o COMMUNICATING
KNOWLEDGE SYSTEMS:
Part I-Big Talk among Small
Systems
CRISTINA SERNADAS
Departmento de Matematica, Instituto Superior
Iecnico, Universidade Iecnica de Lisboa, 1096 Lisboa
Codex, Portugal
HELDER COELHO
Centro de Informatica, LNEC, 1799 Lisboa Codex,
Portugal
GRAl;A GASPAR
DICC, Faculdade de Ciencias de Lisboa, 1300 Lisboa,
Portugal
A new model is imroduced for describing societies of communicating knowledge systems from
an artificial intelligence perspective. The society is defined as a Time-evolving concept con-
taining both static and dynamic knowledge. The static knowledge includes The components of
The society, namely The actor Types, Thecommunication issue Types, and The behavior pattern
Types. The dynamic knowledge specifies how The society can evolve. Each component is a
metatheory described using a specific representation system. The metatheories representing
The actor Types include self-knowledge, knowledge about other actors, and knowledge about
The real world. The communication issues always reflect The knowledge of Thesending actor
and produce some effect on The receiving actor.for instance. enrichments of its knowledge or
conflicts with' iTS own knowledge. Conflias are of two forms: routine conflicts TO be solved by
mediating actors wiTh no inre/ligenr knowledge. or conflicts that require intelligent knowledge
and Thatmust be solved by special aaors called managers. Behavior pattern rules are Theo-
ries that express cause-effect rules regarding The society behavior as a whole. Technically,
modifications of The society are defined as parameterized Theories whose argument specifies
The requirements To be satisfied before a change of the situation and whose body introduces
The modifications TO be performed.
INTRODUCTION
The problem of communication between systems has been addressed in sev-
eral areas of computing sciences, namely data communication, programming
languages, distributed operating systems and services, distributed databases, and
artificial intelligence.
The main issue in the programming language area has been the identification
Editor's Note: Becauseof the length of this paper. it has been divided into two parts. Part I comprises the
first four sections, Part n the last two sections plus appendices. Part n will appear in volume I, issue 4.
Applied Anificiallntelligence: 1:233-260, 1987
Copyright © 1987 by Hemisphere Publishing Corporation 233