From Simulations to Theorems: A Position Paper on Research in the Field of Computational Trust

摘要

Since the millennium, a quickly increasing number of research papers in the field of "computational trust and reputation" have appeared in the Computer Science literature. However, it remains hard to compare and evaluate the respective merits of proposed systems. We argue that rigorous use of formal probabilistic models enables the clear specification of the assumptions and objectives of systems, which is necessary for comparisons. To exemplify such probabilistic modeling, we present a simple probabilistic trust model in which the system assumptions as well as its objectives are clearly specified. We show how to compute (in this model) the so-called predictive probability: The probability that the next interaction with a specific principal will have a specific outcome. We sketch preliminary ideas and first theorems indicating how the use of probabilistic models could enable us to quantitatively compare proposed systems in various different environments.