to be addressed<br />
in the context of the new IEEE FIPA Standards<br />
Committee (http://www.fipa.org).<br />
• Methodology: Method engineering enables<br />
developers to customize system development<br />
methodologies to fit the needs of an agent<br />
project. The general idea is to achieve<br />
modularity through the integration and<br />
customization of multiple methodologies.<br />
Possible areas for standardisation could<br />
include a method fragments meta-model, a<br />
method engineering repository meta-model,<br />
and possibly a MAS infrastructure metamodel.<br />
• Service-Oriented support for Agents: Web<br />
Services do not provide an adequate model for<br />
capturing the expressivity intrinsic to existing<br />
agent models, such as ACL performatives.<br />
A question that needs consideration is how<br />
to structure the formal relationship between<br />
agents and web services. Semantic Web<br />
Service standards are clearly relevant and<br />
were discussed, such as WSDL-S, OWL-S,<br />
and WSMO.<br />
• Auditing multi-agent systems: This would<br />
require monitoring the interactions between<br />
multi-agent systems at different levels, such<br />
as the behavioural level, infrastructure level,<br />
and so on. The notion of debugging agent<br />
systems was considered, and distinguished<br />
from the validation of agent behaviors. It<br />
was generally felt that multi-agent system<br />
debugging was of significant interest, and<br />
required greater consideration.<br />
• Formulation of application-driven<br />
requirements on agent system<br />
capabilities: This involves identifying<br />
the functionalities necessary in an agent<br />
system from the perspective of application<br />
development. This implies proactive<br />
engagement with the applications community<br />
to elicit requirements and identify current<br />
standards that may be impacted them.<br />
• Interaction between agents and<br />
environments: Agent systems rarely operate<br />
in isolation from an environment and as<br />
such, clearer statements concerning this<br />
relationship need to be established. This<br />
requires environment modeling, and defining<br />
the nature of the agent-environment interface.<br />
An important distinction was made between<br />
the wrapping of legacy systems, and the<br />
embedding of multi-agent systems within<br />
larger applications.<br />
Given that <strong>AgentLink</strong> III will officially end in<br />
December 2005, several TFG-CASA promoters<br />
and participants are considering alternate ways<br />
in which it would be possible to continue this<br />
meta-level coordination activity for agent-related<br />
standardisation bodies. Of course, <strong>AgentLink</strong> IV<br />
would represent the most ideal context!<br />
Further details, and copies of the presentations<br />
given at the TFG-CASA in Budapest can be found<br />
at http://www.agentlink.org/casa. Feedback and ideas<br />
are welcome. Please send your input to the TFG-<br />
CASA mailing lists: tfg-casa@agentlink.org.<br />
22<br />
November 2005|<strong>AgentLink</strong> News 19
AGENTLINK TECHNICAL FORUM 3<br />
Environments For Multiagent Systems (ENV)<br />
Danny Weyns and Tom Holvoet, Katholieke Universiteit Leuven, Belgium,<br />
{danny.weyns, tom.holvoet}@cs.kuleuven.be<br />
Recent efforts, including past TFG meetings<br />
(ENV) and the AAMAS workshops on<br />
Environments for Multiagent Systems (E4MAS)<br />
in New York and Utrecht have firmly established<br />
the Environment on the research agenda of the<br />
MAS community. As a result, the TF3-ENV in<br />
Budapest had two main goals: to discuss actual<br />
research issues on environments, and to bridge<br />
ongoing research on environments with other<br />
related research lines; in particular programming<br />
multiagent systems (PROMAS) and selforganisation<br />
(SELFORG).<br />
Several core topics have been identified at<br />
previous meetings, and these formed an<br />
interesting starting point for discussions. The<br />
first topic was: “Definition and Responsibilities<br />
of the Environment”. Previous exploration had<br />
yielded a basic understanding of the dual role of<br />
the environment in MAS: on the one hand, the<br />
environment is an essential part of any MAS; on<br />
the other hand the environment is an exploitable<br />
design abstraction for building MAS. During the<br />
discussion, a number of essential aspects were<br />
identified that should be part of a definition of<br />
environment:<br />
• the environment is a first-order abstraction in<br />
MAS;<br />
• the environment provides the surrounding<br />
conditions for agents to exist;<br />
• the environment enables and regulates both<br />
interaction among agents and access to<br />
resources.<br />
Several environment responsibilities were<br />
discussed: (1) the environment structures the MAS<br />
(physical, communication, social structuring);<br />
(2) the environment provides life-cycle support<br />
and management (wrt. resources, agents, dynamics<br />
in the environment); (3) the environment<br />
provides observability and accessibility; (4) the<br />
environment regulates the MAS (access control<br />
to resources, interaction mediation, norms, laws,<br />
etc.); and (5) the environment provides a domain<br />
ontology. Rather than “being responsibilities of an<br />
environment”, these aspects should be interpreted<br />
as “responsibilities that can be assigned to the<br />
environment”.<br />
The first session concluded with a critical<br />
self-reflection on whether or not traditional<br />
middleware mechanisms already provide the<br />
necessary components to support environments<br />
in MAS The conclusion was no: (1) from a<br />
middleware perspective, MAS require a very<br />
specific set of middleware mechanisms of<br />
which some have not yet been developed (e.g.,<br />
support for societies). Moreover, the integration<br />
of the required mechanisms is itself a research<br />
challenge; (2) designing application-specific<br />
environments (from software architecture down<br />
to implementation) on top of middleware is a<br />
huge research challenge.<br />
The second discussion topic was: “Mechanisms and<br />
Opportunities of the Environment”. Mechanisms<br />
are building blocks of environments that<br />
encapsulate reusable functionality for designing<br />
MASs. The discussion mainly focused on defining<br />
a taxonomy of mechanisms for environments,<br />
which prove to be extremely difficult, and it is<br />
unclear whether such a taxonomy would be of<br />
any use at all at the moment, without a thorough<br />
study of the literature.<br />
During the third discussion session, the group<br />
focussed on “Environment Engineering”.<br />
Three aspects of environment engineering<br />
were discussed: (1) modelling: a basis for<br />
environment engineering is a reference model<br />
that provides a generic functional decomposition<br />
of environment; (2) software architecture: this<br />
describes how the reference model is mapped<br />
onto software elements; and (3) methodology:<br />
should environment engineering be integrated<br />
with existing agent-oriented software engineering<br />
methodologies, or should a general software<br />
engineering perspective be taken. With respect to<br />
the latter, a number of architectural approaches<br />
for environments were discussed, including<br />
modelling the environment as a set of mediating<br />
artefacts that agents can use, or modelling it as a<br />
set of modules that represent different functional<br />
concerns/responsibilities of the environment,<br />
such as communication, perception, action and<br />
interaction.<br />
The topic of the last discussion session was<br />
“Multiagent modelling and simulation: the<br />
role of environment”, which focussed on two<br />
activities regarding environment engineering in<br />
the context of simulation: the first being that the<br />
domain that is simulated must be modelled; and<br />
the second, that the simulation infrastructure<br />
must be engineered. However, several concerns<br />
regarding the relationship between these two<br />
activities were identified, including timing or<br />
simultaneity of actions. An important challenge<br />
in this domain is to develop a reference model<br />
for environments specific to MAS simulation that<br />
relates the concerns to one another.<br />
Now the community has established a basic<br />
understanding of the role of the environment<br />
in MAS, these insights are being consolidated<br />
to provide a basis for future research on<br />
environments. An upcoming JAAMAS Special<br />
Issue on Environments for MAS will aim to<br />
communicate the acquired knowledge of research<br />
on environments to the MAS research community<br />
in general, and focus on environment engineering<br />
in particular. Future meetings will continue, with<br />
the next, E4MAS 2006 being held at AAMAS<br />
2006.<br />
ENV JOINT SESSIONS<br />
The main goal of the joint session with<br />
the PROMAS group was to discuss the<br />
programming of environments in MAS.<br />
A fruitful discussion on the very concept<br />
of environment and its particular role<br />
in MAS emerged from this, with one<br />
interesting topic of discussion being the<br />
role of the environment in organisational<br />
issues of MASs. Alternate viewpoints<br />
were argued: some believed that the<br />
role of the environment should be<br />
kept to a minimum, providing just the<br />
medium for agents to interact; w<strong>here</strong>as<br />
others argued that the environment<br />
should be assigned an active role in the<br />
organisation of a multi-agent system<br />
(as in electronic institutions).<br />
During the joint session with SELFORG,<br />
the group had an open discussion<br />
on the role of environment for selforganising<br />
mechanisms. Whilst it was<br />
accepted that the environment plays an<br />
essential role in many self-organising<br />
systems, from the perspective of the<br />
designer of self-organising systems,<br />
this role cannot be separated from the<br />
role of the agents in self-organising<br />
mechanisms. Such mechanisms<br />
that involve the environment should<br />
t<strong>here</strong>fore be considered as an integral<br />
concern.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
23
AGENTLINK TECHNICAL FORUM 3<br />
Multiagent Resource Allocation (MARA)<br />
Ulle Endriss<br />
University of Amsterdam, The Netherlands<br />
ulle@illc.uva.nl<br />
The allocation of resources is a significant issue<br />
in both Computer Science and Economics<br />
disciplines. To emphasise the fact that resources<br />
are being distributed amongst several agents<br />
and that these agents may influence the<br />
choice of allocation, the field is also known as<br />
Multiagent Resource Allocation. The questions<br />
investigated by computer scientists are often<br />
of a procedural nature (“how do we find an<br />
allocation”), while economists are more likely<br />
to concentrate on qualitative issues (“what makes<br />
a good allocation”). A comprehensive analysis<br />
of the problem at hand, however, requires an<br />
interdisciplinary approach. It is <strong>here</strong> w<strong>here</strong> the<br />
Multi-Agent system paradigm offers an excellent<br />
framework in which to study these issues.<br />
The Technical Forum in Budapest has been the<br />
second meeting of the <strong>AgentLink</strong> Technical<br />
Forum Group on Multi-Agent System<br />
(TFG-MARA). The group emphasises the<br />
interdisciplinary aspect of the field and aims at<br />
providing an environment for the exchange of<br />
new ideas and also at fostering collaborations<br />
between different European research groups.<br />
The Budapest meeting hosted over 20 researchers<br />
from a dozen different institution countries as<br />
diverse as Hungary, France, Germany, Romania,<br />
Spain, the Netherlands, and the UK.<br />
After the first TFG-MARA meeting, held in<br />
Ljubljana earlier in 2005, ten of the participants<br />
came together to write an extensive survey paper<br />
on Multi-Agent Resource Allocation, which<br />
is both available on the Internet and due to be<br />
published in the Informatica journal. This survey<br />
was presented and discussed during most of the<br />
first session of the meeting in Budapest.<br />
The MARA Survey reviews various languages<br />
used to represent the preferences of agents over<br />
alternative allocations of resources, as well as<br />
different measures of social welfare, to assess<br />
the overall quality of an allocation. It also<br />
discusses pertinent issues regarding allocation<br />
procedures and presents important complexity<br />
results. The presentation of theoretical issues was<br />
complemented by a discussion of software packages<br />
for the simulation of agent-based market places.<br />
The survey also introduces four major application<br />
areas for Multiagent Resource Allocation, namely:<br />
Industrial Procurement; sharing of satellite<br />
resources; manufacturing control; and Grid<br />
computing.<br />
At the previous TFG-MARA meeting, we had<br />
identified Fair Division as a topic that is of<br />
particular interest to the community, and that could<br />
greatly benefit from a closer interaction between<br />
computer scientists and economists. To address<br />
this requirement, we decided to invite an expert<br />
in the field to deliver an in-depth presentation<br />
from the viewpoint of Economics at the Budapest<br />
meeting. Thibault Gajdos, CNRS researcher at<br />
the EUREQua research group at the University<br />
of Paris 1, gave an invited tutorial on Fairness<br />
and Uncertainty. The tutorial examined various<br />
connections between decision theory and social<br />
choice and demonstrated how decision theory<br />
can provide a useful perspective in the context<br />
of the analysis of resource allocation problems,<br />
and more specifically, for the identification and<br />
implementation of fair allocations.<br />
The remainder of the meeting was filled with<br />
several contributed talks reporting on recent<br />
