Evaluating the impact of aspects on inconsistency detection effort: A controlled experiment

Abstract

Design models represent modular realizations of stakeholders' concerns and communicate the design decisions to be implemented by developers. Unfortunately, they often suffer from inconsistency problems. Aspect-oriented modeling (AOM) aims at promoting better modularity. However, there is no empirical knowledge about its impact on the inconsistency detection effort. To address this gap, this work investigates the effects of AOM on: (1) the developers' effort to detect inconsistencies; (2) the inconsistency detection rate; and (3) the interpretation of design models in the presence of inconsistencies. A controlled experiment was conducted with 26 subjects and involved the analysis of 520 models. The results, supported by statistical tests, show that the effort of detecting inconsistencies is 20 percent lower in AO models than in their OO counterparts. On the other hand, the inconsistency detection rate and the number of misinterpretations are 43 and 37 percent higher in AO models than in OO models, respectively. © 2012 Springer-Verlag.