Research

Paper

AI LLM March 24, 2026

Between Rules and Reality: On the Context Sensitivity of LLM Moral Judgment

Authors

Adrian Sauter, Mona Schirmer

Abstract

A human's moral decision depends heavily on the context. Yet research on LLM morality has largely studied fixed scenarios. We address this gap by introducing Contextual MoralChoice, a dataset of moral dilemmas with systematic contextual variations known from moral psychology to shift human judgment: consequentialist, emotional, and relational. Evaluating 22 LLMs, we find that nearly all models are context-sensitive, shifting their judgments toward rule-violating behavior. Comparing with a human survey, we find that models and humans are most triggered by different contextual variations, and that a model aligned with human judgments in the base case is not necessarily aligned in its contextual sensitivity. This raises the question of controlling contextual sensitivity, which we address with an activation steering approach that can reliably increase or decrease a model's contextual sensitivity.

Metadata

arXiv ID: 2603.23114
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-24
Fetched: 2026-03-25 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.23114v1</id>\n    <title>Between Rules and Reality: On the Context Sensitivity of LLM Moral Judgment</title>\n    <updated>2026-03-24T12:08:16Z</updated>\n    <link href='https://arxiv.org/abs/2603.23114v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.23114v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>A human's moral decision depends heavily on the context. Yet research on LLM morality has largely studied fixed scenarios. We address this gap by introducing Contextual MoralChoice, a dataset of moral dilemmas with systematic contextual variations known from moral psychology to shift human judgment: consequentialist, emotional, and relational. Evaluating 22 LLMs, we find that nearly all models are context-sensitive, shifting their judgments toward rule-violating behavior. Comparing with a human survey, we find that models and humans are most triggered by different contextual variations, and that a model aligned with human judgments in the base case is not necessarily aligned in its contextual sensitivity. This raises the question of controlling contextual sensitivity, which we address with an activation steering approach that can reliably increase or decrease a model's contextual sensitivity.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CY'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.HC'/>\n    <published>2026-03-24T12:08:16Z</published>\n    <arxiv:comment>preprint</arxiv:comment>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Adrian Sauter</name>\n    </author>\n    <author>\n      <name>Mona Schirmer</name>\n    </author>\n  </entry>"
}