Agents Selecting Trustworthy Recommendations in Mobile Virtual Communities

  • Authors:
  • G. Lenzini;N. Sahli;H. Eertink

  • Affiliations:
  • Telematica Instituut, Enschede, The Netherlands 7500 AN;Telematica Instituut, Enschede, The Netherlands 7500 AN;Telematica Instituut, Enschede, The Netherlands 7500 AN

  • Venue:
  • Trust in Agent Societies
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

We propose a trust model for recommender systems in mobile virtual communities that improves the quality of recommendations. High-quality recommendations come from trustworthy recommenders and are supported by convincing (from a user's viewpoint) arguments. Such recommendations match better a user's profile and his/her way of reasoning. We also propose a mobile open multi-agent framework for our model of trust. In this framework each member is represented by an embedded agent, which resides in the user's mobile, and by a delegate agent, which runs in a meeting infrastructure (called Virtual Agora) where it interacts with other delegates. Each delegate builds and maintains two weighted and subjective networks of trust: a register of recommenders of proven (un)trustworthiness and a register of rated items. In our setting, delegates generally prefer opinions that come from their network of recommenders; moreover delegates are able to play argumentation games. The implemented system is also privacy preserving. Agents receive different ratings depending on their trustworthy recommenders and these ratings are not necessarily public. Since we address mobile communities of users, the Virtual Agora is designed to reduce communication traffic between agents. Finally, we illustrate these concepts in a recommender system for restaurants.