When Can Knowledge-based Systems Be Called Agents?

Distributed Artiicial Intelligence (DAI) has become an established research domain in recent years. As opposed to classical AI approaches, where the metaphor of intelligence is based on an individual human behaviour and the emphasis is put on knowledge representation and inference methods, the metaphor used in DAI is based on social behaviour, and its emphasis is put on actions and interactions. An "intelligent" social behaviour can arise either in the case of coarse-grained "intelligent" members of the society (called cognitive agents) or in the case of ne-grained "unintelligent" members of the society (called reactive agents). Interests in this eld may be divided into two major areas: Distributed Problem Solving (DPS) and Multi-Agent Systems (MAS). One can consider the DPS approach as one phase of the MAS approach. In either of these areas, the entities are called agents, and not knowledge-based systems as in classical AI approaches. The goal of this paper is to put together some ideas and deenitions of the MAS eld, and to give some answers to the question "When can knowledge-based systems be called agents?". To illustrate, we will use as an example a model of a cognitive agent being currently developed at the LIFIA which deals with connicting data.