A Report on JELIA '96

Appeared in Volume 10/1, February 1997

Chandrabose Aravindan

Évora - a small, historic town in Portugal about 120km east of the capital Lisboa - hosted the fifth European Workshop on Logics in Artificial Intelligence, popularly known as JELIA. JELIA is a French acronym that stands for Journées Européennes sur la Logique en Intelligence Artificielle). The workshop was held at the main campus of Universidade de Évora, from September 30th, 1996 to October 3rd, 1996.

This series of workshops on Logics in AI began in Roscoff, France in 1988 to meet the need for an European forum for the discussion of emerging work in the field of logics in Artificial Intelligence. The workshop brings together AI researchers from various areas, such as automated reasoning, modal logics, default logics, temporal logics, spatial logics, logic programming, and non-monotonic reasoning.

This fifth workshop had 57 submissions, of which the programme committee selected 25 for presentation and publication (Springer-Verlag LNAI 1126). Among these 25, 6 papers qualified for the special issue of the Journal of Automated Reasoning due out in 1997. About 40 researchers went to Évora to attend the workshop.

After opening remarks by the workshop chair José Alferes and programme chairs Ewa Orlowska and Luís Pereira, Ulrich Furbach opened the first session on automated reasoning with a talk on Hyper Tableaux (joint work with Peter Baumgartner and Ilkka Niemelä). He described a tableaux approach for theorem proving that incorporates many useful ideas from the resolution field. Hans de Nivelle came next to discuss an optimized unification algorithm suitable for resolution based theorem provers. The work by Christophe Bourley, Gilles Défourneaux, and Nicolas Peltier on an analogy based resolution framework was presented next. Finally, Anatoli Degtyarev talked about rigid E-unification (joint work with Andrei Voronkov) and defined a proof procedure for first order logic with equality.

The session on Modal Logic, consisted of two talks on tableaux based proof procedures for modal logic. First to be discussed was the work by Alberto Artosi, Paola Benassi, Guido Governatori, and Antonio Rotolo, that combined tableau-like calculus with analytic cut and label formalism to obtain a proof procedure for normal modal logics. Later, the work of Francesco Donini, Fabio Massacci, Daniele Nardi, and Riccardo Rosati showed that it is possible to have semantic tableaux calculi for wide range of non-monotonic modal logics.

The next session hosted two works that outlined potential applications of modal logic. Peter Fröhlich, Wolfgang Nejdl, and Michael Schroeder presented a modal logic based meta language to express diagnostic process and design strategies. Filipe Santos and José Carmo used modal logic for the specification and analysis of organizations.

What kind of negation do we need for effective knowledge representation and reasoning? Classical? Default? Strong? Explicit? or no negation at all!? Luís Pereira talked about this issue (joint work with José Alferes and Teodor Przymusinski) and convinced the audience that at least two kinds of negations are necessary. Later, Joeri Engelfriet insisted that only persistence can make non-monotonicity monotonous! He characterised the sentences that can be added to premises without destroying the consequences. The final talk of the session was given by Konstantinos Georgatos, who spoke about characterizing rational inference relations by orders among the formulas.

It is unfortunate that two authors, Artur Mikitiuk and Viorica Ciorba, could not attend the workshop to give their talks. Consequently, the session on default logic was shorter. Joeri Engelfriet's talk was about extending Reiter's default logic by allowing infinite sets of justifications to capture beliefs of reasoning agents (joint work with Wiktor Marek, Jan Treur, and Miroslaw Truszczynski). Later, Mary-Anne Williams talked about strategies for computing extensions of constraint default theories.

The session on logic programming started with my talk on an abductive framework for capturing the semantics of negation in disjunctive logic programming. It was followed by Jürgen Dix who presented (joint work with Stefan Brass) two important characterizations of D-WFS, a disjunctive well-founded semantics. Finally the work of Vasco Pedro and Luís Monteiro on verifying the partial correctness of logic programs with modules was presented.

The session on temporal and spatial logics opened with work by Manuel Enciso, Inma de Guzman, and Carlos Rossi on an automated theorem prover called TAS-FNext for temporal logic. It was followed by a talk on temporal logic with 'next' and 'unless' by Regimantas Pliuskevicius. The final talk of the session was given by Brandon Bennett on constructing theory of spatial regions.

Perhaps the most interesting part of the workshop were the invited talks, which aroused a lot of stimulating discussion. There were three invited talks given by well-known researchers in the field of logic and AI: Michael Genesereth, Gert Smolka, and Robert Kowalski.

Michael Genesereth originally planned to talk about MaCarthy's vision on AI from today's perspective, and the proceedings carries the abstract of this proposed talk. But he changed his mind after going through the proceedings, and discussed his current project on virtual information system. His talk on 'InfoMaster' included a lively on-line demonstration over the Internet. See the Web page at http://infomaster.stanford.edu/ for more information.

The second invited talk by Gert Smolka was about a general purpose concurrent constraint programming language called Oz. Oz is very well suited for developing multi-agent, concurrent, or constraint based systems. Smolka provided insights into the concepts and design of the language, and argued that the lack of a declarative semantics is not a serious problem. More information on Oz can be obtained from http://www.ps.uni-sb.de/oz/.

Robert Kowalski gave an interesting talk about programming with logic and integrity constraints. He discussed a proof procedure for such a system, and its applications. Many papers related to this talk are available from http://laotzu.doc.ic.ac.uk/UserPages/staff/rak/rak.html

The last session of the workshop was devoted to works on belief revision and paraconsistency. Paola Forcheri, Paolo Gentilini, and Maria Molfino introduced an informational logic based on logical entropy and discussed its applications in automated theorem proving. Michael Kaminski talked about semantical definition of extensions for open default theories and its correspondence with the domain closure assumption (joint work with Johann Makowsky and Michael Tiomkin). Later, Cees Witteveen presented (joint work with Wiebe van der Hoek) an approach for revision and updating based on a back-up semantics. The final talk of the session and the workshop was given by Philippe Besnard, who presented his joint work with Torsten Schaub on signed system for paraconsistent reasoning.

Overall, it was a well-organized and successful workshop. Two minus points were the lack of a good Internet connection and some noise problems.

During the course of the workshop, it was decided that the next JELIA will be held in Germany, with Jürgen Dix as the workshop chair and Ulrich Furbach and Luis Farinas del Cerro as program committee chairs. I wish them all the best for a successful workshop in 1998.

Chandrabose Aravindan
Universität Koblenz-Landau
Fachbereich Informatik
Rheinau 1, D-56075 Koblenz, Germany
Email: arvind@informatik.uni-koblenz.de