Abstract
Software agents have become a useful paradigm for developing intelligent, distributed software architectures in which proactive software entities collaborate with each other to achieve goals. Given the sophistication and variety of agent functionalities and Multi-Agent System implementations, agent-oriented software engineering (AOSE) tools are needed to aid the user in designing, implementing, debugging and maintaining these systems of agents. Software comprehension, which is essential for maintaining software systems, has been a manual process, involving the interpretation of agent behavior in the implemented system. This paper presents the Designer's Agent Creation and Analysis Toolkit (DACAT) for the architectural design phase of the software engineering lifecycle, and describes the Tracer tool to automate interpretations and explanations of the implementation, thus helping users comprehend implemented agent behaviors. By explicitly modeling the user's comprehension of the implemented system as background knowledge for Tracer, feedback can be provided as to whether the user's comprehension accurately represents the implementation's behavior. The contributions of this research are two AOSE tools – DACAT offers design tools that provide visualization and automated evaluation of the agent architecture, and the Tracer tool offers automation of the manual tasks, such as verifying that agents are behaving as expected, identifying unexpected behavior, and generating explanations.