Next: Results Up: Description Previous: Description

Objective: a Common Formal Model of ``Agent''

(This section is detailed in appendix , and summarized in , .)

The goal of ModelAge is to establish a common formal model of the concept of agent, usable across a wide number of fields of Information Technology, where its importance is now appearing:

  1. Requirements Engineering: identifying agents and their goals is a main step of RE.
  2. Software Design: setting up a society of agents is a crucial point in modern design; in its simplest form, it means designing concurrent programs and their communication protocols.
  3. Concurrency: formal model of concurrent programs include the notions of state (unstructured), communication, action, behaviour. Its enrichment with commitments, goals would ease the understanding of these concurrent programs.
  4. Data Bases: federative data bases group agents, each having its knowledge (an incomplete and sometimes incorrect representation of the external world), access protocol, and goals. The dynamic, or behavioural, aspect becomes thus essential.
  5. Artificial Intelligence: knowledge-based systems are often designed as a set of cooperating agents (experts); each of them has particular knowledge, communication protocols, and goals of cooperation with some other agents.
  6. Organization Theory: agents may represent individuals, organizations, or computer systems having a function (= a goal), a mental state, and fixed means of action and communication (protocols).

In each of these fields, the key facets of an agent appear to be:

  1. goal: each agent is trying to achieve through cooperation a precise objective under given constraints.
  2. state: agents have an internal, imperfect representation of the world (including the state of agents), on which their decisions are based.
  3. behaviour: agents act, communicate and perceive, showing thus an external behaviour that obeys the constraints given.
  4. heterogeneity: agents may be software, hardware systems or human organisations.
The collaborations already undertaken between members of ModelAge make us believe that a common model may be found. To ensure precision, the model will formal, i.e. mathematically described. In any case, the cross-fertilisation provided by this research will be very valuable, as shown in the past by members of ModelAge e.g. by the application of deontic logics, originally designed to formalise human organisations, to software specification; the introduction of the notion of behaviour and goals into data base theory; the use of belief logics (also originally modelling humans) in concurrent systems; the introduction of behaviour in software specification. The many formalisms used in each field could then be grounded (given a formal semantic definition) in this common semantic framework, allowing an integrated multi-formalism approach. Then the most promising of these languages could be extended to make fuller use of the richness of the underlying model. This approach would result in large savings for the development of information systems requiring multidisciplinary techniques, which are acquiring an ever-increasing importance.

The definition of a general purpose common language is not among the objectives of ModelAge, as we believe that multiple languages are necessary, each one offering a natural treatment of a specific domain, that are integrated by semantic rather than syntactic means.




Next: Results Up: Description Previous: Description


Pierre-Yves Schobbens
Fri Nov 25 10:58:38 WET 1994