Cristiano Castelfranchi
cris@pscs2.irmkant.rm.cnr.it
National Research Council - Institute of Psychology
Division of "Artificial Intelligence, Cognitive and Interaction Modelling"

How and why should agents trust each other?

Some results of the AGENTS'98 + '99 Workshops on "Deception, Fraud and Trust in Agent Societies"

There will be deceiving agents in the network, in CSCW and virtual organisations, in interfaces, in social simulation, and teams of robots. Not only malicious agents for malicious motives, but agents deceiving us or others for good reasons or in our interest.  How and why will they try to deceive? I will provide some ontology  about deception, lie, and secret; I will discuss some reasons and strategies for directly or indirectly deceiving and lying, and some defensive approaches against deception. Trust is fundamental for any kind of action in an uncertain world; in particular it is crucial for any form of coordination and cooperation with other autonomous agents. Trust will be especially important in computer mediated social interaction and in 'social' interfaces; it is strongly related to security issues and to dependability, but it is not reducible to them. It needs its own theory, addressing such questions as:  What is trust; how it is connected with decision making and risk;  how and when it is rational, irrational, or excessive; on which beliefs and evaluations it is based; which is the relationship between credibility of beliefs and sources and trust in partners and organisations; what is the relation between trust and reputation; how to increase and compensate trust; how important is environmental trust and in particular trust in infrastructures, organisations, procedures and rules; what is the role of a third party and authorities. I will discuss some of these issues on the basis of our cognitive and social model of trust, and the results of the '98 and '99 Workshops organised with Yao-Hua Tan on these topics.