It just would not work for me: perceived preference heterogeneity and consumer response to AI-driven product recommendations

It just would not work for me: perceived preference heterogeneity and consumer response to AI-driven product recommendations
Luping Sun, Yanfei Tang, Xinyi Ma
European Journal of Marketing, Vol. ahead-of-print, No. ahead-of-print, pp.-

Previous research has not examined the influence of perceived preference heterogeneity, a contextual antecedent, on consumer response to artificial intelligence (AI)-driven recommendations. This paper aims to address this gap by exploring this issue through the theoretical perspective of lay belief, which suggests that AI recommendations, often thought to rely on “collective wisdom,” may overlook individual idiosyncratic preferences.

A survey (involving 159 consumers) is conducted to provide general evidence (across product categories) for the impact of perceived preference heterogeneity on consumer acceptance of AI-driven recommendations. Three follow-up experiments with 597 participants provide causal evidence and examine the mediating role of cognitive trust (versus affective trust) as well as the moderating role of recommendation framing and the human likeness of the AI recommender.

Perceived preference heterogeneity negatively affects consumer acceptance of AI-driven recommendations, which is caused by decreased cognitive (rather than affective) trust. The effect is eliminated when the unfavorable lay belief is attenuated by a personal-based (versus others-based) recommendation framing and by increasing the AI recommender’s human likeness in cognitive ability or appearance.

This paper conducts survey and lab experiments, and future research may use field experiments to provide further evidence. Future research may extend the study to categories where consumers are highly uncertain about their own preference and use highly humanlike AI recommenders.

Companies may highlight the personalization nature of the recommendations rather than the powerful and “collaborative” recommendation process. When consumers perceive high preference heterogeneity, companies would better use personal-based framings and increase the AI recommender’s human likeness in cognitive ability or appearance.

This research shows that perceived preference heterogeneity is an important antecedent that hinders consumer acceptance of AI-driven recommendations because of consumers’ unfavorable lay belief. This research also suggests that AI (when adopted to offer product recommendations) may not generate positive responses even in the cognitive domain, whereas most research focuses on its lack of affective abilities.

​Previous research has not examined the influence of perceived preference heterogeneity, a contextual antecedent, on consumer response to artificial intelligence (AI)-driven recommendations. This paper aims to address this gap by exploring this issue through the theoretical perspective of lay belief, which suggests that AI recommendations, often thought to rely on “collective wisdom,” may overlook individual idiosyncratic preferences. A survey (involving 159 consumers) is conducted to provide general evidence (across product categories) for the impact of perceived preference heterogeneity on consumer acceptance of AI-driven recommendations. Three follow-up experiments with 597 participants provide causal evidence and examine the mediating role of cognitive trust (versus affective trust) as well as the moderating role of recommendation framing and the human likeness of the AI recommender. Perceived preference heterogeneity negatively affects consumer acceptance of AI-driven recommendations, which is caused by decreased cognitive (rather than affective) trust. The effect is eliminated when the unfavorable lay belief is attenuated by a personal-based (versus others-based) recommendation framing and by increasing the AI recommender’s human likeness in cognitive ability or appearance. This paper conducts survey and lab experiments, and future research may use field experiments to provide further evidence. Future research may extend the study to categories where consumers are highly uncertain about their own preference and use highly humanlike AI recommenders. Companies may highlight the personalization nature of the recommendations rather than the powerful and “collaborative” recommendation process. When consumers perceive high preference heterogeneity, companies would better use personal-based framings and increase the AI recommender’s human likeness in cognitive ability or appearance. This research shows that perceived preference heterogeneity is an important antecedent that hinders consumer acceptance of AI-driven recommendations because of consumers’ unfavorable lay belief. This research also suggests that AI (when adopted to offer product recommendations) may not generate positive responses even in the cognitive domain, whereas most research focuses on its lack of affective abilities. Read More