An empirical study on using stereotypes to improve understanding of UML models

Author(s): L. Kuzniarz, M. Staron, and C. Wohlin
Venue: 12th IEEE International Workshop on Program Comprehension
Date: 24-26 June 2004


Page(s):14 - 23

This paper was a study to determine the effects of UML stereotypes on program comprehension.
Graphical icons were used to represent stereotypes. For instance, if a class was a ``sender'' (that
is it sent signals but did not receive them) then an icon of a radio tower was placed next
to its class diagram. They used a class of 44 software engineering students for their experiment.
They gave them a 45 minute lecture on the idea of UML stereotypes prior to the experiment.

The experiment itself was done in two phases. In each phase, the students were presented with
a UML diagram and a questionnaire asking questions about the diagram in order to measure their
comprehension of the diagram. In one phase, the UML diagram used stereotypes. In the other, it
did not. Some students got the stereotyped version first. Others got the non-stereotyped version
first. The students were also rated on how quickly they filled out the questionnaires. After
the experiment, they were asked to fill out a questionnaire on their background and prior
knowledge of UML, UML stereotypes, and the domain area of the experiment (telecommunications).

39 of the 44 students' questionnaires were used for the results. The others were disqualified
for various reasons such as not turning in one of the questionnaires. The results of nearly
all the students were in favor of stereotypes. There was a mean increase of correct responses
to the questions of 52% when stereotypes were used. The mean time spent on the questionnaire was
25% lower when stereotypes were used. And the ratio of correct answers to time spent increased
by 52% with stereotypes.