A Dynamic Bayesian Network approach for agent group trust evaluation

Trust, an essential concept in human society, plays a crucial role in multi-agent systems by giving agents the confidence in making decisions as well as maintaining the well-being of transactions in the systems. Research of trust in multi-agent systems has focused heavily on trust evaluation of individual agents. When agent groups are becoming essential parts of agent society, existing trust models have shown several defects in addressing the highly dynamic behaviours of these targets. To this end, this paper presents a Dynamic Bayesian network Approach for Trust Evaluation (DBATE). It combines personalised criteria with multiple observations obtained from the interaction context to reveal the trustworthiness of the evaluated targets. We demonstrate the advantages of the model in enhancing the accuracy of trust computation as well as potential applications compared with other methods.