Can We Trust an AI Agent? Interaction Effects of Its Machine Learning Performance and Digital Character

Abstract

AI-powered digital characters (i.e., AI agents) are expanding their scope of application to various fields. However, research on the key factors influencing AI users' attitude is insufficient. This study investigates the role of machine learning (ML) performance (as the behavioral/intelligence realism of AI agents) in determining users' trust. The study further investigates the interaction role of different forms of digital character (as the form realism of AI agents) in the relationship between ML performance and trust. The findings achieved from an experimental setting provide a novel understanding of human-AI interaction, expand academic understanding of AI anthropomorphism, and suggest new research directions for AI-powered digital characters. The results will also guide business practitioners in developing various AI services.