Formal trust model for multiagent systems

TLDR: Despite a more subtle definition than previous approaches, this paper establishes a bijection between evidence and trust spaces, enabling robust combination of trust reports and provides an efficient algorithm for computing this bijection.

Abstract: Trust should be substantially based on evidence. Further, a key challenge for multiagent systems is how to determine trust based on reports from multiple sources, who might themselves be trusted to varying degrees. Hence an ability to combine evidence-based trust reports in a manner that discounts for imperfect trust in the reporting agents is crucial for multiagent systems. This paper understands trust in terms of belief and certainty: A's trust in B is reflected in the strength of A's belief that B is trustworthy. This paper formulates certainty in terms of evidence based on a statistical measure defined over a probability distribution of the probability of positive outcomes. This novel definition supports important mathematical properties, including (1) certainty increases as conflict increases provided the amount of evidence is unchanged, and (2) certainty increases as the amount of evidence increases provided conflict is unchanged. Moreover, despite a more subtle definition than previous approaches, this paper (3) establishes a bijection between evidence and trust spaces, enabling robust combination of trust reports and (4) provides an efficient algorithm for computing this bijection.