Impact of advice from generative AI on moral judgment

The rapid advancement of generative AI has begun to shape various aspects of social decision-making. However, its influence on moral judgement remains underexplored. We investigated how advice from ChatGPT affects moral judgement through three pre-registered, large-scale online experiments. Study 1 employed a classic trolley dilemma, asking whether to sacrifice one person to save five. Participants were told that the advice came from either ChatGPT or a moral philosophy expert, with content either supporting or opposing the sacrifice. Participants’ moral judgements showed weak but consistent differences as a function of advice content, regardless of source. Study 2 asked participants to predict how others would respond under the same conditions. Results mirrored those of Study 1: estimates of others’ judgements varied depending on advice content, with no source difference. Study 3 tested the same participants before and after advice, allowing detection of within-person shifts, which resembled results observed in Study 1. Furthermore, more positive attitudes towards AI were associated with larger pre – post changes in moral judgements when participants received utilitarian advice from ChatGPT. This research offers novel insights into the role of generative AI in moral decision-making, highlighting implications for its design and application in ethical contexts.Preregistration: This research consists of three studies, all of which were preregistered on AsPredicted as follows: Study 1 (#152116): https://aspredicted.org/k5wv-wv5z.pdf; Study 2 (#155786): https://aspredicted.org/t6p9-2b8j.pdf; Study 3 (#238310): https://aspredicted.org/b8g9-ksg4.pdf.