Estimating the Impact of “Humanizing” AI Assistants

Abstract

More and more product designers are adopting anthropomorphic design strategies to facilitate the widespread use of AI assistants. However, existing studies show that the ideal boundary of the degree of anthropomorphism is still unclear. Therefore, we design two scenario experimental studies to explore the influence of different degrees of anthropomorphism on human–AI interaction quality. We also consider the difference in impact between two usage contexts: the hedonic and utilitarian contexts. The results show that different degrees of anthropomorphism significantly affect human–AI interaction quality in different ways; when AI assistants have a medium-level degree of anthropomorphism, the positive effect is pronounced. Furthermore, the positive relationship between anthropomorphism and human–AI interaction quality is more robust in the hedonic usage context; this positive effect disappears in the utilitarian usage context. We expect this study to provide practical guidance for AI product designers to achieve their products’ long-term feasibility and sustainability.