In centralized collaborative information recommendation, there is a bottleneck for the scalability and the availability. Yenta is a decentralized approach. Information can be recommended through the clustering. The clustering of similar agents in Yenta is based on the similarities between the content of their interest models. So the agents must represent the interests of users or the models of other agents in the same way. This paper introduces the TRUST! system. In TRUST!, the agent's user evaluates the information recommended from other agents. The trust degree between different agents is decided by the user's evaluation. The similarities between different agents are measured by the trust connections. So the agents can represent the interests of users or the models of other agents in different ways. TRUST! is a distributed multiagent system for Internet information propagation and recommendation. It is based on limited friends list and trust relationship. To update trust, we introduce the PID (proportion, integral, and differential coefficient) arithmetic. Through emulation experiments, we give some analysis of the distributed clustering.
[1]
Michael N. Huhns,et al.
Agents on the Web: Benevolent Agents
,
1999,
IEEE Internet Comput..
[2]
Yoav Shoham,et al.
Learning to surf: multiagent systems for adaptive web page recommendation
,
1998
.
[3]
Tomas Olsson.
Information Filtering with Collaborative Interface Agents
,
1998
.
[4]
Bradley N. Miller,et al.
GroupLens: applying collaborative filtering to Usenet news
,
1997,
CACM.
[5]
Loren Terveen,et al.
PHOAKS: a system for sharing recommendations
,
1997,
CACM.
[6]
Stephen Marsh,et al.
Formalising Trust as a Computational Concept
,
1994
.
[7]
Robert B. Ross,et al.
Impact: a platform for collaborating agents
,
1999,
IEEE Intell. Syst..
[8]
N. Luhmann.
Familiarity, Confidence, Trust: Problems and Alternatives
,
2000
.
[9]
Olli Lagenspetz,et al.
Legitimacy and Trust
,
1992
.