When AI is fairer than humans: The role of egocentrism in moral and fairness judgments of AI and human decisions

Algorithmic fairness is a core principle of trustworthy Artificial Intelligence (AI), yet how people perceive fairness in AI decision-making remains understudied. Prior research suggests that moral and fairness judgments are egocentrically biased, favoring self-interested outcomes. Drawing on the Computers Are Social Actors (CASA) framework and egocentric ethics theory we examine whether this bias extends to AI decision-makers, comparing fairness and morality perceptions of AI and human agents. Across three experiments (two preregistered, N = 1880, Prolific US samples), participants evaluated financial decisions made by AI or human agents. Self-interest was manipulated by assigning participants to conditions where they either benefited from, were harmed by, or remained neutral to the decision outcome. Results showed that self-interest significantly biased fairness judgments—decision-makers who made unfair but personally beneficial decisions were perceived as more moral and fairer than those whose decisions benefited others (Studies 1 & 2) or those who made fair but personally costly decisions (Study 3). However, this egocentric bias was weaker for AI than for humans, mediated by a lower perceived mind and reduced liking for AI (Studies 2 & 3). These findings suggest that fairness judgments of AI are not immune to egocentric biases, but are moderated by cognitive and social perceptions of AI versus humans. Our studies challenge the assumption that algorithmic fairness alone is sufficient for achieving fair outcomes. This provides novel insight for AI deployment in high-stakes decision-making domains, highlighting the need to consider both algorithmic fairness and human biases when evaluating AI decisions. • Self-interest biases moral and fairness judgements of both humans and AI decision-makers, leading people to perceive unfair decisions as more moral and fairer when they benefit from them. • The self-interest bias is stronger and more substantial for human decision-makers than for AI. • AI is judged more harshly due to lower perceived mind capabilities and reduced liking compared to human agents. • Findings challenge the notion that algorithmic fairness alone ensures just outcomes as human biases in fairness perception must also be addressed when designing AI systems.