research activities within the European MARA<br />
community. The topics covered included<br />
complexity issues, applications in grid computing<br />
and crisis response, combinatorial auctions,<br />
heuristics and experimentation in negotiation,<br />
fair division procedures, connections between<br />
sociological theory and Multi-Agent systems, and<br />
Multi-Agent simulation.<br />
Full details of the presentations, as well as the<br />
MARA Survey, are available at the MARA website<br />
(http://www.illc.uva.nl/~ulle/MARA/).<br />
Programming Multi-Agent Systems (PROMAS)<br />
Mehdi Dastani, Utrecht University, The Netherlands, mehdi@cs.uu.nl<br />
Jorge Gómez Sanz, Universidad Complutense de Madrid, Spain, jjgomez@sip.ucm.es<br />
The aim of the PROMAS technical forum was<br />
to promote collaboration between industry and<br />
academy by providing an informal platform w<strong>here</strong><br />
attendants can present their work and discuss<br />
issues related to the implementation of multi-agent<br />
systems. A little under 40 participants took place at<br />
the meeting in Budapest, which was organised into<br />
four sessions; each focussing on one of the topics<br />
below:<br />
• “Programming languages/issues for individual<br />
agents”. Speakers: Viviana Mascardi (University<br />
of Bologna), Birna van Riemsdijk (Utrecht<br />
University), and Robert Ross (Universität<br />
Bremen).<br />
• “Programming languages/issues for multiagent<br />
environments”. Speakers: Danny Weyns<br />
(Katholieke Universiteit Leuven) and Mirko<br />
Viroli (Università di Bologna a Cesena).<br />
• “Programming languages/issues for multiagent<br />
organisation/coordination structures”.<br />
Speakers: Andrea Omicini (Università di<br />
Bologna a Cesena), Mehdi Dastani (Utrecht<br />
University), and Jorge Gómez (Universidad<br />
Complutense de Madrid).<br />
• “Debugging and IDEs for multi-agent<br />
systems”. Speakers: Laszlo Gulyas (AITIA<br />
company), Juan Botía (Universidad de<br />
Murcia), and Alexander Pokahr (Universität<br />
Hamburg).<br />
The development of agent-oriented programming<br />
languages is still a main issue in this community.<br />
Many agent-oriented programming languages<br />
have been proposed, e.g., IMPACT, SOCS,<br />
3APL, Jason, Jadex, CLIMA, or Jack; however,<br />
few have reached a level of maturity necessary<br />
for widespread adoption, as t<strong>here</strong> is still a need<br />
for more powerful agent-dedicated programming<br />
constructs, debugging facilities, formal<br />
ontologies, primitives to deal with distributed<br />
code, and development environments. Also, we<br />
need to explore programming principles applied<br />
to conventional programming languages, like<br />
modularity, compositionality, and ease of use, and<br />
introduce them in agent-oriented programming<br />
languages. The importance of primitives for<br />
invoking other languages, like Java and C++,<br />
from an agent program were discussed, as was the<br />
need for guidelines and manuals, and examples<br />
that illustrate how these languages should be<br />
applied.<br />
A joint session was held with the ENV technical<br />
forum. Here, the focus was the implementation<br />
of the environment in multi-agent systems. The<br />
environment of multi-agent systems is different<br />
from multi-agent infrastructures, individual<br />
agents, and their organisation structures. The<br />
ENV-TFG presented two possible approaches<br />
to the implementation of the environment for<br />
multi-agent systems: by means of a set of artefacts<br />
which can be exploited by agents to achieve their<br />
24<br />
November 2005|<strong>AgentLink</strong> News 19
AGENTLINK TECHNICAL FORUM 3<br />
goals, and by a dedicated component like the<br />
Delta framework. In either approach, laws can<br />
be built in to codify, for example, the effect of<br />
actions.<br />
The third session was about coordination and<br />
organisation of multi-agent systems. It was<br />
argued that the organisation of multi-agent<br />
systems is a specific kind of coordination, and<br />
thus the implementation of such organisations<br />
can be realised by means of various technologies;<br />
including tuple centres, coordination artefacts,<br />
and message oriented communications.<br />
Finally, the meeting concluded with a debate<br />
about the maturity of debugging techniques<br />
applied to multi-agent systems. T<strong>here</strong> are complex<br />
systems such as agent-based social simulations,<br />
which require specialised tools able to detect<br />
bugs and errors through thousands of messages.<br />
An example of such a tool is ACLAnalyzer, a<br />
JADE third-party add-on. Moreover, inspection<br />
and debugging of the internal states of individual<br />
agents could be easier using concepts inspired<br />
by agent meta-models and BDI models. Based<br />
on conventional software engineering expertise,<br />
a wish-list of desirable debugging features<br />
was identified, including breakpoints, asserts,<br />
automated testing, and tracing facilities. As<br />
methods for distributed debugging were less well<br />
understood, the community agreed to examine<br />
them in preparation for future meetings.<br />
More details, as well as presentation slides, can be<br />
found at the PROMAS web page:<br />
http://www.cs.uu.nl/~mehdi/al3promas.html.<br />
Self-Organisation in Multi-Agent Systems (SELFORG)<br />
Di Marzo Serugendo 1 , Giovanna, University of Geneva, Switzerland, dimarzo@cui.unige.ch<br />
Gleizes, Marie-Pierre, University Paul Sabatier, France, gleizes@irit.fr<br />
Karageorgos, Anthony, University of Thessaly, Greece, karageorgos@computer.org<br />
Decentralised control, self-organisation and<br />
emergent behaviour are issues of major interest<br />
in distributed complex systems, such as MAS.<br />
The aim of the Self-Organisation in Multi-<br />
Agent Systems (SelfOrg) TFG was to discuss<br />
and investigate these issues through examination<br />
of representative case studies and organisation<br />
of interdisciplinary collaborative interaction<br />
activities.<br />
The TFIII meeting in Budapest was the third<br />
meeting of the TFG Self-Organisation in MAS,<br />
and concentrated on three main subjects: 1)<br />
finalising the definition of the emergence concept;<br />
2) extending the understanding of existing selforganisation<br />
mechanisms via presentations of<br />
relevant work; and 3) finalising the case study<br />
exercises, consolidating lessons learned, and<br />
highlighting issues and directions requiring<br />
further work.<br />
The first part of the meeting focused on the case<br />
studies that were developed during the previous<br />
meetings. The decision to establish three case<br />
studies was taken at the first meeting in Rome.<br />
Subsequently, the second meeting in Ljubljana<br />
focused on the description of the selected case<br />
studies and on propositions of agent-based selforganisation<br />
approaches for modelling and/or<br />
for implementing solutions to the problems they<br />
describe. In Budapest, Paul Verstraete presented<br />
a solution to the Manufacturing Control Case<br />
Study, w<strong>here</strong> self-organisation is examined from<br />
an environment perspective. After considering<br />
this alternative view, the question of whether<br />
or not the selected case studies could be useful<br />
as reference points (benchmarks) for evaluating<br />
self-organisation mechanisms was discussed. The<br />
responses received indicated agreement on the<br />
selection, but suggested that additional work<br />
was required if the case studies were to better<br />
serve the benchmark role. Such improvements<br />
include extending and detailing the case study<br />
descriptions (i.e. input data files) and clarifying<br />
the measurements needed for a proper evaluation.<br />
Furthermore, it was agreed that additional case<br />
studies should be proposed, that new solutions to<br />
the case studies should be contributed and added<br />
to the TFG web site, and that an important final<br />
objective of this work should be to compare<br />
different mechanisms on the same case study.<br />
The second part of the meeting consisted of<br />
presentations of three self-organising applications.<br />
Luca Gardelli from University of Bologna, Italy,<br />
presented cognitive stigmergy; an extension of the<br />
well-known stigmergy mechanism for cognitive<br />
agents. Gerald Silverberg from Maastricht<br />
University, MERIT, The Netherlands, presented<br />
a model of innovation dynamics based on<br />
invasive percolation by adding endogenous R&D<br />
search by economically motivated firms. The last<br />
presentation by Giuseppe Vizzari, University<br />
Bicocca, Milan, Italy, focused on situated MAS<br />
for the detection of emergent links in adaptive<br />
Web sites.<br />
A large part of the meeting was devoted to fruitful<br />
discussions concerning topics relevant to selforganisation,<br />
and a trial to define the concept of<br />
“emergence” in the context of artificial systems,<br />
such as MAS. To this end, the group had to<br />
discuss the emergence concept from three points<br />
of view: The first one was regarding engineering,<br />
and focused on questions such as:<br />
• Can we control systems with emergent<br />
functionality<br />
1 Giovanna Di Marzo Serugendo was not able to participate at this third meeting due to an important<br />
event beyond her control.<br />
• What is the key issue in the design of systems<br />
with emergent functionality<br />
• Is the environment fundamental in these<br />
systems<br />
The second view concerned analysis of the selforganisation<br />
results with questions including:<br />
• Do we need an observer<br />
• If yes, w<strong>here</strong> is the location of the observer<br />
• What is its role<br />
The last was the operational point of view, with<br />
questions such as:<br />
• Can we speak about emergence in artificial<br />
systems<br />
• Why and under which conditions can we<br />
qualify that the phenomenon emerges in an<br />
artificial system is emergent<br />
• How can we define the notion of emergence<br />
in an artificial system (what could be an<br />
operational definition)<br />
These questions were aimed at guiding<br />
participants to form a definition of the emergence<br />
concept. Although an initial definition was given,<br />
additional discussion will be needed to refine and<br />
assess this definition based on further insight<br />
gained from additional elaboration on case<br />
studies.<br />
Beyond TFIII, the activities of the group will<br />
continue remotely. Our aim is to improve the<br />
case studies towards benchmarks which can be<br />
used by the community to evaluate and compare<br />
self-organisation mechanisms. Furthermore, we<br />
would like to propose a detailed formal definition<br />
of the concept of emergence in artificial systems.<br />
Finally, the results of the TFG’s collective efforts<br />
will be refined and comprehensively presented<br />
in a text-book about self-organising artificial<br />
systems, planned to appear in 2006.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
25
EVENT REPORT<br />
Semantic Web Days Munich<br />
6-7 October, 2005<br />
Andrea Kulas<br />
webXcerpt Software GmbH<br />
Germany<br />
ak@webxcerpt.com<br />
ALSO IN THIS SECTION...<br />
The 1 st International Workshop on Context for Web Services<br />
The 8 th Biennial Israeli Symposium on Foundations of<br />
Artificial Intelligence<br />
E4MAS 2005: Environments for Multiagent Systems<br />
Int. Workshop on Agent-Based Models for Economic Policy Design<br />
The 2005 IEEE/WIC/ACM Int. Joint Conference<br />
on Web Intelligence and Intelligent Agent Technology<br />
The 3 rd Workshop on Agents Applied in Health Care<br />
Agents in Space<br />
1 st Int. Workshop on Security and Trust Management<br />
Alain Léger,<br />
France Telecom R&D<br />
France<br />
Alain.leger@rd.francetelecom.com<br />
Semantic Web Days 2005 provided a forum for<br />
innovative companies and research institutions<br />
with a strong desire to accelerate the uptake<br />
of Semantic Web technologies to meet and<br />
exchange experiences, challenges and solutions.<br />
With attendees travelling from all over Europe<br />
and countries as far a field as the US, and Korea,<br />
the main goal of the two-day conference, held<br />
in October in Munich, was to present the latest<br />
Semantic Web technologies which are very<br />
promising or already in use. This international<br />
event was organised jointly by the EU Network<br />
of Excellence REWERSE (REasoning on the<br />
WEb with Rules and SEmantics) and by the<br />
EU Network of Excellence Knowledge Web<br />
(Realizing the Semantic Web).<br />
One of the central questions addressed over<br />
the two days focused on the degree of maturity<br />
of Semantic web technologies i.e. Ontology,<br />
Reasoning, Business rules, facing real needs<br />
in industry i.e. data, services and business<br />
automated integration. The message, which<br />
was also discernable in the concluding panel<br />
discussion, was that procedures to describe<br />
semantic co<strong>here</strong>nces presently mainly exist in<br />
the form of pilot studies, but as yet t<strong>here</strong> has<br />
been little uptake of such procedures in major<br />
industrial projects. Although only a limited<br />
number of projects have so far been realized<br />
with Semantic Web technologies, the power and<br />
benefits of using Semantic Web approaches, as<br />
well as the need for catch-up, became obvious as<br />
Semantic Web Days 2005 proceeded. The fact<br />
that t<strong>here</strong> were more than 100 participants at the<br />
Semantic Web Days illustrated that t<strong>here</strong> can be<br />
no doubt of the high potential for the application<br />
of Semantic Web technologies; especially given<br />
the already convincing pre-deployment of several<br />
Semantic Web applications.<br />
At present, established projects are focused<br />
on the development of corporate Knowledge<br />
Management, Telecommunication applications,<br />
Business integration for agile enterprises and<br />
