From Adam Smith to artificial intelligence: an experimental exploration of emotion in humanomics

Abstract This study examines whether individuals can effectively express their emotions or “fellow feelings” to a machine, an impartial observer, or individuals who have caused them harm, and the effects of these emotional expressions on behavior. Our findings indicate that the impartial observer is the most preferred outlet for emotional expression, with neither the machine nor the offending party being able to adequately fulfill this role. Negative emotions are more commonly expressed to an impartial observer and a machine, whereas positive emotions are more often directed toward the offending party. Additionally, our findings indicate that emotional expression cannot replace costly punishment and that communication channels do not affect the behavior of the offending party. By exploring the nuanced ways emotions are conveyed, our study offers valuable insights into the evolving discourse on the impact of emotions on human-AI interactions.