Perspectives in Knowledge Representation Luigia Carlucci Aiello, Daniele Nardi Dipartimento di Informatica e Sistemistica Universit` a di Roma “La Sapienza” Via Buonarroti 12, I-00185 Roma, Italia Abstract In this paper we discuss recent developments of the research on Knowledge Represen- tation, focusing on hybrid formalisms, nonmonotonic reasoning and formalisms for rea- soning about knowledge and reasoning in a multi-agent scenario. 1 Introduction We believe that, in the future of Artificial Intelligence, an important role will be played by the research in Knowledge Representation and Knowledge Representation formalisms. It has been largely so in the past, even though sometimes theory and application looked disconnected from one another, and the impact of the research was difficult to appreciate in terms of new market products. In fact, the construction of intelligent systems involves two main ingredients: for- malisms that capture the relevant features of common sense reasoning; methodologies for building knowledge-based systems. In the present discussion we focus on the former aspect, warning the reader that the latter is equally important. In order for a system to behave intelligently, it must embody various kinds of knowl- edge. In particular, we expect an intelligent system to incorporate a model of the world, of self, and of the user. Roughly speaking, this means the system should know about the above entities, and be capable of reasoning about them. This is provided by the repre- sentation, which not only reifies these concepts in terms of computer data structures, but provides algorithms which effectively implement the reasoning processes going on within such models. An intelligent system should therefore keep an explicit representation of the world, and reason about it and its changes; it should be introspective, namely able to access its own representation and reason about it; and finally, the system should have a representation of the knowledge possessed by other agents in the world, such as for example the user, and be able to understand their behaviour. Research in Knowledge Representation aims at providing formalisms that allow one to represent effectively all these different kinds of knowledge, and at constructing systems 1