Health care. Furthermore, what with more<br />
than 12 companies and institutions featured<br />
at the exhibition (as part of the Semantic Web<br />
Days event), it was clear that a number of SMEs<br />
already have products on the basis of Semantic<br />
Web technologies.<br />
The keynote speaker was Hermann Friedrich,<br />
who is responsible at Siemens AG for the<br />
development of the knowledge database used<br />
internally in the company. He spoke about<br />
initial experiences with the Semantic Web: “We<br />
also have the possibility to classify documents<br />
in content management, but with ontologies<br />
it is possible to demonstrate more complex<br />
co<strong>here</strong>nces.”<br />
Modeling complex causal co<strong>here</strong>nces was also a<br />
focus for Thomas Syldatke, one of the panelists at<br />
the concluding panel discussion “Earning Money<br />
with Semantic Web Technologies – examples of<br />
best practice and outlook for promising projects<br />
of the future”. He is responsible at Audi AG for<br />
the development of software for testing engines,<br />
and noted that: “…this is a very complex<br />
procedure, t<strong>here</strong> are a lot of rules which we are<br />
integrating with ontologies which need to be<br />
considered…”. Ten percent of the rules have<br />
been integrated in the first pilot application,<br />
which was finished in June. He then went on to<br />
say“…we asked the system which should be the<br />
next control step for the engine and compared<br />
the result with the program which has been in<br />
use already…”. Plans are currently underway<br />
to integrate 100 percent of the rules with the<br />
ontology by December this year.<br />
An exciting field for demonstrating the power<br />
of semantic technologies is the integration and<br />
application of rules. Researchers at the University<br />
of Munich demonstrated the possibilities with<br />
a geographical information system based on<br />
scalable vector graphics (SVG). Ontologies<br />
model rules which form the basis for adapting<br />
the graphics. Using this system, different views<br />
<strong>AgentLink</strong> III ATC – 2005<br />
Stockholm, 18 th October<br />
The <strong>AgentLink</strong> III annual Agent Technology<br />
Conference (ATC) facilitates the sharing of<br />
experiences in deployment of agent-based<br />
solutions among commercial organisations<br />
and with discussion of the benefits this<br />
technology brings their businesses. Several<br />
talks were presented, covering a range of<br />
issues from the technologies themselves to<br />
commercial relevance and strategy<br />
• Modeling the Aerospace Aftermarket with<br />
Multi-agents Systems<br />
Ken Woghiren (Lost Wax)<br />
• Intelligent Agents for Networked Home<br />
Environment<br />
Götz Brasche (Microsoft Innovation<br />
Center)<br />
• Contribution of Agent Technologies to<br />
Flexibility in Production and Product<br />
Development<br />
Kurt Fessl (Profactor)<br />
• Administrable Services in a Mobile, Dynamic<br />
World<br />
David Marples (Telcordia)<br />
• Intelligent Agents for Banking and Insurance<br />
Industries<br />
Chris van Aart (Y’all)<br />
• How to Sell Agent Technology into<br />
Commercial IT<br />
James Odell<br />
(James Odell Associates)<br />
In the afternoon, demonstrations of existing<br />
agent-based solutions were presented<br />
by: CalicoJack, Certicon, Erxa, Lostwax,<br />
Magenta, Profactor, Rockwell and Y’All.<br />
http://www.agentlink.org/agent-Stockholm/<br />
26<br />
November 2005|<strong>AgentLink</strong> News 19
EVENT REPORT<br />
can be modeled. On being asked if they were<br />
just rebuilding existing tools, such as Google<br />
Earth, the developer Hans Jürgen Ohlbach<br />
replied “… when compared with conventional<br />
navigation systems, we have the advantage of<br />
being independent of certain tools”.<br />
Four workshops on the topics “Industrial<br />
Applications of Semantic Web technologies”,<br />
“Vocabularies and Rules for Enterprise<br />
Applications”, “Semantic Web Services<br />
in Industry” and “Semantic Web for Life<br />
Sciences”, as well as a presentation on<br />
“Geospatial Information Processing for the<br />
Web” demonstrated more in detail that t<strong>here</strong> is<br />
high potential for semantic technologies such as<br />
ontologies and rules to provide better and faster<br />
solutions than conventional methods. Pilot<br />
projects presented at the Semantic Web Days<br />
proved those potential advantages.<br />
Overall, the organizers have been extremely happy<br />
with the success of the event, and their impression<br />
has been backed by the tenor of the participants,<br />
who wished to establish this conference so that<br />
developments and applications of Semantic Web<br />
Technologies could be communicated between<br />
companies and research institutions in a timely<br />
manner. Last, but not least, the participants list<br />
of the Semantic Web Days demonstrated that the<br />
goal of exchange between business and research<br />
has been achieved, as half of the participants<br />
came from companies while the rest came from<br />
research institutions. Qualified presentations,<br />
the participants themselves, nice surroundings<br />
and many opportunities for social networking<br />
contributed to the success of the two days.<br />
The First International Workshop on Context for Web Services (CWS’05)<br />
Djamal Benslimane, Lyon 1 University, France, djamal.benslimane@liris.cnrs.fr<br />
Chirine Ghedira, Université Claude Bernard, France, chirine.ghedira@iuta.univ-lyon1.fr<br />
Zakaria Maamar, Zayed University, U.A.E, zakaria.maamar@zu.ac.ae<br />
The First International Workshop on Context<br />
for Web Services (CWS’2005) that was held<br />
in conjunction with The Fifth International<br />
and Interdisciplinary Conference on Modeling<br />
and Using Context (CONTEXT’2005) took<br />
place in Paris, France, on July 5th, 2005. The<br />
workshop aimed at bringing together researchers<br />
and practitioners who are actively engaged in the<br />
fields of information systems, context and Web<br />
services (provide more detail on the themes).<br />
Nearly 20 people attended the workshop.<br />
The workshop received 17 submissions, from<br />
which 7 were selected for presentation and<br />
inclusion in the workshop’s proceedings. A<br />
keynote presentation by David L. Martin from<br />
SRI International in California is also featured<br />
in the workshop’s program. David L. Martin has<br />
worked extensively in the fields of software design<br />
environments, software agent frameworks, and<br />
the Semantic Web. In the latter area, he is the<br />
chair of the OWL-S coalition and a co-chair of<br />
the Semantic Web services language committee<br />
of the Semantic Web services initiative (SWSI).<br />
Several topics have been discussed during the<br />
one-day workshop including ontologies and<br />
context for Web services, context-based semantic<br />
matching for Web services composition, security<br />
of context during Web services interactions, just<br />
to cite a few.<br />
Web services are nowadays emerging as a major<br />
technology for deploying automated interactions<br />
between distributed and heterogeneous<br />
applications. In general, composing Web services<br />
rather than accessing a single Web service is<br />
essential and provides better benefits to users.<br />
Composition primarily addresses the situation<br />
of a user’s request that cannot be satisfied by any<br />
available Web service, w<strong>here</strong>as a composite Web<br />
service obtained by combining available Web<br />
services might be used. Several questions have<br />
been raised regarding Web services composition<br />
and execution including which businesses have<br />
the capacity to provision Web services, when and<br />
w<strong>here</strong> the provisioning of Web services occurs,<br />
and how Web services from independent parties<br />
coordinate their activities during execution so that<br />
conflicts are avoided. To address some of these<br />
questions, it is recommended to consider the<br />
context in which the composition and execution<br />
of Web services occur. Context is generally<br />
perceived as the information that characterizes<br />
the interaction between humans, applications,<br />
and the surrounding environment. From a Web<br />
services perspective, it is expected that context<br />
should define a set of common data about the<br />
current status of a Web service and its capability<br />
of collaborating with other peers (i.e. webservices<br />
or also non–web services), possibly enacted by<br />
distinct providers.<br />
The keynote speaker’s talk was geared towards<br />
semantic Web services technology and the<br />
mechanisms it provides to deal with context. He<br />
considered what kinds of contextual knowledge<br />
Web services-based systems need to handle, shown<br />
how selected semantic Web services approaches<br />
might be applied to these systems, and speculated<br />
about what else might be needed. According<br />
to David (Martin), a Web service is normally<br />
conceived as a neatly encapsulated module of<br />
functionality that can be easily reused, so long<br />
as the inputs, outputs, and messaging protocol<br />
are conformant with its description. However,<br />
to support automated discovery and selection<br />
of world-changing Web services, for example,<br />
Web service descriptions must be unambiguous<br />
about what situations will guarantee successful<br />
uses of Web services, and what new situations<br />
will result from those uses. In some scenarios,<br />
a Web service’s behavior may vary with time,<br />
location, user history, pre-existing contractual<br />
commitments, and so on. T<strong>here</strong>fore, descriptions<br />
of such distinctions can quickly become complex.<br />
Moreover, many aspects of Web service use and<br />
management may require knowledge that is not<br />
normally captured in Web service descriptions.<br />
When considering the full range of Web servicerelated<br />
activities, it becomes clear that dealing<br />
with context is a major challenge, requiring far<br />
greater expressiveness and reasoning capabilities<br />
than are supported by the current widely accepted<br />
building blocks of the Web services stack.<br />
Titles of some papers presented at the workshop<br />
are as follows: “Generic framework for the remote<br />
control of high technology instruments based on<br />
composite Web service supporting user context”<br />
(etc) by C. Gravier and J. Fayolle (name their<br />
institution France), “Context management for<br />
adaptive information systems” by C. Cappiello,<br />
M. Comuzzi, E. Mussi, and B. Pernici ((name<br />
their institution Italy) and “A system architecture<br />
for context-aware service discovery” by C.<br />
Doulkeridis and N. Loutas ((name their institution<br />
Greece). The complete list of presented papers is<br />
available at http://www710.univ-lyon1.fr/~dbenslim/<br />
CWS05. The post-proceedings of the workshop<br />
will appear as Electronic Notes in Theoretical<br />
Computer Science (ENTCS) of Elsevier.<br />
The workshop was supported by The European<br />
Co-ordination action for Agent-based<br />
Computing (AGENTLINK), The Object Web<br />
Consortium, The Lyon Research Center for<br />
Images and Intelligent Information Systems,<br />
The ISTASE institute, and The Computer<br />
Science Department of the IUT A Institute of<br />
Lyon I University.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
27
EVENT REPORT<br />
The Eighth Biennial Israeli Symposium<br />
on Foundations of Artificial Intelligence (BISFAI)<br />
Claudia V. Goldman<br />
University of Haifa<br />
Israel<br />
clag@cri.haifa.ac.il<br />
The topic of the first satellite workshop was<br />
recommender systems. Participants enjoyed a full<br />
day of invited talks presenting the state-of-theart<br />
in recommender systems and their prospects.<br />
Invited speakers included Prof. Joseph A.<br />
Konstan from the University of Minnesota, Prof.<br />
Robin Burke from De Paul University, Prof. Luc<br />
Moreau from the University of Southampton,<br />
Prof. Derek Bridge from University College<br />
Cork, Prof. Oded Shmueli from the Technion<br />
and Prof. Daniel Lehmann from the Hebrew<br />
University of Jerusalem. The workshop ended<br />
with demonstrations of several systems including<br />
a guided search assisting a user in an electronic<br />
commerce scenario and a system that interacts<br />
with a user with the purpose of recommending<br />
music and refining its recommendation<br />
mechanisms.<br />
Prof. Martin C. Golumbic, director of the Caesarea<br />
Rothschild Institute (CRI), and Dr. Gideon<br />
Ariely, representing the Ministry of Science<br />
and Technology, opened the main conference.<br />
The first invited talk, from Prof. Yoav Shoham<br />
of Stanford University, discussed the topic of<br />
multi-agent learning and distinguished between<br />
different types of problems and perspectives.<br />
This was followed by two regular talks on game<br />
theory and preferences. The second invited talk,<br />
by Prof. Richard<br />
Korf from UCLA,<br />
enlightened us<br />
about new search<br />
algorithms for<br />
successful queries<br />
in very large data<br />
domains. This was<br />
followed by two<br />
parallel tracks:<br />
one on problem solving and the other on control<br />
of agent behaviour and learning in uncertain<br />
environments. The first day of the symposium<br />
ended with two further parallel sessions: one on<br />
E-commerce agents including talks by the leading<br />
groups in Israel from the Hebrew University and<br />
Bar Ilan University, and the other on information<br />
retrieval.<br />
Prof. Shlomo Zilberstein’s invited talk opened the<br />
second day with a presentation on decentralized<br />
decision-making under uncertainty. The<br />
subsequent regular talks developed this theme,<br />
one with a presentation of a heuristic algorithm<br />
to solve infinite horizon decentralized POMDPs<br />
and the other on learning how to communicate<br />
in decentralized environments. The morning<br />
