Social Conceptions of Knowledge and Action: DAI Foundations and Open Systems Semantics

This article discusses foundations for Distributed Arti cial Intelligence (DAI), with a particular critical analysis of Hewitt's Open Information Systems Semantics (OISS). The article sets out to do ve things: It presents a brief overview of current DAI research including motivations and concepts, and discuss some of the basic problems in DAI. It introduces several principles that underly a fundamentally multi-agent (i.e., social) conception of action and knowledge for DAI research. These principles are introduced to provide de nitions, to delimit the discussion of OISS and as background against which to assess its contributions. It analyzes the main points of OISS in relation to these principles. It shows how attention to these principles can strengthen OISS approach to foundations for DAI. It traces some of the implications of this synthesis for theorizing and system-building in AI. The OISS approach productively challenges some conceptions of knowledge, reasoning, and action in classical AI research. However, it sometimes ignores the sophistication and richness of contemporary DAI research. Several of the key concepts of OISS are not clearly enough de ned or operationalized, and the article points out several ways to strengthen the OISS approach.