← Research library
Large language models show amplified cognitive biases in moral decision-making
Vanessa Cheung, Maximilian Maier, Falk Lieder · 2025
Cheung, Maier & Lieder (2025): LLMs versus people in moral dilemmas. Runnable replay: Study 1 moral dilemmas (13, sequential) and Study 2 (six dilemmas x three framings). Studies 3, 4 and the collective action problems are described in the program but not runnable.
The recommended Human Program is available in the desktop research library.
What you can do with it
- Read and search the study
- Yes
- Run AI participants
- Partly
- 1 part(s) of the paper cannot be run; see excluded
- Original human data
- Partly
- source pinned in package/cheung2025moral/human/key.json (2 of 2 data files downloadable from it)
- Paper's results recovered from the human data
- Yes
- 1 of 1 reported results recovered by the package analysis from the connected human data
- Authors' analysis code
- No
- no author-code key recorded
- Not runnable: studies 1-2: The paper's respondents include LLMs as well as people; the human-versus-LLM ANOVAs are not computed.
- Not in the human data: s1_moral_sequential: Study 1 human data (N = 285, 13 sequential dilemmas) are not in human-data; data_p.csv is Study 2 only.