Reputation, Trust and Recommendation Systems in Peer-to-Peer Environments Position Paper

The tremendous popularity of the Internet has brought about the notion of peer-to-peer computing, whose reliance on a central authority (let alone a central server) is absolutely minimal. It seems fair to say that one of the Great Promises of the Internet is that such unmoderated direct interaction between users will reduce much of the traditional overhead due to the “man in middle ” taking his share: either financially (in the context of e-commerce) or conceptually (in the context of opinion shaping, say). The flip side of this prospect, of course, is the danger that the system will deteriorate into a lawless jungle: some users in a peer-to-peer system, possibly coordinated, might exploit honest users to their advantage, since it appears that there is no effective way to enforce the rules of the game. Consider eBay for example, where users can buy and sell stuff. The immense success of this application of peer-to-peer computing is quite surprising for people who are used consider worst-case scenarios, since eBay is “ripe with the possibility of large scale fraud and deceit ” [5]. The answer, according to most experts, lies in eBay’s closely watched reputation system [8]: after every transaction, the system invites each party to rate the behavior of the other party (the grades are “positive,” “negative ” and “neutral”). The system maintains a public billboard that records all feedbacks received for each user. It is common knowledge that consulting the billboard is a key step before making