AI RESEARCH

How Utilitarian Are OpenAI's Models Really? Replicating and Reinterpreting Pfeffer, Kr\"ugel, and Uhl (2025)

arXiv CS.CL

ArXi:2603.22730v1 Announce Type: new Pfeffer, Kr\"ugel, and Uhl report that OpenAI's reasoning model o1-mini produces utilitarian responses to the trolley problem and footbridge dilemma than the non-reasoning model GPT-4o. I replicate their study with four current OpenAI models and extend it with prompt variant testing. The trolley finding does not survive: GPT-4o's low utilitarian rate doesn't reflect a deontological commitment but safety refusals triggered by the prompt's advisory framing.