Artificial intelligence (AI), don't surprise me and stay in your lane: An experimental testing of perceiving humanlike performances of AI

Abstract

This study tests how evaluations of various performances of artificial intelligence (AI) influence the willingness to use AI technologies. We examined the influences of (a) whether expectations about AI performances are met, (b) the valence of the violation, and (c) the types of AI performances, using a 2 (expectancy violation vs. confirmation) × 2 (positive vs. negative evaluation) × 2 (humanlike vs. machine-distinctive performance) design. The relationship between attitude toward AI and intention to use AI was also analyzed. Participants (n = 238) in an online survey read randomly assigned reading materials about AI performances. We found that people prefer to use AI with machine-distinctive performances instead of AI with humanlike performances. Also, the violation of expectations about AI performances was found to be a factor that affects the intention to use AI technologies. Finally, there was a two-way interaction effect between AI performance types and the expectancy violation. The implications for expectancy violations theory and perceiving the humanness of AI performances are discussed.