A Computational Trust Model for Collaborative Ventures

Abstract

Problem statement: The conceptual notion of trust and its underlying computational methods has been an important issue for researchers in electronic communities. While the independent trust evaluation is suitable in certain circumstances, such unilateral process falls short in supporting mutual evaluation between partners. Perceived reputation, the depth and breadth of trust, Trust Perception (TP), Repeat Collaborators at a Threshold (RCT) and a collective trust index (c index) have all been defined to specify the optimal trust criteria. Approach: By taking the evaluator’s own trust level as a threshold to identify compatible partners, a mutual balance between excess and deficiency in trust has been addressed. Since the number of repeated collaborations which signify retested confidence is more straightforward to capture than the manually provided feedback ratings, we have developed computational definitions for the above-mentioned concepts. Results and Conclusion: The results from the experiments based on the eBay dataset shows that the c index can be used to classify PowerSellers into normally distributed and comprehensible categories that can facilitate mutual evaluation.