Decision Making in Dynamic Environments Based on Interdependent Trust States1

Many models of trust consider the trust an agent has in a trustee as the result of experiences with that specific trustee in combination with certain personality attributes. Such trust states are a key element for decision making, thereby for instance deciding what agent to request for a certain piece of information, or to delegate a task to. In order to understand this entire process within humans and also enable computational formation and utilization of trust, it is important to model the various aspects that play a crucial role: (1) to develop a model representing the dynamics of trust states in individuals, and (2) to describe a decision making mechanism that considers the trust states when making decisions. For the former part, it is important to realize that formation of trust on multiple trustees over time can involve interdependencies between trustees, whereby an experience with one trustee may affect the trust in another trustee. Trust models which incorporate this interdependence phenomenon are presented in this paper, both in the form of a new model and as an extension of previously developed models. Second, a decision making model is presented which uses the trust states formed by the aforementioned models and outputs decisions on which trustee to consult. Hereby, a trade-off between exploration and exploitation is made. The models are evaluated and compared to other existing models by means of simulations.