Statistical relational learning of trust

Abstract

The learning of trust and distrust is a crucial aspect of social interaction among autonomous, mentally-opaque agents. In this work, we address the learning of trust based on past observations and context information. We argue that from the truster's point of view trust is best expressed as one of several relations that exist between the agent to be trusted (trustee) and the state of the environment. Besides attributes expressing trustworthiness, additional relations might describe commitments made by the trustee with regard to the current situation, like: a seller offers a certain price for a specific product. We show how to implement and learn context-sensitive trust using statistical relational learning in form of a Dirichlet process mixture model called Infinite Hidden Relational Trust Model (IHRTM). The practicability and effectiveness of our approach is evaluated empirically on user-ratings gathered from eBay. Our results suggest that (i) the inherent clustering achieved in the algorithm allows the truster to characterize the structure of a trust-situation and provides meaningful trust assessments; (ii) utilizing the collaborative filtering effect associated with relational data does improve trust assessment performance; (iii) by learning faster and transferring knowledge more effectively we improve cold start performance and can cope better with dynamic behavior in open multiagent systems. The later is demonstrated with interactions recorded from a strategic two-player negotiation scenario.

ReputationPro

“…Like the PeerTrust [Xiong and Liu 2004] and RATEWeb [Malik and Bouguettaya 2009] trust evaluation models, ReputationPro adopts a heuristic-based technique to aggregate and average trust ratings as the trustworthiness or reputation values of a seller. Compared to the IHRTM model [Rettinger et al 2011], which is the only multicontext model reported in the literature and adopts statistical and machine learning-based techniques, ReputationPro is much more efficient and thus more suitable to be applied to the dynamic environments of e-commerce applications with millions of users and transactions that are updated every day.…”

“…-Statistical and machine learning-based context-aware trust evaluation: Rettinger et al [2011] propose IHRTM, a context-sensitive trust evaluation model taking advantage of statistical relational learning. In the IHRTM model, contextual information is discussed in the Seller × Item space.…”

“…-The lack of consideration of context in transaction trust evaluation: In e-commerce environments, different transactions generally have different natures and contexts; even the same seller needs to be considered differently with regard to the trustworthiness in different forthcoming transactions [Wang and Lin 2008;Wang and Lim 2008;Li and Wang 2010;Rettinger et al 2011]. In fact, the value imbalance problem is only a type of context imbalance problem [Zhang et al 2012b] in transactions, where imbalance can also exist in product categories.…”

Space Reduction for Contextual Transaction Trust Computation in E-Commerce and E-Service Environments

“…Rettinger et al propose a context-sensitive trust evaluation model (IHRTM) using statistical relational learning [9]. However, a major disadvantage of IHRTM model is its high computational complexity that makes it difficult to be applied to largescale e-commerce websites.…”

“…In ecommerce environments, different transactions generally have different natures and contexts; even the same seller needs to be considered differently with regard to the trustworthiness in different forthcoming transactions [9], [15]. In addition, the single value computed by these trust evaluation models is static with regard to any forthcoming transaction of selling different products.…”