30 AICOM Vol.1 No.4 December 1988 CONFERENCE REPORTS 2nd Int. Conf. on Theoretical Aspects of Reasoning about Knowledge Monterey, California, 7-9 march 1988 Fabrizio Sebastiani Istituto di Elaborazione dell'InJormazione - CNR Via S. Maria, 46 - 56126 Pisa (Italy) First of all, let us do away with a potential misunder- standing: the topic of "Theoretical Aspects of Rea- soning about Knowledge" was not knowledge representation. Or rather, this was not the only topic. The word "knowledge" in the title is used in the linguistic sense of the term, as opposed to the metalinguistic reading which is implicit in the notion of KR. In fact, if we adhere to the by now widely accepted definition of KR proposed by Brian Smith, the concept of KR presumes an implicit attribution of knowledge to the system that exhibits this hypotheti- cal property by an observer external to the system itself; if we want to maintain the parallel with KR, the topic of the Monterey conference was rather the representation of the very concept of knowledge, its problems and its applications. We might say, with some over-simplification, that there was no talk of "Clyde is an elephant", but rather of "Mary knows that Clyde is an elephant"; i.e. topics were discussed pertaining to realities that may be modelled by means of ("epistemic") logics endowed with operators of type KNOWS. T ARK is a biennial congress, sponsored by NSF and AAAI and organized by IBM San Jose; together with its sister conference "Theoretical Aspects of Reason- ing about Actions and Plans" TARK constitutes by now a fundamental point of reference for scholars engaged in the recently emerging "theory of intelli- gent autonomous agents" and, in general, in non- standard logics for AI. TARK is not, however, a specifically AI conference, but an interdisciplinary gathering of scholars that are brought together by some kind of interest in the very notion of knowledge (to understand the importance of this notion to AI proper one need only recall that at recent outstanding AI conferences (such as AAAI'84, nCA1'85, nCA1'87) the best paper prizes have been awarded to articles related to it). Confirming a trend that is now being consolidated within theoretical AI, the organizers of TARK'88 have successfully strived to create a small but highly qualified workshop, characterized by a limited number of both participants (about 80) and technical papers (22 accepted out of 108 submitted). The lack of parallel sessions has fostered lively exchanges of ideas among scholars from different disciplines: phi- losophy of language (its representatives included Kamp, Lenzen, Thomason), logic (Barwise), knowledge representation (Konolige, Israel, Lifschitz, McCarthy), game-theory (Bicchieri, Werlang), analysis of distributed systems (Fagin, Lamport, Vardi) and cryptography (Tompa, Tuttle). The debate that followed stressed once again the fact that AI can no longer avoid looking, both for inspiration and endorsement, at a diversity of disciplines, among which the philosophy of language emerges as that with the highest potential for cross-fertilization with artificial intelligence as a whole. The majority of the authors of accepted papers were from the US and Israel, with the European share being limited to Peter Giirdenfors and Eric Werner. Giirdenfors' paper ("Revisions of knowledge systems using epistemic entrenchment") was one of the richest with implications for AI as a whole and, in particular, for the now increasingly popular field of belief revision. This work (co-authored by David Makinson and part of a broader-scope research on the logics of theory change carried out in collaboration with Carlos Alchourm) attacks the problem of knowledge base updating, presenting a formal account of the opera- tions involved (expansions, contractions, revisions); this account is not only formal, but also constructive, in the sense that each such operation is univocally determined by an ordering of importance ("epistemic entrenchment") which is defined on knowledge base formulae. If, on one side, the notion of EE allows an impeccable formalization of updating operations, its philosophical grounding (together with the ghosts of necessity and analyticity which it evokes) is likely to be the object of hot debate.