A change model for credibility partial order

LUCIANO H. TAMARGO; MARCELO A. FALAPPA; ALEJANDRO J. GARCÍA; GUILLERMO R. SIMARI

Conferencia; 5th International Conference on Scalable Uncertainty Management (SUM 2011); 2011

 In a multi-agent system (MAS), an agent may often receiveinformation through a potentially large number of informants. We willconsider the case where the informants are independent agents who havetheir own interests and, therefore, are not necessarily completely reliable;in this setup, it will be natural for some agent to believe an informantmore than other. The use of the notion of credibility will allow agents toorganize their peers in a partial order that will reflect the relative credi-bility of their informants. It is also natural that the assigned credibilitywill change dynamically, leading to changes in the associated partial or-der. We will investigate the problem of updating the credibility order toreflect the change in the perceived agent?s credibility, seeking to definea complete change theory over the agents? trust and reputation. The fo-cus will be on the characterization and development of change operators(expansion, contraction, and revision) for modeling the dynamics of thispartial order of agents. These operators, characterized through postu-lates and representation theorems, can be used to dynamically modifythe credibility of informants to reflect a new perception of informant?splausibility, or admit the arrival of a new agent to the system.