A MULTI-ISSUE TRUST MODEL IN MULTIAGENT SYSTEMS: A MATHEMATICAL APPROACH

Abstract

In the recent years, trust has become a crucial issue in studying agentbased distributed autonomous systems such as Semantic Web or Peerto-Peer, in which software agents need to select the most trustworthy partners to interact. Most current computational trust models are mainly based on two basic factors:  ersonal experience trust and reference trust on a single issue of trust. These models may be not very fruitful when applying to trust systems with multi-issue, in which agents need to infer a trust of some new issue from trusted issues. This status occurs due to lack of information or uncertainty on both experience trust and reference trust of the issue. In this paper, we introduce a trust model that is an extension of the single issue trust one to a multi-issue trust one. Our approach is to investigate a new type of trust - inference trust, and then to integrate it into this extension model. The new trust may enable agents to discover his local knowledge about their partners to infer the new trust of their partners on some issue.