Product Perceptual Similarity Evaluation: From Attributive Error to Human Knowledge Hierarchy

Abstract

A crucial part of any recommendation system for manufactured goods is the method used to evaluate the similarity of products. However, despite being central to the performance of many online retailers, current ways of evaluating product similarity are inconsistent with several characteristics of human perception and consequently often generate obvious errors when applied in practice. This paper proposes a new approach that uses the results of a perceptual evaluation of respondents to determine their “inherent knowledge hierarchy.” Using this characterization of the user, we propose a new strategy for determining the perceived similarity of products (PSIM strategy). After a preliminary verification of the proposed approach, we found that the performance of the PSIM strategy is much better than current strategies in terms of both accuracy and robustness. Beyond the application of PSIM in product recommendation systems, the findings of this study have the potential to help designers and companies better understand their customers’ emotional needs.