Object-Oriented Modeling with UML: A Study of Developers’ Perceptions

Author(s): Ritu Agarwal and Atish P. Sinha
Venue: , Communications of the ACM, Volume 46 , Issue 9, Why CS students need math, SPECIAL ISSUE: Virtual extension
Date: 2003


This paper shows the results of a study that sought to assess the general usability of
UML diagrams. A class of undergrad and grad students developed software over the duration
of a course on Object-Oriented Analysis and Design using UML diagrams. They worked in
teams of 4 or 5 on various types of software projects. Throughout the course, they
used case, class, state, and interaction UML diagrams in the development of their programs.
They used a tool called Rational Rose to help them draw the various UML diagrams.

Too acquire the actual results from the study, two surveys were performed. The first was
given at the beginning of the course and was used to gather background information about
each student with regards to their experience and knowledge level about various areas
of software engineering. The second survey was given at the end of the class and was
designed to assess the usability of the various UML diagrams. They answered a number
of questions about the various UML diagrams where they said whether they agreed with
various statements. Answers were given on a scale of 1 to 7 -- highly disagree to highly agree.
The answers were that used to calculate the results using various fancy measurement
scales and techniques.

They found that the students had positive perceptions of the usability of UML diagrams in
general but that they didn't really rate any particular diagram type as highly usable.
Out of the 4 types of UML diagrams used, case and state diagrams were found to be easier
to use than class and interaction diagrams. Interestingly, they that that prior experience in
conceptual data modeling or system development had no significant effect on the usability
of the diagrams. However, those with prior experience in OO analysis and design techniques
found the class and interaction diagrams to be easier to use than those without.

Surprisingly, they found that prior experience in process-oriented analysis
and design techniques had a positive effect on the usability of the UML diagrams which
tends to contradict previous studies. Those doing this study surmise that it may be due
to the use of use case diagrams in this study.

All in all, UML diagrams were found to be fairly usable be the students. All 4 diagram
types were found to have a mean above 4 out of 7 on the usability scale (which would
have meant neutral instead of positive or negative). However, no diagram type achieved
much higher than a mean of 5, so the UML diagrams were certainly useful but not
overwhelmingly so.