session continued with two parallel tracks on<br />
collaborative activity in multi-agent systems and<br />
on information and machine learning. Prof. Bart<br />
Selman from Cornell University gave the second<br />
plenary talk on developing automated reasoning<br />
methods for complex scenarios. The day ended<br />
with two additional sessions on cooperative<br />
activity and machine learning. The evening social<br />
activity allowed participants to enjoy an Israeli<br />
dinner, with music and informal discussions in a<br />
great atmosp<strong>here</strong>.<br />
We were honored to start the third day with a<br />
plenary talk on simple deterministic free will by<br />
Prof. John McCarthy of Stanford University. The<br />
rest of the day included a variety of interesting talks<br />
on logic, active databases, constraint satisfaction<br />
problems, swarm intelligence, machine learning<br />
and robotics-laws.<br />
Prof. Helmut Prendinger of the Japanese National<br />
Institute of Informatics gave the second invited<br />
talk, discussing design for affective interaction<br />
with embodied interfaces. This also began the<br />
second satellite<br />
w o r k s h o p ,<br />
whose theme<br />
was affective<br />
computing, and<br />
this continued<br />
through the<br />
rest of the day<br />
closing the whole<br />
symposium. It<br />
included exciting invited talks by Gideon Dror on<br />
learning facial attractiveness, Andrew Ortony on<br />
emotions and intelligent agents, Fabio Pianesi on<br />
a database of kinetic facial expressions and Moshe<br />
Koppel on profiling of anonymous authors.<br />
Prof. Helmut Prendinger<br />
discussed design for<br />
affective interaction with<br />
embodied interfaces<br />
We would like to thank the Caesarea Rothschild<br />
Institute and the Center for Scientific and<br />
Technological Research (ITC-irst) for hosting the<br />
event at the University of Haifa, Israel, <strong>AgentLink</strong>,<br />
the organizers of the symposium Prof. Martin C.<br />
Golumbic, Prof. Uri Schild and Dr. Claudia V.<br />
The 8 th Biennial<br />
Israeli Symposium on<br />
Foundations of AI (BISFAI)<br />
took place from the 27 th to the 30 th<br />
of June 2005 in Haifa Israel. The<br />
symposium included two satellite<br />
workshops: one on recommender<br />
systems and another on affective<br />
computing.Thirty nine reviewed<br />
papers were presented in two<br />
parallel tracks. Each day of the<br />
main symposium started with a<br />
plenary invited talk. A second<br />
plenary talk opened the afternoon<br />
session.<br />
BISFAI is the Israeli main event<br />
on artificial intelligence, which<br />
gathers all the researchers and<br />
students involved in AI research.<br />
This year we also stressed<br />
the participation of industry<br />
at the symposium. Although<br />
the conference is a national<br />
gathering, it attracted international<br />
speakers from Belgium, France,<br />
Italy, Ireland, the UK and the<br />
United States. Abstracts and<br />
the technical program (including<br />
photos!) are available from the<br />
symposium web-site at<br />
www.cri.haifa.ac.il/events/2005/<br />
bisfai05/bisfai05.php.<br />
Goldman, the program committee members and<br />
all the participants. The next BISFAI is planned<br />
to take place in Haifa in 2007.<br />
28<br />
November 2005|<strong>AgentLink</strong> News 19
EVENT REPORT<br />
E4MAS 2005: Environments for Multiagent Systems<br />
Danny Weyns<br />
Katholieke Universiteit Leuven<br />
Belgium<br />
danny.weyns@cs.kuleuven.be<br />
Eric Platon<br />
Sokendai, National Institute of Informatics<br />
Japan<br />
platon@nii.ac.jp<br />
Fabien Michel<br />
Universite de Reims<br />
France<br />
fmichel@leri.univ-reims.fr<br />
The second edition of the International<br />
Workshop on Environments for Multiagent<br />
Systems (E4MAS) was held with the Fourth Joint<br />
Conference on Autonomous Agents and Multi-<br />
Agent Systems (AAMAS 2005) in Utrecht. The<br />
main goal of the E4MAS workshop series is to<br />
provide a discussion forum on the emerging<br />
research topic of Environments for Multiagent<br />
Systems. E4MAS 2005 attracted 25 researchers.<br />
Like the first edition, the workshop was organized<br />
in a highly interactive and discussion oriented<br />
style. The main topics of discussion were: (1)<br />
Role of the environment, scope and models; (2)<br />
Environment engineering; and (3) Multiagent<br />
system applications with an explicit notion of<br />
environment. 7 full and 4 short papers, selected<br />
from 16 submissions, provided input for the<br />
discussions.<br />
Organization<br />
In the first part of the morning session, authors<br />
of papers got a strict time window (5 minutes for<br />
short papers and 10 for full papers) to present their<br />
work. Instead of traditional paper presentations,<br />
speakers were asked to answer a number of<br />
questions on open issues from the perspective<br />
of their work, aiming to boost the subsequent<br />
discussion sessions. During the remaining part of<br />
the morning session, three parallel groups started<br />
the discussion on the main topics mentioned<br />
above. After lunch, discussions were resumed.<br />
During the second part of the afternoon session,<br />
all participants joined a plenary session to discuss<br />
the results of each group and to outline challenges<br />
for future research. The workshop concluded with<br />
an open discussion on ongoing and new efforts<br />
to further develop research on Environments for<br />
Multiagent Systems.<br />
Outcome<br />
Role of the Environment, Scope, and Models<br />
The main focus of the discussion session was<br />
on regulating properties of the environment<br />
in multiagent systems. On the one hand, the<br />
environment enforces rules in multiagent systems,<br />
e.g. laws in a physical domain, or communication<br />
properties in a network; on the other hand<br />
the environment can be exploited to mediate<br />
interactions among agents and interactions<br />
with resources, typical examples are norms and<br />
organizational rules in virtual organizations.<br />
During the discussion, regulating properties of the<br />
environment were related to a deployment model<br />
of multiagent systems that was presented at the<br />
workshop. This model distinguishes three layers<br />
in a multiagent system (from top to bottom):<br />
Multiagent System Application layer; Execution<br />
Platform, and Physical Infrastructure. Different<br />
kinds of regulation pertain to each layer:<br />
• Specific multiagent system application rules:<br />
e.g. norms, interaction laws, etc.<br />
• Execution platform rules: e.g. coordination<br />
rules of middleware, scheduling policies, etc.<br />
• Physical infrastructure rules: communication<br />
bandwidth, topological constraints, physical<br />
laws, etc.<br />
Regulation appears as a feature of the<br />
environment that spans the three layers in which<br />
the rules of lower layers typically constrain rules<br />
at higher layers. This observation is important for<br />
environment engineering.<br />
Environment Engineering<br />
A main point of discussion in this discussion<br />
session was the dependency between environment<br />
engineering and agent engineering. One<br />
viewpoint states that environment engineering<br />
should be independent from agent engineering.<br />
The general idea of this viewpoint is that it should<br />
be possible to develop a library of environmental<br />
mechanisms that designers can reuse when<br />
building multiagent system applications. The<br />
opposite viewpoint states that the environment<br />
cannot be engineered independently from agents.<br />
At the very least the interface between agents<br />
and the environment cannot be uncoupled.<br />
Moreover, if the environment is actively involved<br />
in the mediation of agent interaction, concerns<br />
appear that crosscut agents and the environment,<br />
indicating the need for an integrated engineering<br />
process.<br />
Despite opposing arguments, the two points<br />
of view are compatible. Some concerns of the<br />
environment can be engineered separately, e.g.<br />
internal dynamics of a pheromone infrastructure,<br />
while others concerns must be considered in<br />
relation to agent engineering, e.g. roles and social<br />
structures. Further research is necessary to clarify<br />
these essential engineering issues.<br />
Applications<br />
A number of interesting multiagent system<br />
applications with an explicit notion of<br />
environment were presented at the workshop.<br />
One application showed how web sites can be<br />
considered as the concrete environment for<br />
web users (i.e. the agents) and demonstrated<br />
how this environment can be used to provide<br />
useful advice to users in their current context.<br />
Another industrial application showed how<br />
a physical environment was enhanced with a<br />
virtual environment layer that allows flexible and<br />
distributed coordination among a set of automatic<br />
guided vehicles. These and other applications<br />
highlighted several opportunities arising from an<br />
explicit environment:<br />
• Providing abstraction and hiding low-level<br />
complexity of a physical entity (e.g. fusion of<br />
sensory information)<br />
Research on Environments for Multiagent Systems considers the environment<br />
as a first-order abstraction in multiagent systems. The environment is an explicit part<br />
of a multiagent system that encapsulates essential parts of a multiagent system that<br />
conceptually do not belong to agents, such as infrastructure for communication, the<br />
topology of a spatial domain, or laws of an e-institution. Basically, the environment provides<br />
the surrounding conditions for agents to exist, it offers an abstraction of the external world<br />
to agents in which they can act and interact. This abstraction bridges the conceptual gap<br />
between the agent abstraction and low-level issues, such as details of communication, or<br />
resources access. Moreover, the environment provides an exploitable design abstraction<br />
for building multi-agent systems. The environment can also serve as a medium for agents<br />
to share information and coordinate their behaviour.<br />
November 2005|<strong>AgentLink</strong> News 19<br />
29
EVENT REPORT<br />
• Resolving coordination conflicts (e.g.<br />
access control to resources, support for<br />
transactions)<br />
• Providing extended functionality for<br />
coordination (e.g. exploiting marks in the<br />
environment)<br />
• Simulating future-world state space (e.g.<br />
what-if analysis)<br />
• Maintaining past-world state space (e.g.<br />
enable causal predictions)<br />
Follow-up<br />
Selected and revised papers of the workshop,<br />
complemented with several invited papers will be<br />
collected in a second volume on Environments<br />
for Multiagent Systems that will be published<br />
early 2006 as volume 3830 in the Lecture Notes<br />
in Computer Science series of Springer-Verlag.<br />
In addition, a Special Issue on Environments<br />
for Multiagent Systems for the Journal of<br />
Autonomous Agent and Multiagent Systems<br />
is in preparation. The aim of this volume is to<br />
communicate the consolidated knowledge in<br />
research on environments to the multiagent<br />
system research community in general.<br />
Finally, the organizing committee plans a third<br />
edition of the E4MAS workshop at AAMAS 2006<br />
to be held in Hakodate, Japan. General goals for<br />
E4MAS 2006 are:<br />
• To make a step forward in environment<br />
engineering<br />
• To broaden the perspective on environment<br />
by investigating new uses of environments in<br />
multiagent systems.<br />
Detailed information about the E4MAS<br />
workshop series is available on the web: http://<br />
www.cs.kuleuven.ac.be/~distrinet/events/e4mas/<br />
Bibliography<br />
[1] Environments for Mulitagent Systems: Stateof-the-Art<br />
and Research Challenges, Danny<br />
Weyns, Van Parunak, Fabien Michel, Tom<br />
Holvoet, Jacques Ferber, First International<br />
Workshop on Environments for Multiagent<br />
Systems, New York, 2004. Available at:<br />
http://www.cs.kuleuven.ac.be/~danny/e4mas_<br />
survey_2004.pdf<br />
[2] Environments for Multiagent Systems I,<br />
Danny Weyns, Van Parunak, Fabien Michel<br />
(Eds.), Lecture Notes in Computer Science,<br />
Volume 3374, Springer Verlag, Berlin,<br />
Heidelberg, New York, 2005.<br />
[3] Environments for Multiagent Systems, Danny<br />
Weyns, Michael Schumacher, Mirko Viroli,<br />
Alessandro Ricci, Tom Holvoet, Knowledge<br />
Engineering Review, Cambridge University<br />
Press (to appear). Available at: http://www.<br />
cs.kuleuven.ac.be/~danny/ENV-KER-2005.pdf<br />
International Workshop on Agent-Based Models for Economic Policy<br />
Design (ACEPOL05), June 30 – July 2, 2005, Bielefeld, Germany.<br />
Herbert Dawid, University of Bielefeld, Germany, hdawid@wiwi.uni-bielefeld.de<br />
The International Workshop on Agent-Based<br />
Models for Economic Policy Design (ACEPOL05)<br />
took place from June 30 – July 2, 2005 at the<br />
Center for Interdisciplinary Research (ZiF) at<br />
Bielefeld University. The scientific coordinator<br />
of the meeting was Herbert Dawid (Bielefeld<br />
University). The goal of the workshop was to<br />
assess and further advance the usefulness of agentbased<br />
computational models for the evaluation<br />
and design of economic policy measures.<br />
In particular questions concerning the design and<br />
validation of agent-based models in economic<br />
policy applications as well evaluation of robustness<br />
of simulation results were discussed.<br />
T<strong>here</strong> was strong interest in the meeting from<br />
researchers in several fields. The workshop<br />
programme consisted of 5 plenary and 25 regular<br />
talks, selected from 55 submissions. More than<br />
50 participants from 11 countries attended the<br />
workshop w<strong>here</strong> the background of participants<br />
included economics, management science,<br />
computer science, social science, engineering<br />
and demography. This multi-disciplinary setting<br />
