The Fallacy of Endogenous Discounting of Trust Recommendations

Recommendations are widely used in recommender systems, reputation systems, and trust-based security systems. Some existing reputation systems and trust-based security systems use the flawed notion of endogenous discounting. Endogenous discounting is the idea that claims deviating from prior expectations should be ignored or discounted, which introduces confirmation bias. To show the fallacy of endogenous discounting, we construct a semantic meta-model that captures the key notions surrounding recommendations. We prove that any model covered by the meta-model can be formulated in a divide-and-conquer fashion. We show how divideand-conquer clashes with endogenous discounting. Lastly, we discuss the implications on existing work that applies endogenous discounting, and provide suggestions for future work. Categories and Subject Descriptors I.2.11 [Distributed Artificial Intelligence]: Multiagent Systems Keywords Endogenous Discounting; Recommendation; Trust System; Rating

• Type: Journal
• Year: 2015
• Where: ATAL
• Authors: Tim Muller, Yang Liu, Jie Zhang
• Keywords: ATAL 2015, Intelligent Agents