Computational Quantification of Trust Updates
Arthur Ramer · 2006
A computational model for expressions of trust values is outlined. It is based on the proposal by Jonker and Treur to base trust updates on reported experiences. The model handles arbitrary sequences of experience inputs; its such updates are fully commutative and associative. It satisfies all the axiomatic properties suggested for the trust values. Trust is interpreted as family of probabilistic beliefs on the space of possible experience reports. Expansion of the space of reports gives rise to inverse conditioning of probability distributions and thus of trust values. Belief and trust changes follow the AGM structure. Inverse conditioning is put into effect through a suitable application of maximum entropy principles