12.07.2015 Views

View - ResearchGate

View - ResearchGate

View - ResearchGate

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Developing Agents Who Can Relate to Us 41Although such agents do not have access to the basic decision makingprocesses they do have access to and can report on their linguistic selfmodelwhich is a model of their decision making (which is, at least,fairly good). Thus, they do have a reportable language of thought, butone which is only a good approximation to the underlying basic decisionmaking process.The model allows social and self reflective thinking, limited only bycomputational resources and ingenuity - there is no problem with unlimitedregression, since introspection is done not directly but via a modelof one’s own thought processes.5. Towards Implementing Self-Constructing AgentsThe above model gives enough information to start to work towards an implementation.Some of the basic requirements for such an implementation arethus:1 A suitable social environment (including humans)2 Sufficiently rich communicative ability - i.e. a communicative languagethat allows the fine-grained modelling of others’ internal states leadingto action in that language3 General anticipatory modelling capability4 An ability to distinguish the experience of different types, including theobservation of the actions of particular others; ones own actions; andother sensations5 An ability to recognise other agents as distinguishable individuals6 Need to predict other’s decisions7 Need to predict one’s own decisions8 Ability to reuse model structures learnt for one purpose for anotherSome of these are requirements upon the internal architecture of an agent,and some upon the society it develops in. I will briefly outline a possibility foreach. The agent will need to develop two sets of models.1 A set of models that anticipate the results of action, including communicativeactions (this roughly corresponds to a model of the world includingother agents). Each model would be composed of several parts: - acondition for the action - the nature of the action - the anticipated effectof the action - (possibly) its past endorsements as to its past reliability

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!