Should AI Trust Its Gut? The Cost of Deliberative Moral Reasoning in Humans and AI
Abstract
People prefer deliberation to intuition, when judging how others solve hard problems, and they extend this to artificial agents. Yet in the moral domain, the preference may reverse as intuitive moral responses appear to signal good character. One explanation is that fast, intuitive responses show real emotional engagement. Since machines are thought to lack emotion, it is unclear whether a preference for intuition in the moral domain extends to AI. Across three preregistered studies (N = 2,173), we find that it does, but that it appears to be driven by commitment to moral norms and precedents rather than emotion. In Study 1, participants preferred intuitive reasoning for moral decisions and systematic reasoning for non-moral decisions, for human and AI agents alike. Intuitive moral reasoning signalled better character but was not clearly expected to produce better outcomes. In Study 2, the character preference persisted when the AI was described as being incapable of emotion. In Study 3, we held the decision fixed and varied how firmly the agent held the relevant moral norm. An agent who would never reconsider applying a moral norm was rated above one who would reconsider given sufficiently high stakes; and both above one who weighed costs and benefits case by case. Commitment raised character ratings more than expected-outcome ratings. The character-based cost of moral deliberation appears to be a cost of treating a moral norm as negotiable, and we found strong evidence that people apply this inference to machines and people alike.
Related
- NMR-Challenge for LLMs: Evaluating Chemical Reasoning in Humans and AI
- Public trust and moral perceptions as predictors of anticipated preventive behavior toward future epidemics among adults in Riyadh, Saudi Arabia: a cross-sectional study.
- Consolidation of shared sustainability visions through deliberative multicriteria evaluation in Ecuadorian cocoa farming communities
- A Qualitative Exploratory Study of Pathways to Moral Repair and Moral Injury in U.S. Veterans
- Physicians in the Image of Christ the Healer: The Vocation and Moral Limits of Medicine
BibTeX
@misc{myers2026should,
title = {{Should AI Trust Its Gut? The Cost of Deliberative Moral Reasoning in Humans and AI}},
author = {Simon Myers and Nick Chater},
howpublished = {PsyArXiv},
year = {2026},
month = oct,
doi = {10.31234/osf.io/2fv9t_v1},
url = {https://osf.io/preprints/psyarxiv/2fv9t_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Should AI Trust Its Gut? The Cost of Deliberative Moral Reasoning in Humans and AI
AU - Simon Myers
AU - Nick Chater
PY - 2026
DA - 2026/10/22
PB - PsyArXiv
DO - 10.31234/osf.io/2fv9t_v1
UR - https://osf.io/preprints/psyarxiv/2fv9t_v1/
AB - People prefer deliberation to intuition, when judging how others solve hard problems, and they extend this to artificial agents. Yet in the moral domain, the preference may reverse as intuitive moral responses appear to signal good character. One explanation is that fast, intuitive responses show real emotional engagement. Since machines are thought to lack emotion, it is unclear whether a preference for intuition in the moral domain extends to AI. Across three preregistered studies (N = 2,173), we find that it does, but that it appears to be driven by commitment to moral norms and precedents rather than emotion. In Study 1, participants preferred intuitive reasoning for moral decisions and systematic reasoning for non-moral decisions, for human and AI agents alike. Intuitive moral reasoning signalled better character but was not clearly expected to produce better outcomes. In Study 2, the character preference persisted when the AI was described as being incapable of emotion. In Study 3, we held the decision fixed and varied how firmly the agent held the relevant moral norm. An agent who would never reconsider applying a moral norm was rated above one who would reconsider given sufficiently high stakes; and both above one who weighed costs and benefits case by case. Commitment raised character ratings more than expected-outcome ratings. The character-based cost of moral deliberation appears to be a cost of treating a moral norm as negotiable, and we found strong evidence that people apply this inference to machines and people alike.
ER -