Team Dynamics and Conflict Resolution

Understanding intelligent agents: analysis

and synthesis

John Fox, Martin Beveridge and David Glasspool

London Research Institute, Cancer Research UK, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK

E-mail: {john.fox,martin.beveridge,david.glasspool}@cancer.org.uk

Abstract. Current views of intelligent agent technologies are reviewed with respect to (a) their general cognitive capabilities

and (b) the classic Belief-Desire-Intention (BDI) model. A benchmark agent model is developed as a basis for analyzing and

comparing agent systems. PROforma is an agent technology that has grown out of work in modeling medical expertise and the

benchmark is used to carry out a case study analysis of this technology, looking at it from three contrasting points of view:

logic programming, object-oriented programming and agent-oriented programming. These viewpoints yield different insights

into the strengths and weaknesses of PROforma and lead to a clarification and consolidation of the benchmark agent features. The

consolidated model offers a useful framework for analysis and comparison of other agent systems in medicine or other domains.

Keywords: Agent programming, object oriented programming, logic programming

1. Introduction

Research on “intelligent agents” has diverse motivations,

ranging from general theoretical interests in AI

to practical objectives of building flexible distributed

systems in specific application domains. As agent technologies

mature so also does the interest in applying

them in medicine. Agent technologies offer advanced

platforms for building expert systems to assist individual

clinicians in their work [20] and distributed agent

systems have the potential to improve the operation of

healthcare organizations, where failures of communication

and coordination are important sources of error

[6].

As computer science advances, however, there is a

danger that older technologies are neglected. Worse

still the advocates of different computing paradigms

sometimes engage in unhelpful disputes about the relative

weaknesses of their preferred approaches, rather

than looking for ways of combining their different

strengths. Historical examples include the debates

about logic programming versus functional programming,

object-oriented databases versus relational databases

and rapid prototyping versus formal software engineering.

No one view has a monopoly on insight,

however, and it would be desirable if novel paradigms

did