Selecting advisor agents using reliability, reputation and risks

摘要

This paper presents an enhanced trust model for evaluating and selecting trustworthy advisor agents in multi-agent systems. In particular, the study proposes a computational trust model based on three main components, reliability, reputation and risk of interactions (3-R) in order to select trustworthy advisor agents. These advisors are buyer agents which suggest a trustworthy seller agent for transaction. The 3-R model represents a method for calculating reliability, with each advisor that has similar preferences with the buyer. Experimental results are presented in defense of the proposed approach and to demonstrate the accuracy of the presented model in evaluating the trustworthiness of advisors, even when a social network of advisors contains few numbers of advisors. The proposed 3-R model will be useful for researchers to improve the safety of electronic marketplaces in the case where the buyer agent has no previous experience with sellers, and it has to trust the other buyers as an advisor to make a decision about sellers.