Reasoning about Uncertain Information and Conflict Resolution through Trust Revision

Abstract

In information driven MAS, information consumers collect information about their environment from various sources such as sensors. However, there is no guarantee that a source will provide the requested information truthfully and correctly. Even if information is provided only by trustworthy sources, it can contain conflicts that hamper its usability. In this paper, we propose to exploit such conflicts to revise trust in information. This requires a reasoning mechanism that can accommodate domain constraints, uncertainty, and trust. Our formalism - 5DL-Lite - is an extension of a tractable subset of Description Logics with Dempster-Shafer theory of evidence. SDL-Lite allows reasoning about uncertain information and enables conflict detection. Then, we propose methods for conflict resolution through trust revision and analyse them through simulations. We show that the proposed methods allow reasonably accurate estimations of trust in information in realistic settings. Copyright © 2013, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.