Exploring the Role of Post-hoc Explanations in Mitigating Algorithm Aversion in Identity-Based Consumption: An Eye-Tracking Study

Customers have a general tendency to discount algorithmic over human recommendations, a phenomenon commonly known as “algorithm aversion.” Within areas driven by identity-based consumption such as fashion, designing efficient recommender systems is particularly challenging due to highly individualistic preferences and tastes. In this study, we analyze algorithm aversion towards fashion recommender systems with regards to social and personal identity and post-hoc explanations of algorithmic recommendations. In line with self-categorization theory and theory of planned behavior, we hypothesize that, to minimize algorithm aversion, the post-hoc explanations of algorithmic recommendations need to target customers’ salient identity. Accordingly, we propose a 3 × 3 between-subject experiment with eye-tracking, where participants are shown several pairs of algorithm- or human-based fashion recommendations. In the treatment groups, we either activate customers’ social or personal identity, while the explanations of algorithmic recommendations emphasize the customers’ mainstream or unique taste. Furthermore, we expect that consumers with activated social or personal identity are more likely to report a different preference than their preference measured by the first and total number of eye fixations. Thereby, we expect to extend IS research on algorithm aversion and post-hoc explanations of algorithms towards identity-based consumption. In addition, our findings have practical implications for online retailers.