Bootstrapping Trust with Partial and Subjective Observability

Assessment of trust and reputation typically relies on prior experiences of a trustee agent, which may not exist, e. g. especially in highly dynamic environments. In these cases stereotypes can be used, where traits of trustees can be used as an indicator of their behaviour during interactions. Communicating observations of traits to witnesses who are unable to observe them is difficult, however, when the traits are interpreted subjectively. In this paper we propose a mechanism for learning translations between such subjective observations, evaluating it in a simulated marketplace. CCS Concepts •Computing methodologies → Multi-agent systems;

Authors

• Phillip Taylor
• Nathan Griffiths
• Lina Barakat
• Simon Miles

Keywords

• Trust and reputation
• Stereotypes
• Machine learning