27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

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.

TABLE I.<br />

COMPARISON OF THE ACCEPTABILITY OF OVERHEAD TIME<br />

BETWEEN THE TWO GROUPS<br />

Question Interactive Background P-Value<br />

Please rate the<br />

acceptability of the time<br />

spent on recording links<br />

Please rate the<br />

acceptability of time<br />

spent on applying rules<br />

TABLE II.<br />

Ave: 2.4<br />

± 0.9<br />

Ave: 3.0<br />

± 1.3<br />

Ave: 2.2<br />

± 0.8<br />

N/A<br />

Note: Scale of 1 to 5, where 1= “Highly Acceptable” and<br />

5=”Unacceptable”<br />

0.341<br />

N/A<br />

LEVEL OF DISTRACTION IN LINK CAPTURE WITH ACTS<br />

Question Interactive Background P-Value<br />

Did you find the turning on<br />

and off of the recording<br />

button distracting?<br />

Did you find the rules to be<br />

distracting?<br />

Ave: 2.8<br />

± 1.2<br />

Ave: 2.2<br />

± 1.0<br />

Ave: 3.9<br />

± 1.1<br />

N/A<br />

0.027<br />

N/A<br />

Note: Scale of 1 to 5, where 1= “Very Distracting” and 5=”Not at all”<br />

I. EVALUATION<br />

To assess the extent to which ACTS addre sses challenges<br />

from the three perspectives, we ran case studies in vari ous<br />

contexts: lab settings, a proprietary project, and an open source<br />

project. We focused on the following questions:<br />

Q1. Is the cost associated with capturing a traceability<br />

link acceptable to users? (Economic and social)<br />

Q2. Are architecture-centric traceability links usable?<br />

(Technical and social)<br />

Q3. Would users use the tool in future projects?<br />

(Social, economic and technical)<br />

A. Case studies in laboratory setting<br />

Case studies in a laboratory setting allow us to perform a<br />

screen capture of the user actions with the tool for later<br />

analysis. W e solicited feedb ack from 18 participants, 15 of<br />

whom are PhD stude nts in either Co mputer Science or<br />

Informatics at the University of California, Irvine. In addition,<br />

seven of these participants ha ve held indus try positions ( e.g.,<br />

programmer, architect, manager, and intern), allowing them to<br />

assess the feasibility of the tool based on their roles in industry.<br />

The participants were asked to use the AC TS tool in the<br />

following task: capture trac eability links while editing a<br />

structural design in A rchStudio and while viewing or editing<br />

documentation files. Half of the user s were asked to<br />

interactively apply rules and half applied the same rules in the<br />

background. All users were also asked to retrieve the captured<br />

links. Prior to the task, users were given a brief tutorial on how<br />

to use the tool (about 5 m inutes). After t he task, users were<br />

then asked to provide feedback via a questionnaire and semistructured<br />

interview. Based on the screen capture, we also<br />

measured the time for capt uring and retrieving traceabili ty<br />

links.<br />

Q1: Both groups concur that time spent is a cceptable (see<br />

Table 1). An independent one-tailed t-test shows th at there is<br />

no significant difference bet ween the ratings of both groups,<br />

suggesting that both groups view the overhead as e qually<br />

acceptable. Users commented that the tool “make[s] the lin king<br />

job easier” and “[t]he tool saved me lots of time. Thanks!”<br />

Using the recorded screen ca pture, we also measured the<br />

time it takes for users to c apture traceability links and to<br />

retrieve the captured links. The box plot in Figure 2 show s the<br />

distribution of overall tim es between users who applied the<br />

rules interactively and users who applied the rules in the<br />

background. The box plot shows that most of the users are<br />

more efficient when applying rules in the background. A onetailed<br />

independent t-test co mparison between the two groups<br />

yields a p- value of 0.005, strongly suggesting that appl ying<br />

rules in the background is m ore efficient than interactive<br />

application.<br />

Meanwhile, the participants in the interacti ve rules group<br />

felt that the manual application of rules was less ac ceptable<br />

(average rating of 3 in Table I). These users felt that turning the<br />

record button on and off was “somewhat distracting” while<br />

users who applied the rules au tomatically in the background<br />

found it less distracting (see Table II). An independent onetailed<br />

t-test between the two groups on this question shows p =<br />

0.027, indicating there is a significant difference between the<br />

views of the groups. A user from the interactive rules gr oup<br />

described turning the recording on and off to be “tedious.”<br />

Q2: Linking artifacts to the architecture (or design) is a<br />

useful feature to some of our pa rticipants. Five participants<br />

(three with industry expe rience) identified this feature as the<br />

feature they liked about the t ool. One of these participants, an<br />

architect/programmer, reasoned that “ [i]n practice [it is]<br />

necessary to link architectu re to doc uments (requirements<br />

specs). [T]his makes the documents more useful.”<br />

Q3: Over hal f of the particip ants indicated interest in<br />

continued usage of the tool for their own software development<br />

projects. Nine participants answered “yes”, and three<br />

additional participants answered “yes” with conditions, such as<br />

if the user is a system architect using ArchStudio or if the user<br />

needed to link the documentation with the design. Three<br />

additional participants answered “maybe”.<br />

Most of the participants l iked the abilit y to link design<br />

elements to documentation, the ability to link to specific points<br />

within the documentation, and the ease of navigating to the<br />

documentation from design. One participant liked the ability to<br />

view the artifacts in their native tool: “Simplifies the process,<br />

everything is integrated into one workspace.”<br />

Figure 2: Time to capture and retrieve traceability links<br />

415

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

Saved successfully!

Ooh no, something went wrong!