A Trust Model for Multiagent Recommendations

This paper describes a trust model for multiagent recommender systems. A user’s request for a travel recommendation is decomposed by the system into subtasks, corresponding to travel services. Agents select tasks autonomously, and accomplish them using knowledge derived from previous solutions or with the help of other agents. Agents maintain local knowledge bases and, when requested to support a user in a travel planning task, they may collaborate exchanging information stored in their local bases. During this exchange process trusting other agents is fundamental. It helps agents to improve the quality of the recommendations and to avoid communication with unreliable agents. In the proposed model, the trust is also used to allow agents to become experts in particular subtasks, helping them to generate better recommendations. In this paper, we propose and validate a multiagent trust model showing the benefits of such model in a travel planning scenario.