Revisiting Trust and Reputation in Multi-agent Systems

Trust and reputation have been recognized as a key issue in the area of multi-agent systems. And till now, a number of researchers have, in different perspectives, reviewed/surveyed the notions, techniques or models of trust and reputation. However, the notion of trust and reputation still remains somewhat vague, and the techniques and model is yet understood in a disunited way. This paper categorizes the notions, techniques, and models of trust and reputation in terms of the trust management process which can be characterized through three questions: a) why does an agent trust another; b) how do agents judge or evaluate the trustworthiness of others; c) what does an agent do after obtaining the trustworthiness of others. These questions present a unified view of trust and reputation. The two aspects implied in the first question suggest two different understanding of the notion of trust. The second question refers to the techniques of getting and dealing with two different types of trust evidence. And it comes to the third question when trust models take agents' post-evaluation actions and the effect of these actions on their partners/opponents into account to make themselves incentive-compatible.