Towards a Set of Measures for Evaluating Software Agent Autonomy

Abstract

Agent-oriented software is an established research field. For this reason, it is important to develop comprehensive measures of excellence to evaluate this software. No set of measures defining the overall quality of an agent has been developed to date. Some attempts at evaluating agent quality have addressed certain agent features, like the development process. We believe that agent quality can be determined as a function of well-defined characteristics. Evaluated using appropriate measures, these characteristics will assure an agent's reliability and correct functionality. This paper deals with an important agent feature, namely, autonomy. Autonomy is considered to be the agent's ability to operate independently, without the need for human guidance or the intervention of external elements. The article proposes a set of measures used to evaluate the autonomy of a agent and presents a case study analyzing the behavior of these measures.