Metric-Based Evaluation of Multiagent Systems Models

The use of Multiagent Systems (MAS) has been increasing over the years due to their capacity of dealing with problems in a variety of domains. The modeling of such systems is not trivial: besides the knowledge and skills on agent-oriented software engineering and a basic understanding of the target domain to be modeled, it demands familiarity with agent-oriented modeling methodologies. This is not always the case, though, especially for newcomers in the field. This work proposes a set of guidelines in the form of a questionnaire for the evaluation of MAS models, aiming at supporting the verification of their quality. The questionnaire is built upon the results of a systematic mapping conducted to identify how MAS models have been evaluated and what metrics have been used. The proposed guidelines were evaluated through (i) a peer review by experts and (ii) its actual application by graduate students applying modeling methodologies in the context of Guardian Angel (GA), a patient-centered health system that automatically supports patients suffering from chronic diseases. Participants provided an overall positive feedback and proposed some improvements on the questionnaire, most of which were promptly incorporated.