gave rise to a fruitful exchange of ideas between<br />
disciplines and lively discussions.<br />
The five plenary talks addressed different<br />
important issues concerning the potential of<br />
agent-based models for policy evaluation and<br />
policy design. Han LaPoutre (CWI Amsterdam)<br />
discussed the importance of a careful design of<br />
agent-based models in economics. He argued<br />
that modelling choices and parameter values<br />
have to be sound not only from an economic<br />
modelling point of view but also should make<br />
sure that computational pitfalls like premature<br />
convergence are avoided in the simulations. Luigi<br />
Orsenigo (University of Brescia) demonstrated<br />
how industry simulation models which are able<br />
to capture and reproduce past qualitative changes<br />
in the structure of a certain industry (historyfriendly<br />
models) can be used to gain insights<br />
into the medium- and long-term effects of public<br />
policy like antitrust measures, procurement or<br />
investment in basic research. Leigh Tesfatsion<br />
(Iowa State University) dealt with the application<br />
of agent-based methods for market design.<br />
In addition to a discussion of methodological<br />
issues in this area she presented recent work in<br />
a project w<strong>here</strong> agent-based simulation studies<br />
are used to test the reliability of a wholesale<br />
power market platform proposed by the US<br />
Federal Energy Regulatory Commission. The<br />
complementary relationship between agent-based<br />
economics and the currently flourishing field of<br />
experimental economics was discussed by John<br />
Duffy (University of Pittsburgh). He pointed<br />
out that results from human subject experiments<br />
can be very helpful in providing foundations for<br />
behavioural rules used in agent-based models and<br />
In recent years economists, computer<br />
scientists and social scientists have started to<br />
use an agent-based approach as a normative<br />
tool to construct and evaluate market<br />
designs as well as public policy measures<br />
in simulated economic environments. The<br />
agenda of the workshop was to demonstrate<br />
in which economic policy areas agent-based<br />
methods are already succesfully applied<br />
and to deal with important methodological<br />
challenges for agent-based models in this<br />
area of application.<br />
that carrying out agent-based and human subject<br />
experiments in parallel are a promising way to<br />
evaluate the predictive power of agent-based<br />
models in (simple) economic environments. In<br />
the final plenary talk Nigel Gilbert (University<br />
of Surrey) used examples of his own previous<br />
work and work presented at this workshop to<br />
derive a list of guidelines a researcher in agentbased<br />
economics should keep in mind in order<br />
to fruitfully address policy issues with the help of<br />
agent-based models.<br />
Many of the regular talks in the workshop gave<br />
an overview over the various different fields of<br />
economic policy w<strong>here</strong> agent-based models are<br />
currently applied. The issues addressed included<br />
30<br />
November 2005|<strong>AgentLink</strong> News 19
EVENT REPORT<br />
the design of electricity, labour and lottery<br />
markets, the evaluation of the effects of tax<br />
policy, industrial policy, agicultural policy, the<br />
control of the spread of crime and flood control.<br />
Several talks addressed issues of empirical and<br />
experimental validation of agent-based models in<br />
economics highlighting the fact that these issues,<br />
which have been rather neglected for some time,<br />
have attracted a lot of attention in recent years.<br />
Selected papers from this workshop will be<br />
published in a special issue of the Journal of<br />
Economic Behavior and Organization on Agent-<br />
Based Models for Economic Policy Design.<br />
Financial support from Agentlink III and the<br />
Center for Interdisciplinary Research (ZiF) met<br />
the costs of accomodation of all speakers at this<br />
event. The full programme of the workshop<br />
including most papers and presentation files<br />
is available at: http://www.wiwi.uni-bielefeld.de/<br />
~dawid/acepol/index.htm .<br />
The 2005 IEEE/WIC/ACM International Joint Conference<br />
on Web Intelligence and Intelligent Agent Technology<br />
(WI’05 and IAT’05)<br />
Compiegne University of Technology, Compiegne,<br />
France 19-22 Sep 2005<br />
www.hds.utc.fr/WI05/<br />
Pierre Morizet-Mahoudeaux<br />
Universite de Technologie<br />
France<br />
pmorizet@hds.utc.fr<br />
Autonomous entities capable of independent<br />
actions, self-adaptation, communication and<br />
cooperation that characterise living sentient<br />
organisms are one of fundamental inspirations<br />
of the agent computing paradigm. Research<br />
progress on intelligent agents provides for modern<br />
multi-agent systems aiming at solving intricate<br />
problems involving highly complex processes<br />
in heterogeneous environments, described in<br />
different languages, often under uncertainty<br />
and incomplete knowledge. One of such<br />
environments is the World Wide Web, which has<br />
enjoyed an unprecedented expansion in recent<br />
years. The Web is no longer considered merely as<br />
yet another channel of gathering information, but<br />
that it effectively embodies a whole new paradigm<br />
of computation, communication, interaction<br />
and cooperation. Its size, structure, complexity<br />
and behaviour is gradually resembling those<br />
from the real world, in which many elaborate,<br />
independent, and yet interacting processes occur.<br />
A better understanding, and consequently, proper<br />
modelling of such processes is vital to the design,<br />
development and government of the Web and<br />
Web-based computing. A principal approach<br />
to this task, widely known as Web Intelligence<br />
(WI), is based on the theory of autonomous,<br />
distributed, interacting and cooperating<br />
computing agents, which has been a central<br />
subject of the Intelligent Agent Technology<br />
research community. On the other hand, Agent<br />
Technology researchers see the fast-expanding<br />
Web as an exciting environment and a fresh, new<br />
testbed for their theories and methods, as well<br />
as a crucial complement to other models such as<br />
grid or ubiquitous computing.<br />
The IEEE/WIC/ACM WI-IAT Joint Conferences<br />
attempt to provide a platform of exchange<br />
for researchers on many of these problems.<br />
Following WI-IAT’04 held in Beijing, China,<br />
this year WI-IAT’05 was hosted by Compiegne<br />
University of Technology, Compiegne, France.<br />
Out of 328 submissions, 153 papers by 436<br />
authors from 34 countries have been accepted<br />
for WI’05. IAT’05 hosted 134 papers by 322<br />
authors from 25 countries selected from a total<br />
of 305 submissions. During four days, from<br />
19 to 22 September 2005, 6 keynote speeches,<br />
5 workshops, 5 tutorials, Industry and Demo<br />
tracks along with 37 regular sessions have been<br />
held.<br />
The conference’s first keynote speaker, Peter<br />
Schuster (University of Vienna), gave an<br />
insightful lecture on the evolution of autonomous<br />
simple systems and their optimisation, along<br />
with a vivid illustration of described principles<br />
on the Darwinian evolution of DNA and RNA<br />
structures. He stressed the comparison of the<br />
agent computing paradigm to behaviours of<br />
living organisms. The speech was followed by<br />
another keynote lecture on agent agreement<br />
technologies by Nick Jennings from University<br />
of Southampton, who thoroughly outlined<br />
a number of crucial issues pertaining to<br />
communicating networked software agents that<br />
seek to reach agreements by way of negotiations<br />
in a context-dependent environment.<br />
Another key lecturer, Pat Langley from Stanford<br />
University, presented in details the project<br />
ICARUS, an adaptive architecture for physical<br />
agents. It concentrates on exploring intelligent<br />
behaviours in physical domains and encompasses<br />
many profound aspects of cognition, perception,<br />
inference, and planning actions. Wray Butine<br />
(Helsinki Institute for Information Technology),<br />
in turn, introduced to the audience many exciting<br />
facets of the world of web search engines, with a<br />
particular description of those developed with<br />
open sources.<br />
The last day of the conference has been marked<br />
with two keynote speeches. Henry Lieberman<br />
(MIT Media Laboratory), in picturesque detail,<br />
presented a novel approach to providing online<br />
assistance, expertise transfer and understanding<br />
users’ language. Patrick Doherty from Linkšping<br />
University gave a vivid and amusing presentation<br />
of his unmanned aerial vehicle project, which<br />
attempts to perform advanced tasks such as<br />
automatic monitoring and surveillance using<br />
approximate reasoning with complex concepts.<br />
A broad range of issues, problems, approaches<br />
and methods were presented and widely discussed<br />
at the conference’s sessions. Most notably, a<br />
paper describing myopic approaches to agent’s<br />
communication by Raphen Becker, Victor Lesser,<br />
and Shlomo Zilberstein and another dealing with<br />
web mining using rough sets theory by Chi Lang<br />
Ngo and Hung Son Nguyen were awarded best<br />
papers for IAT’05 and WI’05 tracks, respectively.<br />
The Best WI/IAT’05 Demo Award went to David<br />
Sislak, Martin Rehak, and Michal Pechoucek<br />
for their multi-agent platform with advanced<br />
simulation and visualisation support (for more<br />
information on this work, see the article on page 9<br />
in <strong>AgentLink</strong> News 17). Robert Nolker and Lina<br />
Zhou were awarded the WI’05 Best Student Paper<br />
Award for their article on a method of identifying<br />
member roles in online communities.<br />
Apart from its extensive scientific program, the<br />
conference’s organizers came up with pleasant<br />
social activities, including visits to the historic<br />
city hall and the Pierrefonds medieval castle.<br />
The next year IEEE/WIC/ACM WI-IAT<br />
conference will be held in Hong Kong, together<br />
with the IEEE International Conference Data<br />
Mining (ICDM), on 18-22 December 2006. For<br />
more information, please visit http://www.comp.<br />
hkbu.edu.hk/~wii06/<br />
November 2005|<strong>AgentLink</strong> News 19<br />
31
EVENT REPORT<br />
Third Workshop on Agents Applied in Health Care,<br />
at the 19th International Joint Conference on Artificial Intelligence<br />
(IJCAI 2005)<br />
Antonio Moreno, University Rovira i Virgili, Spain, antonio.moreno@urv.net<br />
The 3rd workshop on Agents Applied in Health<br />
Care was held on July 31st at Edinburgh,<br />
Scotland, as part of the 19th International Joint<br />
Conference on Artificial Intelligence. It was<br />
organised by Antonio Moreno (Univ. Rovira<br />
i Virgili, Tarragona, Spain), Ulises Cortes<br />
(Technical University of Catalonia, Barcelona,<br />
Spain),John Nealon (Oxford Brookes University,<br />
Oxford, UK) and John Fox (Cancer Research<br />
UK, London, UK). Previous editions of the<br />
workshop had been held at ECAI 2002 (Lyon,<br />
France) and ECAI 2004 (Valencia, Spain). The<br />
main topic of these meetings is the use of agent<br />
technology and multi-agent systems in all aspects<br />
of health care. The workshop organisers are also<br />
the main promotors of the <strong>AgentLink</strong> TFG on<br />
Agents Applied in Health Care.<br />
This, the third workshop in the series, was<br />
attended by 20 delegates, with eight talks<br />
coming from countries as diverse as Italy, Czech<br />
Republic, The Netherlands, Germany, United<br />
Kingdom and Spain. The main topics covered by<br />
the papers were: the use of agents in diagnostics<br />
and monitoring tasks, the management of health<br />
care of old or disabled citizens, the simulation<br />
and optimisation of medical processes, and the<br />
retrieval of medical information from Internet.<br />
T<strong>here</strong> were three papers in the diagnostics and<br />
monitoring field. Aoda Valls (Univ. Rovira i<br />
Virgili, Spain) presented the PalliaSys project,<br />
in which a multi-agent system that monitors the<br />
evolution of palliative patients in a hospital is<br />
currently being developed.<br />
Lenka Lhotska (Czech Technical Univ., Czech<br />
Republic) explained the first steps in the design of<br />
ADAM (Agent architecture for Diagnostics And<br />
Monitoring in medicine). Beatriz Lopez (Univ.<br />
of Girona, Spain) gave a report on the Masictus<br />
multi-agent system, which gives support to the<br />
diagnosis of acute stroke and the coordination of<br />
ambulance services.<br />
Roberta Annicchiarico (Fondazione Santa Lucia,<br />
Italy) presented the latest developments in the<br />
on-going e-tools project, in which an agentcontrolled<br />
wheel chair is being developed to<br />
improve the quality of life of disabled people.<br />
Richard Hill (Sheffield Hallam Univ., UK) argued<br />
the need of having formal tools to represent<br />
appropriately health care workflows within multiagent<br />
systems.<br />
Loes Braun (Inst. for Knowledge and Agent<br />
Technology, The Netherlands) described an<br />
agent-based system that allows doctors to retrieve<br />
from Internet the medical information that may<br />
be relevant to treat their particular patients.<br />
Reiner Herrler (Univ. of Würzburg, Germany)<br />
described a hospital simulation kit based on<br />
the SeSAM simulation tool. Finally, Pancho<br />
Tolchinsky (Technical Univ. of Catalonia, Spain)<br />
explained the latest work in the Carrel project,<br />
in which argumentative agents are being used to<br />
increase the number of available human organs<br />
for transplant.s<br />
The workshop also featured an invited talk by<br />
the IJCAI conference chair, Fausto Giunchiglia<br />
(Univ. of Trento. Italy). In his talk he described<br />
the concept of context ontologies, and their<br />
representation in the language C-OWL.<br />
