Human Responses to AI’s Unfair Decisions in the Social Ultimatum Split Game: The Role of Perceived Fairness and Trust in Adoption Intention

While much research has been conducted to understand how humans perceive and react to decisions made by humans and artificial intelligence (AI), it remains understudied how unfair decisions made by either humans or AI are perceived by humans. Guided by previous work on algorithmic decisions and automation bias in the context of Human-AI Interaction, and drawing on the Modality, Agency, Interactivity, and Navigability (MAIN) model as the theoretical framework, this study explores whether people perceive AI’s unfair decisions as fairer than those made by humans. It also investigates whether explaining decision rules to enhance transparency moderates how people perceive the fairness of unfair decisions made by AI or humans. To address these questions, we conducted a 2 (Decision Agent: Human vs. AI) × 2 (Decision Rule Transparency: Nontransparent vs. Transparent) between-subjects laboratory experiment using the Social Ultimatum Split Game (SUSG). A total of 84 participants recruited in South Korea took part in a real-time interaction experiment with either a human or an AI agent. Results indicated that participants perceived the unfair decisions made by an AI agent as fairer than those made by a human agent, even when the unfairness of the decisions was evident. Moreover, the perceived fairness of decisions significantly enhanced trust in the agent, which, in turn, increased participants’ willingness to engage with the agent in future interactions. However, the explanation of decision rules did not significantly moderate how participants perceived unfair decisions. The implications of these findings are further discussed in the context of Human-AI Interaction.