AI’s Humanoid Appearance Can Affect Human Perceptions of Its Emotional Capability: Evidence from Self-Reported Data in the U.S

Abstract

As artificial intelligence (AI) integration becomes more prominent in society, the infosphere becomes increasingly complex. A better understanding of the human-AI relationship is crucial during these rapid technological and social changes. Emotions are complex responses of the human mind – a highly advanced information processing system. Thus, examining how humans perceive AI’s emotional capability may provide helpful insights. Human physiological properties such as humanoid appearance and gender are suggested to be potential influencing factors. Following the information processing approach, we employed Bayesian Mindsponge Framework (BMF) analytics on a dataset of 266 residents of the United States. Results show that interacting with an AI agent with human-like physical features is positively associated with the belief in that AI agent’s capability of experiencing emotional pain and pleasure. This effect is stronger in males than females. A possible explanation for these patterns is reliance on cognitive strategies used in interpersonal interactions, including gender-specific physiological properties. This study also demonstrates the effectiveness of using a compatible information processing analytical method such as BMF for studying humans’ perceptions of AI.