The organisers would like to thank the financial<br />
contribution of <strong>AgentLink</strong> III, that kindly<br />
provided some funding to partially cover the<br />
attendance costs of 3 PhD students that were the<br />
first authors of accepted papers.<br />
An special issue on agents applied in health care<br />
of the IEEE Intelligent Systems journal, edited<br />
by Antonio Moreno, is planned for next year.<br />
It will include revised and extended versions<br />
of papers presented at the workshop, as well as<br />
papers contributed by other researchers in the<br />
field. Papers will have to be submitted for review<br />
around. May 2006.<br />
Agents in Space<br />
Michael Fisher<br />
University of Liverpool<br />
United Kingdom<br />
M.Fisher@csc.liv.ac.uk<br />
July 2005 saw a one-day symposium on<br />
“Agents in Space” held in Liverpool, UK, and<br />
jointly organized by the Research Institute for<br />
Advanced Computer Science (RIACS) at the US<br />
National Aeronautics and Space Administration<br />
(NASA), space systems company SciSys Ltd,<br />
and the Departments of Computer Science and<br />
Engineering at the University of Liverpool. The<br />
symposium co-chairs were Professor Michael<br />
Fisher (Head of the Liverpool Verification<br />
Laboratory) and Dr. Roger Ward (Head of<br />
On-board Software at SciSys). The aims of the<br />
symposium were to highlight the use of agentbased<br />
software systems in space exploration and<br />
to explore the use of formal methods to verify<br />
that software systems in aerospace applications<br />
would work as intended.<br />
With the development of deep space and planetary<br />
exploration, communications between spacecraft<br />
and mission control on earth are subject to<br />
extended delays. These delays could have serious<br />
impacts on operations, and so t<strong>here</strong> is increasing<br />
need for spacecraft to act autonomously, that is,<br />
without constant reference to mission control.<br />
Such autonomy is often naturally captured by<br />
the notion of an agent. However, with the use<br />
of autonomous agents in safety and missioncritical<br />
situations come several challenges. This<br />
symposium concerned three of these, namely:<br />
how to control autonomous agents; how to<br />
organise effective collaboration between humans<br />
and agents; and how to verify the behaviour of<br />
autonomous agents. The symposium brought<br />
together some of the leading experts in agents,<br />
autonomy in space, and formal verification, with<br />
the aim of stimulating further developments in<br />
this exciting area.<br />
At the symposium, Dr Maarten Sierhuis from<br />
NASA spoke on teamwork between robots and<br />
humans on Mars or the Moon, while Dr Willem<br />
Visser spoke about techniques for verifying<br />
aerospace software. In addition, scientists from<br />
SciSys, Cranfield University, and the Universities<br />
of Durham, Southampton and York in the UK<br />
gave talks on related topics during the day. This<br />
novel and exciting area is set for significant increase<br />
through the strategic directions being taken by<br />
various space agencies: -- NASA; the European<br />
Space Agency, ESA; and the Japanese Aerospace<br />
Exploration Agency, JAXA. Verification of space<br />
software, particularly in mission critical or safety<br />
critical situations, is vital.<br />
The symposium programme is available from:<br />
http://www.csc.liv.ac.uk/~michael/agents-in-space05.<br />
html<br />
32<br />
November 2005|<strong>AgentLink</strong> News 19
EVENT REPORT<br />
First International Workshop on Security and Trust Management<br />
(STM 2005)<br />
15 September 2005<br />
Milan, Italy<br />
Sjouke Mauw<br />
Eindhoven University of Technology<br />
The Netherlands<br />
sjouke@win.tue.nl<br />
In September 2005 the STM working group<br />
of ERCIM organized a successful workshop on<br />
Security and Trust Management in Milan, Italy.<br />
The initiative for this workshop was taken at the<br />
founding meeting of the STM working group<br />
in January 2005 in order to fulfill some of the<br />
goals of the working group to bring researchers<br />
together and stimulate scientific discussion.<br />
The organization of the workshop, co-located<br />
with ESORICS’05 (European Symposium On<br />
Research In Computer Security), was in the<br />
hands of Fabio Martinelli, Pierangela Samarati<br />
(general co-chairs), Valerie Issarny, Sjouke Mauw<br />
(PC co-chairs) and Cas Cremers (vice-chair).<br />
Due to the high number of submissions (36)<br />
the PC could select nine papers of high quality,<br />
which were presented at the workshop and will<br />
appear in a special issue of ENTCS (Electronic<br />
Notes in Theoretical Computer Science. Is t<strong>here</strong><br />
a also series or theme). Sponsorship was received<br />
from ERCIM (write out: the European Research<br />
Consortium for Informatics and Mathematics)<br />
and <strong>AgentLink</strong>.<br />
The workshop had a broad scope, ranging from<br />
cryptography and formal methods to physical<br />
security. The link between the major topics,<br />
security and trust, was made by keynote speaker<br />
Prof. Dieter Gollmann (Technische Universität<br />
Hamburg, Harburg), who challenged the<br />
audience with his presentation entitled “Why<br />
Trust is Bad for Security”. He argued that<br />
the notion of trust has many different (often<br />
conflicting) interpretations, while t<strong>here</strong> is a need<br />
for clarity and precision.<br />
The first session of the workshop covered smart<br />
dust security, a formal approach to multiparty<br />
contract signing and the development of security<br />
models for mobile agent security. In the second<br />
session, the notions of credit and responsibility<br />
were formalized, a new scheme for trapdoor<br />
hash functions was presented, and access control<br />
mechanisms based on trust were studied.<br />
The third and final session was completely<br />
dedicated to trust management. It covered<br />
the extension of role based trust m anagement<br />
languages with non-monotonicity, assigning<br />
trust values to metadata, and a new authorization<br />
strategy for distributed environments.<br />
Given the success of this first STM workshop,<br />
and the interest that it raised, we believe that<br />
this will be the first of a series of successful<br />
workshops.<br />
The 8th edition of the<br />
European Agent Systems Summer School<br />
(EASSS’06)<br />
Université de Savoie, Annecy, France, 17-21 July 2006<br />
http://www.esia.univ-savoie.fr/easss06<br />
Researchers and teachers in this field are invited to submit a proposal for a course. A typical course is<br />
4 hours in total, but deviations (e.g. 2 or 6 hour) are possible. Programs of previous editions of EASSS<br />
can be found at the web page.<br />
Tutors are encouraged to submit a propsosal for a course to Mehdi Dastani (mehdi@cs.uu.nl), including<br />
a 1 page description mentioning experience of the tutor, level of the course (beginners, advanced),<br />
duration and needs for equipment.<br />
Deadline for Proposals: December 17, 2005<br />
Acceptance Notification: January 23, 2006<br />
Submission of final Teaching Material (< 26 pp): April 23, 2006<br />
November 2005|<strong>AgentLink</strong> News 19<br />
33
Semantics<br />
Semantics<br />
PROJECT REPORT<br />
ATHENA:<br />
Advanced Technologies for interoperability of<br />
Heterogeneous Enterprise Networks and their Applications<br />
Klaus Fischer<br />
Deduction and Multiagent<br />
Systems, DFKI GmbH<br />
Germany<br />
Klaus.Fischer@dfki.de<br />
ATHENA is an Integrated Project sponsored by the<br />
European Commission in support of the Strategic<br />
Objective “Networked businesses and government”<br />
as set out in the IST 2003-2004 Work programme<br />
of FP6 (Framework Programme). Building upon<br />
an ambitious Vision Statement that “By 2010,<br />
enterprises will be able to seamlessly interoperate<br />
with others”, ATHENA aims to make a major<br />
contribution to interoperability by identifying<br />
and meeting a set of inter-related business,<br />
scientific & technical, and strategic objectives. The<br />
ATHENA programme of work is geared towards<br />
producing results that span the full spectrum of<br />
interoperability from technology components<br />
to applications and services, from research &<br />
development to demonstration & testing and from<br />
training to evaluation of technologies for social<br />
impact. In ATHENA, different Research and<br />
Development projects (inside Action Line A) are<br />
executed in close synergy and collaboration with<br />
Community Building activities (inside Action<br />
Line B), thus ensuring that solutions for multidisciplinary<br />
research challenges are of optimal<br />
industrial relevance leading to a broad uptake<br />
by the end user. Figure 1 depicts the structure of<br />
ATHENA and the relation between the projects<br />
running in the ATHENA program.<br />
ATHENA is committed to creating a long-term<br />
impact for advancing interoperability which is<br />
mainstream, inclusive and has critical mass. To this<br />
end, ATHENA is initiating an open, neutral and<br />
independent Enterprise Interoperability Centre<br />
(EIC) in which all stakeholders, both in private<br />
and public sectors, are invited to participate.<br />
Fig 1: ATHENA’s Strcuture<br />
The establishment of the EIC<br />
is an action of the European<br />
Union’s updated eEurope 2005<br />
Action Plan for implementing<br />
the Lisbon Strategy. The EIC<br />
Executive Summary is available<br />
as part of the public consultation<br />
with interoperability stakeholders<br />
in respect of the establishment of<br />
the EIC.<br />
The ATHENA consortium<br />
currently comprises of 19 leading<br />
organisations in research, academia, industry and<br />
other stakeholder communities including SMEs,<br />
working collaboratively in pursuit of a common<br />
set of objectives in interoperability. The ATHENA<br />
programme of work represents one of the largest<br />
investments made in interoperability through<br />
the co-financing arrangements of the European<br />
Commission’s research programme (the total<br />
committed project budget is 26.6 million Euro,<br />
including 14.4 million Euro financing by the EC).<br />
ATHENA started on the 1st of February 2004 and<br />
is currently contracted for 3 years, with a 5 year<br />
time horizon and planning.<br />
ATHENA’s Research and Development<br />
projects<br />
Research and Development projects, under Action<br />
Line A, will provide research results that fit into<br />
an Interoperability Framework. The General<br />
ATHENA Framework is a conceptual model<br />
which looks at different aspects of interoperability,<br />
combining and leveraging business knowledge<br />
Enterprise A<br />
Business<br />
Knowledge<br />
ICT Systems<br />
Enterprise B<br />
Business<br />
Knowledge<br />
ICT Systems<br />
Fig 2: General Athena Framework<br />
and modeling, Information and Communication<br />
Technologies (ICT) platforms and Semantics<br />
research.<br />
This general framework is represented via three<br />
main layers: business, knowledge and ICT<br />
(Information and Communication Technology)<br />
solutions, as illustrated in Figure 2.<br />
In order for organizations to successfully<br />
conduct business, both internally and with<br />
other organisations, they must be aware of the<br />
interdependency among the three layers as<br />
mentioned in Figure 2. In addition, Athena has<br />
added a fourth dimension regarding the semantics<br />
that cut across the three layers. It was introduced<br />
to promote an understanding of expression of<br />
similarities and differences among businesses, the<br />
way they operate, and their software solutions.<br />
Research and Development has the primary<br />
objectives of:<br />
• Studying, analysing, enabling and managing<br />
the business processes and the operations<br />
of individual networked enterprises and<br />
between collaborating enterprises;<br />
• Studying, analysing, and managing<br />
organizational roles, skills, competencies and<br />
knowledge assets of individual networked<br />
enterprises and between collaborating<br />
enterprises;<br />
• Designing and developing an Interoperability<br />
Framework providing integration on a<br />
conceptual, application and technical level;<br />
• Studying, analysing and improving techniques<br />
to increase interoperability between<br />
applications, data and communication<br />
components;<br />
• Studying, analysing, and developing concepts,<br />
tools, and methods to leverage and share<br />
enterprise knowledge.<br />
34<br />
November 2005|<strong>AgentLink</strong> News 19
PROJECT REPORT<br />
Figure 3 presents ATHENA’s operational vision<br />
to interoperability. In this context three activities<br />
of ATHENA are most interesting for agent<br />
technologies:<br />
1) Planned and Customisable Service-<br />
Oriented Architectures (SOA): SOA are<br />
being viewed as the next wave of technology<br />
to impact the computing landscape. They do<br />
this by enabling the utilisation of distributed<br />
components by allowing software vendors<br />
to provide not only applications to market,<br />
but a suite of services. These services can be<br />
utilised by a wider audience and charged<br />
for through an access or usage business<br />
model. This project is concerned with the<br />
development of service-oriented solutions<br />
that can be more easily planned and then later<br />
customised when they are being deployed.<br />
This is intended to provide better industry<br />
focused solutions that can be adapted for<br />
deployment into client environments. This<br />
is to directly counter the problem of high<br />
costs of customisation and integration of<br />
highly generic industry solutions. The main<br />
objectives of this project are to develop<br />
modelling and specification systems that<br />
accurately express services and serviceoriented<br />
architectures and particularly<br />
assist in the planning of solutions and the<br />
