Interaction-Oriented Programming: Concepts, Theories, and Results on Commitment Protocols

Abstract

Unlike traditional information systems, modern systems are _open_, consisting of autonomous, heterogeneous parties interacting dynamically. Yet prevalent software techniques make few accommodations for this fundamental change. Multiagent systems are conceptualized for open environments. They give prominence to flexible reasoning and arms-length interactions captured via communications. On the backdrop of multiagent systems, Interaction-Oriented Programming is the idea of programming with interactions as first-class entities instead of, e.g., objects. Protocols are to interactions as classes are to objects: because of their key nature, protocols have obtained a lot of research attention. Modeling protocols suitably for open environments meant modeling their content, not just the surface communications. In a number of important cases, such as business processes and organizations, the content is best understood using the notion of commitments of an agent to another agent in an appropriate context. Our theory of protocols supports flexible enactment of protocols, a treatment of refinement and composition of protocols, and their relationship with organizations and contracts, thus reducing the gap between agents and conventional computer science. This talk will review the key concepts, theories, and results on commitment protocols, and some important challenges that remain..