Large language models show amplified cognitive biases in moral decision-making

• 成果类型: Article
• 署名作者: Cheung, Vanessa; Maier, Maximilian; Lieder, Falk
• 署名单位: University of London; University College London; University of California System; University of California Los Angeles
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-10336
• DOI: 10.1073/pnas.2412015122
• 发表日期: 2025-06-24
• 关键词: omission; dilemmas; JUDGMENT; advice; CONSEQUENCES; utilitarian; Consistency; PSYCHOLOGY; inclusion; doctrine

摘要

As large language models (LLMs) become more widely used, people increasingly rely on them to make or advise on moral decisions. Some researchers even propose using LLMs as participants in psychology experiments. It is, therefore, important to understand how well LLMs make moral decisions and how they compare to humans. We investigated these questions by asking a range of LLMs to emulate or advise on people's decisions in realistic moral dilemmas. In Study 1, we compared LLM responses to those of a representative U.S. sample (N = 285) for 22 dilemmas, including both collective action problems that pitted self-interest against the greater good, and moral dilemmas that pitted utilitarian cost-benefit reasoning against deontological rules. In collective action problems, LLMs were more altruistic than participants. In moral dilemmas, LLMs exhibited stronger omission bias than participants: They usually endorsed inaction over action. In Study 2 (N = 474, preregistered), we replicated this omission bias and documented an additional bias: Unlike humans, most LLMs were biased toward answering no in moral dilemmas, thus flipping their decision/advice depending on how the question is worded. In Study 3 (N = 491, preregistered), we replicated these biases in LLMs using everyday moral dilemmas adapted from forum posts on Reddit. In Study 4, we investigated the sources of these biases by comparing models with and without fine-tuning, showing that they likely arise from fine-tuning models for chatbot applications. Our findings suggest that uncritical reliance on LLMs' moral decisions and advice could amplify human biases and introduce potentially problematic biases.