marking of intended customisations for<br />
the better deployment of a solution into<br />
a wide range of client environments; to<br />
develop technologies that enable the easier<br />
composition of services and the brokering,<br />
mediation and ultimately the negotiation<br />
of pre-specified but customisable services;<br />
to develop an execution framework for<br />
Company<br />
Company B<br />
planned and customisable service-oriented<br />
architectures utilising user scenarios from<br />
the ATHENA community building<br />
activities; and to evaluate the developed<br />
technologies and execution framework in<br />
proto-typical implementations.<br />
2) Model-Driven and Adaptive Interoperability<br />
Architectures: the main objective activity is<br />
to provide new and innovative solutions for the<br />
problem of sustaining interoperability through<br />
change and evolution, by providing dynamic<br />
and adaptive interoperability architecture<br />
approaches. The objective is to advance state of<br />
the art in this field, by applying the principles of<br />
model-driven, platform independent architecture<br />
specifications, and dynamic and autonomous<br />
federated architecture approaches. This includes<br />
the following sub-goals: to support requirements<br />
and validate solutions for the involved sectors<br />
specified in the ATHENA community<br />
building activities; to provide meta models and<br />
methodologies for interoperability architecture<br />
solutions; to evaluate and extend multiple<br />
adaptive autonomous and federated architecture<br />
approaches, including agent and peer-to-peer<br />
technologies and the model-driven architecture<br />
approach; to provide support for non-functional<br />
interoperability aspects, through a modeldriven<br />
approach: to apply the use of ontologies<br />
and semantics in model and service registry/<br />
repositories for better semantic interoperability;<br />
and to provide semantic mapping and mediation<br />
technologies that provide executable frameworks<br />
and support for active models.<br />
3) Knowledge Support and Semantic<br />
Mediation Solutions: This project activity<br />
examines the development of methods and<br />
Service Wrappers / Interoperability Management<br />
Evaluation & Negotiation of Available<br />
Functionality<br />
Execution Environment<br />
Service Interconnection Bus<br />
Fig 3: ATHENA’s operational vision to interoperability<br />
Repository<br />
Existing Enterprise Applications<br />
tools for enterprise ontology management,<br />
with a focus on supporting enterprise<br />
knowledge integration and interoperability<br />
for enterprises and software applications.<br />
The main objective is to build an ontology<br />
environment aimed at supporting the<br />
integration of the different sorts of knowledge<br />
that can be found in an enterprise. Enterprise<br />
knowledge can be divided in two main<br />
categories. The first category comprises<br />
the knowledge represented by all sorts of<br />
documents: from technical reports to emails,<br />
from scientific papers to circuit blueprints;<br />
this category is mainly conceived to be<br />
used by human users. A second category is<br />
represented by formal symbolic knowledge,<br />
e.g., modelled by using semantic nets or<br />
description logics, stored in an ontology,<br />
which is mainly organized to be exploited<br />
by a computer. The proposed ontology<br />
enterprise environment, based on ontology<br />
knowledge, will be used in two main<br />
enterprise areas: to support the integration<br />
of documental knowledge and to achieve<br />
enterprise software interoperability. Both of<br />
them can be achieved by applying semantic<br />
annotation techniques, although our focus<br />
will be on the latter. On the technological<br />
level, semantic annotation will be used to<br />
associate meaning to computable services<br />
(e-services, agents, software components,<br />
etc.) selected to be used in advanced software<br />
architectures, such as management domain<br />
architecture, SOA, distributed application<br />
integration. Semantic annotation is one of<br />
the key enabling technologies to implement<br />
a number of solutions for achieving<br />
interoperability, e.g., from service discovery<br />
to matchmaking, to the linking of Business<br />
Processes to computable services.<br />
In the first two activities, agent technologies<br />
are directly investigated in how far they can<br />
contribute to the problems the projects deal with.<br />
Most important are the questions, how agent<br />
technologies can be smoothly integrated with<br />
SOA and how the MDA methodology fits with<br />
an agent-oriented approach to system design.<br />
Because tools for the design of BDI agents are<br />
available that allow to describe the agents and<br />
even teams of agents using UML style models,<br />
BDI agents seem to be the most promising way<br />
to integrate agent-oriented software design with<br />
the model-driven development methodology that<br />
MDA suggests. Still it remains an open question<br />
in how far the meta-models ATHENA comes up<br />
with relate to a meta-model of BDI agents and<br />
in how far the idea of model transformations can<br />
be used to improve the way in which multiagent<br />
systems are designed.<br />
More details on ATHENA can be found at<br />
http://www.athena-ip.org and by sending e-mail<br />
to info@athena-ip.org or, if agent technologies are<br />
concerned, to Klaus Fischer (Klaus.Fischer@dfki.de).<br />
November 2005|<strong>AgentLink</strong> News 19<br />
35
SITE REPORT<br />
Business Decision Making<br />
using Multi-Agent Systems<br />
Nadia Yakounina<br />
Magenta Technology<br />
United Kingdom<br />
yakounina@magenta-technology.com<br />
The complexity of modern business operations<br />
has now increased to the point that without<br />
moving to new dynamic, adaptive systems,<br />
business people face a difficult challenge in<br />
terms of enterprise decision making. Either they<br />
can rely on software systems that ignore the<br />
true complexity of their operations in order to<br />
automate high volume business processes or they<br />
can try to address complexity by employing more<br />
and more smart, experienced people. Neither<br />
alternative is scalable or robust, and certainly<br />
will not provide a competitive advantage. The<br />
smartest software people can reach out for now<br />
are “optimizers” that use elaborate “business<br />
rules”, neither of which can react or respond<br />
effectively under volatile, real-time pressure.<br />
T<strong>here</strong>fore, in many instances, people still need<br />
to get involved to re-work the solution and make<br />
the trade-offs necessary for a practical solution.<br />
To enable a major step-change in enterprises,<br />
Magenta’s approach is focused on combining<br />
several key elements: 1) Continuous Realtime<br />
Planning 2) Adaptation 3) Team-based<br />
Decision-making.<br />
1) Continuous Real-time Planning<br />
The main objective in moving from a cascade/<br />
batch planning environment is to compress<br />
the time taken to complete the Plan-Commit-<br />
Execute cycle. The longer the delay between<br />
planning and starting a task, the higher the<br />
risk that the plan will become stale and nolonger<br />
effective, and hence require some form<br />
of exception handling. Consequently Magenta<br />
has focused on Event-Driven systems that<br />
incrementally react to commit changes to a<br />
complex Schedule. The Magenta approach can<br />
understand Scheduling Events from a flow of<br />
messages.<br />
2) Adaptation<br />
When running practical business systems, the<br />
core knowledge of the business process needs<br />
first to be tuned to fit actual use, and then it<br />
needs to be adapted to respond to the evolving<br />
business needs. By using semantic network<br />
technology, Magenta can describe, capture and<br />
compare information and knowledge on-the-fly<br />
to grow and alter the software over time.<br />
3) Team-base Decision-Making<br />
As business people who are geographically<br />
dispersed typically work asynchronously when<br />
working across a single process, enabling them to<br />
share resources is not straightforward. The endto-end<br />
business process must first be described<br />
in detail, and stored in an industry accepted<br />
standard notation, such as the Business Process<br />
Modelling Notation (BPMN). Each process<br />
typically has a number of layers of decisionmaking<br />
authority and consultation, and should<br />
About Magenta<br />
Magenta Technology is the leading<br />
developer of enterprise-ready multiagent<br />
and semantic web software,<br />
to enable the adaptive enterprise,<br />
utilising a unique combination<br />
of core technologies - Multi-<br />
Agent systems, Semantic Web<br />
technologies and Java/ J2EE. The<br />
UK-based company offers a range<br />
of enterprise-ready products that<br />
release significant value trapped<br />
within supply chain operations;<br />
maximising revenue, lowering<br />
costs and bringing added value to<br />
business activities.<br />
be mapped to the software workflow. Users also<br />
require access to strong visualization tools, alerts,<br />
and access to reports to monitor and understand<br />
how well a process is performing.<br />
Inside the Event-Driven Architecture<br />
Within the Magenta solution, the agents reach<br />
decisions by negotiating with each other, to<br />
achieve the best individual outcome whist<br />
adhering to enterprise goals to maximise value<br />
to the whole system. For example, given the task<br />
of scheduling the delivery of cargo to shipping<br />
vessels, agents representing each cargo should<br />
only seek to minimise their individual cost of<br />
delivery if this does not result in an increase in<br />
cost of delivering the full set of cargo. Figure<br />
2 illustrates the high-level architecture for a<br />
Magenta Technology system to schedule loads<br />
to vehicles for logistics transport planning.<br />
Figure 1: Decision Support Technology Sweet spots. Tools and approaches in software market today force business<br />
people to either simplify the description of their business or increase manual activity to address complexity. Magenta’s<br />
approach resolves this dilemma.<br />
The system utilises XML datasets that describe:<br />
events (orders, cancellations, breakdown of<br />
vehicles etc); the assets used to fulfil orders<br />
(trucks, vans etc); the geography of the transport<br />
network; and perhaps an initial schedule.<br />
This information is used to construct a scene<br />
that represents the real world problem to be<br />
solved, and which is an instance of an ontology<br />
that describes the Business Process situation.<br />
The ontology describes the business process,<br />
including physical attributes of vehicles,<br />
industry regulations, and preferences used in<br />
36<br />
November 2005|<strong>AgentLink</strong> News 19
decision making such as preferred routes, costs<br />
of vehicles etc. The scene-forming component<br />
has both an API and a toolset for managing the<br />
ontology (OMT). The scenes are stored in a<br />
database which provide persistence and a history<br />
of the whole decision making process.<br />
Armed with the business process ontology and<br />
a populated scene of that ontology, a swarm of<br />
agents operating in a virtual market negotiate the<br />
allocation of orders to vehicles. The virtual market<br />
is powered by a multi-agent engine that provides<br />
messaging protocols for the agents. In the agent<br />
swarm, every order and vehicle is represented by an<br />
agent that seeks an optimal plan, thus identifying<br />
the best result for that order and vehicle. To<br />
achieve this, the agents have access to a toolbox of<br />
capabilities, such as distance calculators and key<br />
performance indicator (KPI) calculators. A user<br />
interface (UI) component provides the human<br />
planner with access to the system and reporting<br />
capabilities. The whole system is built on a J2EE<br />
architecture, which gives platform operating<br />
system and database independence, robustness,<br />
reliability, integration with existing systems, and<br />
scalability out of the box.<br />
Figure 2 Magenta Multi Agent System Architecture for Transport Logistics Planning<br />
Results from this system have been compared<br />
with results from traditional optimisation tools.<br />
The quality of the schedule produced by the<br />
multi-agent system is comparable to these tools,<br />
but exhibits the following benefits:<br />
• New orders or cancellations can be scheduled<br />
in near real-time without having to re-plan<br />
the whole schedule. Real time, incremental<br />
planning enables a business to plan new<br />
orders just before loading the vehicle;<br />
• Changes to the business process can be<br />
quickly implemented in the ontology, which<br />
is independent of the agent code;<br />
• The ontology can be changed by a domain<br />
expert, rather than a computer systems<br />
expert;<br />
• The system can be readily used to do scenario<br />
planning because of its incremental planning<br />
capability;<br />
• The system does not impose hard rules on<br />
the plan – all preferences can be stressed, just<br />
like real life;<br />
• The system can cope with very high levels of<br />
complexity and contradictory preferences.<br />
This makes multi-agent systems ideal for business<br />
decision-making in fast moving, complex,<br />
evolving, and uncertain environments.<br />
EUMAS<br />
Brussels, Belgium<br />
7-8 December, 2005<br />
In the last 15 years we have seen a significant increase of interest in agent-oriented technology. This<br />
field is now set to become one of the key technologies in the 21st century and will underpin much of the<br />
next generation of computing that seeks to address issues in Ambient Intelligence, Pervasive Computing,<br />
Grid Computing, the Semantic Web, E-Commerce and many other areas. It is t<strong>here</strong>fore crucial that<br />
both academics and industrialists within Europe have access to a forum at which current research and<br />
