Explainable Recommendation Systems and Their Effects on Consumer Trust and Purchase Decisions

Recommendation systems increasingly explain why an item is recommended, through statements such as “because you viewed” or “customers like you also bought”, on the premise that explanation builds trust and trust drives purchase. This paper examines how explanation type interacts with recommendation accuracy in shaping trust and purchase. A simulation of 1,500 consumers compares no explanation, feature-based explanations that cite the consumer’s own behaviour and preferences, and social-proof explanations that cite other consumers, across recommendation accuracy from 0.5 to 0.95. Feature-based explanations raise trust from 0.42 to 0.60 and purchase from 34% to 45% on average, and their benefit grows with accuracy, since the explanation makes accurate recommendations more credible; social-proof explanations raise trust less (0.46) and lower it below the no-explanation baseline when accuracy is low (0.31 against 0.34), because social proof attached to a poor recommendation signals that the system does not know the consumer. The paper interprets the results through trust theory, in which explanation supports the competence and transparency judgements that trust rests on, derives design guidance that matches explanation type to the system’s accuracy and to the consumer’s data history, and discusses the risk that explanations persuade without informing. A field design is specified.