Capturing Agent Autonomy in Roles and XML

A key question in the field of agent-oriented software engineering is how the kind and extent of autonomy owned by computational agents can be appropriately captured. As long as this question is not answered convincingly, it is very unlikely that agent-oriented software (having "autonomy" as a real property rather than just a catchy label) gets broadly accepted in industry and commerce. In particular, in order to be of practical value an answer to this question has to come in form of concrete techniques which enable developers of agent-oriented software to precisely capture the scope of behavioral freedom and self-control they want to concede to a computational agent. This paper describes two such techniques. First, a formal schema called RNS for specifying the boundaries of autonomous agent behavior. This schema is conceptually grounded in sociological role theory, and employs the concepts of role, norm and sanction to capture agent autonomy. What makes RNS particularly valuable and distinct from related autonomy specification approaches is, among other things, its strong expressiveness and high precision. Second, a software tool called XRNS which enables developers to easily generate RNS-based autonomy specifications in XML format. Encoded in XML, these specifications are easily accessible to all stakeholders in an agent-oriented software under development, and can be even processed directly by XML enabled computational agents.