A declarative approach to software agents and its application to interactive hyperdocuments.

We propose to design a declarative model that is powerful enough to descrihe autonomous software agents, as well as interactions between such agents. As a first approximation, one may think of an agent as a theory in some logic, representing its knowledge, and a goal (which may be complex and include constraints). When interacting with another agent, a dialogue is started which can be regarded as a search process for a feasible compromise where hoth agents are able to prove their goal, suject to certain updates of their theories (these updates need not he the same for all participants).