Trust: A Distributed Cognition Approach. Technical Report 2002-12

We outline the notion of trust and two formal models developed to solve the trust problem in multi-agent systems. The limitations of these models, and their similarities with classical AI models (which stress centrally-stored representations of the world) are examined. We then consider trust as a Distributed Cognition problem, and suggest an agent design framework, inspired by Distributed Cognition. A distributed model of trust is then developed, extending work by Bacharach and Gambetta on trust in signs, and based on Zahavi’s Handicap Principle (a theory of animal signaling that emphasizes the role of costs in ensuring signal reliability). We apply this model to agent systems to suggest a programming language that can act as an institution to partially solve the trust problem.