Cluster‐based improvement rates for trust establishment models in single or distributed multi‐agent systems

Abstract Intelligent agents within open and dynamic multi‐agent systems are becoming increasingly capable in their decision‐making abilities and rely upon the notion of trustworthiness to determine which agents to interact with. To improve the overall performance of trust establishment models which trustees individually select and equip to improve their trustworthiness with trustors, while balancing the resources being spent, a cluster‐based trust establishment model update mechanism is proposed. This cluster‐based approach is applicable to robust trust establishment models which utilize dynamic improvement and disimprovement rate variables to adjust a trustee's behaviors toward trustors to improve or maintain trust with the trustor. By storing a single trust establishment model's dynamic improvement and disimprovement rate variables independently for each trustor and by clustering similar trustors together based on observed experiences, a model can more accurately update a trustee's behaviors toward trustors. Through simulated experiments comparing the performance of the existing integrated trust establishment (ITE) model with and without the cluster‐based approach, with varying trustor to trustee ratios to diversify the agent behaviors, the cluster‐based approach consistently improves a trustee's ability to fully meet a trustor's needs, for less resources than ITE, while minimizing the corresponding impact to the trustee's overall trust.