A research team led by Dr Prof. Hitoshi Yamamoto of Rissho University and Dr Prof. Takahisa Suzuki of Tsuda University explored the conditions under which people would accept the moral judgments of AI. They focused on the behaviour of "not helping people with bad reputations (justified non-cooperation)," which is difficult for people to judge as good or bad, to investigate under what conditions people are more likely to accept AI's judgments over human judgments. The study revealed that people tend to be more accepting of AI's judgments when AI makes positive judgments and humans make negative judgments. The research results were published in the online academic journal Scientific Reports of the Nature Publishing Group on 27th January 2025.
As AI technology becomes more integrated into daily life, understanding public acceptance of AI's decisions is critical. Previous studies have shown that people often hold biases, such as "algorithm aversion" and "algorithm appreciation," where they might distrust or over-trust AI. However, this study addresses the less-explored scenario where individuals find themselves uncertain in their moral judgment, specifically in indirect reciprocity, where individuals decide whether to cooperate with others based on reputation.
The researchers conducted two experiments with Japanese participants, examining how they judged an AI manager's decision compared to a human manager's in a workplace scenario. The key findings showed that participants were more inclined to accept the AI's decision when AI judged a non-cooperative action as positive (justified defection) and a human judged it as negative. The results suggest that individuals may perceive AI's judgment as more objective, especially when human judgments might be perceived as biased or driven by hidden intentions.
The findings contribute to a deeper understanding of the mechanisms behind people's acceptance of AI in moral and social decision-making, highlighting the importance of context in shaping these perceptions. As society continues integrating AI into complex decision-making roles, such insights are essential for designing AI systems that align with human expectations and societal norms.
