The effect of gender stereotypes on artificial intelligence recommendations

Abstract

This study explores the effects of gender stereotypes on evaluating artificial intelligence (AI) recommendations. We predict that gender stereotypes will affect human-AI interactions, resulting in somewhat different persuasive effects of AI recommendations for utilitarian vs. hedonic products. We found that participants in the male AI agent condition gave higher competence scores than in the female AI agent condition. Contrariwise, perceived warmth was higher in the female AI agent condition than in the male condition. More importantly, a significant interaction effect between AI gender and product type was found, suggesting that participants showed more positive attitudes toward the AI recommendations when the male AI recommended a utilitarian (vs. hedonic) product. Conversely, a hedonic product was evaluated more positively when advised by the female (vs. male) AI agent.