application issues are presented and discussed. The aim of this third European Workshop on Multi-Agent<br />
Systems is to encourage and support activity in the research and development of multi-agent systems, in<br />
academic and industrial European efforts.<br />
For Details see http://como.vub.ac.be/eumas2005/<br />
November 2005|<strong>AgentLink</strong> News 19<br />
37
BOOKS<br />
Agent Intelligence through<br />
Data Mining<br />
Andreas L. Symeonidis<br />
Pericles A. Mitkas<br />
ISBN: 0-387-24352-6<br />
2005<br />
http://www.springer.com/sgw/cda/frontpage/<br />
0,,5-147-22-46687939-0,00.html<br />
“Agent Intelligence<br />
through Data<br />
Mining” offers a selfcontained<br />
overview<br />
of a relatively young<br />
but important<br />
area of research:<br />
the intersection of<br />
agent technology<br />
and data mining.<br />
This intersection<br />
is leading to<br />
considerable advancements in the area of<br />
information technologies, drawing the<br />
increasing attention of both research and<br />
industrial communities. It can take two<br />
forms: a) the more mundane use of intelligent<br />
agents for improved data mining and b) the<br />
use of data mining for smarter, more efficient<br />
agents. The book focuses on the second<br />
approach.<br />
Knowledge, hidden in voluminous data<br />
repositories routinely created and maintained<br />
by today’s applications, can be extracted by<br />
data mining. The next step is to transform<br />
this discovered knowledge into the inference<br />
mechanisms or simply the behavior of agents<br />
and multi-agent systems. Agent Intelligence<br />
through Data Mining addresses this issue, as<br />
well as the arguable challenge of generating<br />
intelligence from data while transferring it<br />
to a separate, possibly autonomous, software<br />
entity. This book contains a methodology,<br />
tools and techniques, and several examples<br />
of agent-based applications developed with<br />
this approach. This volume focuses mainly<br />
on the use of data mining for smarter, more<br />
efficient agents.<br />
Agent Intelligence through Data Mining<br />
is designed for a professional audience of<br />
researchers and practitioners in industry.<br />
This book is also suitable for graduate-level<br />
students in computer science.<br />
Multi-Agent Programming<br />
Edited by R.H Bordini, M. Dastani,<br />
J. Dix and A. El Fallah Seghrouchni<br />
ISBN: 0-387-24568-5<br />
http://www.springer.com/sgw/cda/<br />
frontpage/0,11855,3-0-22-53996849-0,00.<br />
htmlreferer=www.springer.com%2F0-387-<br />
24568-5<br />
M u l t i - A g e n t<br />
Programming is an<br />
essential reference<br />
for anyone interested<br />
in the most up-todate<br />
developments in<br />
MAS programming.<br />
P r o g r a m m e r s ,<br />
researchers, and<br />
graduate students will<br />
find this text unique<br />
in its presentation of<br />
the concepts and principles of this fast-growing<br />
field. While previous research has focused<br />
on the development of formal and informal<br />
approaches to analyse and specify Multi-Agent<br />
Systems, this book focuses on the development<br />
of programming languages and tools which<br />
not only support MAS programming, but also<br />
implement key concepts of MAS in unified<br />
frameworks.<br />
Part I describes four approaches that are based on<br />
computational logic or process algebra--Jason,<br />
3APL, IMPACT, and CLAIM/SyMPA. These<br />
programming languages have formal semantics<br />
and use heavy machinery based on formal<br />
methods, but also provide working platforms<br />
for the development of multi-agent systems.<br />
Part II presents agent languages and platforms<br />
that extend or are based on Java--JADE, Jadex,<br />
and JACK TM . Although these have no formal<br />
semantics, the languages are well documented<br />
and the platforms provide a variety of tools<br />
that have been extensively used in practice. Part<br />
III provides two significant industry specific<br />
applications--The DEFACTO System for<br />
coordinating human-agent teams for the future<br />
of disaster response, and the ARTIMIS rational<br />
dialogue agent technology. The book also features<br />
seven appendices, summarising each of the agent<br />
programming languages, hence facilitating<br />
comparison of the approaches. In particular,<br />
Appendix A describes the criteria used for<br />
comparing the agent languages and platforms.<br />
Combinatorial Auctions<br />
Edited by Peter Cramton, Yoav<br />
Shoham, and Richard Steinberg<br />
ISBN 0-262-03342-9<br />
January 2006<br />
http://mitpress.mit.edu/promotions/books/<br />
FL20050262033429<br />
The study of<br />
c o m b i n a t o r i a l<br />
a u c t i o n s - -<br />
auctions in which<br />
bidders can bid<br />
on combinations<br />
of items or<br />
packages--draws<br />
on the disciplines<br />
of economics,<br />
o p e r a t i o n s<br />
research, and computer science. This<br />
landmark collection integrates these three<br />
perspectives, offering a state-of-the art<br />
survey of developments in combinatorial<br />
auction theory and practice by leaders in the<br />
field.<br />
Combinatorial auctions (CAs), by allowing<br />
bidders to express their preferences more<br />
fully, can lead to improved economic<br />
efficiency and greater auction revenues.<br />
However, challenges arise in both design<br />
and implementation. Combinatorial<br />
Auctions addresses each of these challenges.<br />
After describing and analyzing various CA<br />
mechanisms, the book addresses bidding<br />
languages and questions of efficiency. Possible<br />
strategies for solving the computationally<br />
intractable problem of how to compute the<br />
objective-maximizing allocation (known<br />
as the winner determination problem) are<br />
considered, as are questions of how to test<br />
alternative algorithms. The book discusses<br />
five important applications of CAs: spectrum<br />
auctions, airport takeoff and landing slots,<br />
procurement of freight transportation<br />
services, the London bus routes market,<br />
and industrial procurement. This unique<br />
collection makes recent work in CAs<br />
available to a broad audience of researchers<br />
and practitioners. The integration of work<br />
from the three disciplines underlying CAs,<br />
using a common language throughout,<br />
serves to advance the field in theory and<br />
practice.<br />
38<br />
November 2005|<strong>AgentLink</strong> News 19
AGENT EVENTS<br />
CALENDAR<br />
For further details visit:<br />
http://www.agentlink.org<br />
If you would like to announce an event<br />
email: publications@agentlink.org<br />
with the details<br />
2005<br />
ASW 2005 Nov 3-6 1st International Symposium on Agents and the Semantic Web Arlington, Virginia, USA<br />
ExaCt 2005 Nov 3-6 International Symposium on Explanation-aware Computing Washington D.C, USA<br />
MIPSA 2005 Nov 3-6 AAAI 2005 Fall Symposium on Mixed-Initiative Problem Solving Assistants Arlington, Virginia, USA<br />
ROLES 2005 Nov 3-6 2005 AAAI Fall Symposium: Roles, an interdisciplinary perspective Arlington, Virginia, USA<br />
ISWDS 2005 Nov 6-10 The 1st International Semantic Web Doctoral Symposium Galway, Ireland<br />
MICAI 2005 Nov 14-18 The 4th Mexican International Conference on Artificial Intelligence Monterrey, Mexico<br />
MA4CS 2005 Nov 14-18 Multi-Agents for modeling Complex Systems Paris, France<br />
WISE 2005 Nov 20-22 The 6th International Conference on Web Information Systems Engineering New York, USA<br />
TAAI 2005 Dec 2-3 10th Conference on Artificial Intelligence and Applications Kaohsiung, Taiwan<br />
MASTA 2005 Dec 5-8 The 3rd Workshop on Multi-Agent Systems: Theory and Applications Covilha, Portugal<br />
EUMAS 2005 Dec 7-8 The 3rd European Workshop on Multi-Agent Systems Brussels, Belgium<br />
CISSE 2005 Dec 10-20 International Joint Conferences on Computer, Information, and Systems Sciences, and Engineering Online Virtual Event<br />
SOAS 2005 Dec 11-13 The 2005 International Conference on Self-Organization and Adaptation of Multi-agent and Grid Systems Glasgow, Scotland<br />
CIS 2005 Dec 15-19 The International Conference on Computational Intelligence and Security Xi’an, China<br />
IASAA 2005 Dec 16-18 WSEAS Special Session: Intelligent Agents: Standards, Architectures and Applications Tenerife, Canary Islands<br />
IICAI 2005 Dec 20-22 The 2nd Indian International Conference on Artificial Intelligence (IICAI-05) Pune, India<br />
MAS 2005 Dec 20-22 IICIA-05 Special Session on Multi-agent Systems Pune, India<br />
ICDCIT 2005 Dec 22-24 2nd International Conference on Distributed Computing & Internet Technology (ICDCIT 2005) Bhubaneswar, India<br />
2006<br />
ADS 2006 Apr 2-6 Workshop on Agent-Directed Simulation Huntsville, USA<br />
MDAI 2006 Apr 3-5 Modeling Decisions for Artificial Intelligence Vienna, Austria<br />
ACE 2006 Apr 18-20 Agent Construction and Emotions:Modeling the Cognitive Antecedents and Consequences of Emotion Vienna, Austria<br />
EMCSR 2006 Apr 18-21 The biennial European Meeting on Cybernetics and Systems Research Vienna, Austria<br />
AAMAS 2006 May 8-12 The 5th International Joint Conference on Autonomous and Mutli-Agent Systems Hakodate, Japan<br />
FLAIRS 2006 May 13-11 The 19th International FLAIRS Conference FL, USA<br />
AGC 2006 May 16-19 International Workshop on Agent based Grid Computing Singapore, Singapore<br />
CSWWS 2005 June 6 Canadian Semantic Web Working Symposium Quebec, Canada<br />
AMT 2006 June 7-9 The 4th International Conference on Active Media Technology Brisbane, Australia<br />
ESWC 2006 June 11-14 3rd European Semantic Web Conference Budva, Montenegro<br />
ICAC 2006 June 12-16 The 3rd IEEE International Conference on Autonomic Computing Dublin, Ireland<br />
Coordination 2006 June 13-16 8th International Conference on Coordination Models and Languages Bologna, Italy<br />
DEON 2006 Jul 12-14 The 8th International Workshop on Deontic Logic in Computer Science Utrecht, The Netherlands<br />
AAAI 2006 Jul 16-20 21st National Conference on Artificial Intelligence (AAAI-06) Boston, MA, USA<br />
ISWC 2006 Dec 5-9 5th International Semantic Web Conference Athens, GA, USA<br />
November 2005|<strong>AgentLink</strong> News 19<br />
39
www.agentlink.org<br />
About <strong>AgentLink</strong>...<br />
<strong>AgentLink</strong> is the European Coordination<br />
Action for Agent-based Computing,<br />
a network of researchers and developers with a<br />
common interest in agent technology. It is funded<br />
by the European Commission, so membership is<br />
free, but you can only take advantage of <strong>AgentLink</strong><br />
activities if you are a member. If your organisation is<br />
engaged in agent-related activities, you should join.<br />
Contact the coordinator or administrator for details,<br />
or visit www.agentlink.org.<br />
The aim of <strong>AgentLink</strong> News is to provide an<br />
informal way of communicating both what is<br />
happening in <strong>AgentLink</strong> and the agent world generally.<br />
<strong>AgentLink</strong> News offers a range of articles including<br />
features, reports on conferences and workshops,<br />
informal description of research results and new<br />
software, book reviews, and website developments.<br />
ISSN 1465-3842<br />
AGENTLINK PUBLICATIONS AND<br />
WEB COORDINATOR<br />
Serena Raffin<br />
School of Electronics and Computer Science<br />
University of Southampton<br />
Southampton SO17 1BJ<br />
United Kingdom<br />
[e] web@agentlink.org<br />
AGENTLINK ADMINISTRATORS<br />
Adele Maggs<br />
Department of Computer Science<br />
University of Liverpool<br />
Liverpool L69 3BX<br />
United Kingdom<br />
[e] admin@agentlink.org<br />
Rebecca Earl<br />
School of Electronics and Computer Science<br />
University of Southampton<br />
Southampton SO17 1BJ<br />
United Kingdom<br />
[e] publications@agentlink.org<br />
AGENTLINK EVENT COORDINATOR<br />
Catherine Atherton<br />
Department of Computer Science<br />
University of Liverpool<br />
Liverpool L69 3BX<br />
United Kingdom<br />
[e] events@agentlink.org<br />
AGENTLINK TECHNOLOGY<br />
SUPPORT COORDINATOR<br />
Steve Munroe<br />
School of Electronics and Computer Science<br />
University of Southampton<br />
Southampton SO17 1BJ<br />
United Kingdom<br />
[e] web@agentlink.org<br />
AGENTLINK COORDINATORS<br />
Peter McBurney<br />
Department of Computer Science<br />
University of Liverpool<br />
Liverpool L69 3BX<br />
United Kingdom<br />
[e] Peter.McBurney@agentlink.org<br />
Terry Payne<br />
School of Electronics and Computer Science<br />
University of Southampton<br />
Highfield<br />
Southampton SO17 1BJ<br />
United Kingdom<br />
[e] Terry.Payne@agentlink.org<br />
AGENTLINK EXECUTIVE COORDINATORS<br />
Michael Luck, University of Southampton, UK<br />
Michael Wooldridge, University of Liverpool, UK<br />
INDUSTRY ACTION<br />
(Workpackage 1) COORDINATOR<br />
Michal Pechoucek<br />
Czech Technical University<br />
Czech Republic<br />
[e] Michal.Pechoucek@agentlink.org<br />
STANDARDISATION ACTIVITY<br />
(Workpackage 2) COORDINATOR<br />
Monique Calisti<br />
Whitestein Technologies AG<br />
Switzerland<br />
[e] Monique.Calisti@agentlink.org<br />
RESEARCH ACTION<br />
(Workpackage 3) COORDINATOR<br />
Steven Willmott<br />
LSI,Universitat Politècnica de Catalunya<br />
Spain<br />
[e] Steven.Willmott@agentlink.org<br />
STUDENT INTEGRATION PROGRAMME<br />
(Workpackage 4) COORDINATOR<br />
Wiebe van der Hoek<br />
University of Liverpool<br />
United Kingdom<br />
[e] Wiebe.van.der.Hoek@agentlink.org<br />
TECHNICAL FORUM<br />
(Workpackage 5) COORDINATOR<br />
Andrea Omicini<br />
DEIS, Università di Bologna<br />
Italy<br />
[e] Andrea.Omicini@agentlink.org<br />
TECHNOLOGICAL ROADMAP<br />
(Workpackage 6) COORDINATOR<br />
Michael Luck<br />
University of Southampton<br />
United Kingdom<br />
[e] Michael.Luck@agentlink.org<br />
ADDITIONAL COMMITTEE MEMBERS<br />
Onn Shehory<br />
IBM Research<br />
Israel<br />
[e] Onn.Shehory@agentlink.org<br />
Simon Thompson<br />
BT Exact<br />
United Kingdom<br />
[e] Simon.Thompson@agentlink.org<br />
Jörg Müller<br />
Siemens AG<br />
Munich<br />
[e] Joerg.Mueller@agentlink.org