Normative Conformity in Language Models

Overview

This field examines how language models respond to social norms, peer consensus, authority, group identity, and local conversational cues—and whether those responses reflect norm understanding or simple imitation. Researchers use controlled peer-pressure question-answering tasks, social dilemmas, group-chat norm adaptation benchmarks, and multi-agent simulations, often varying whether the prevailing answer is correct and whether a norm is explicit or inferred. Models can change answers under social pressure, including toward incorrect claims, while reasoning prompts sometimes help; behavior also varies with culture, identity, and context. A central challenge is balancing resistance to misleading influence with receptivity to correct advice, while distinguishing a model’s stated conformity from its underlying beliefs or genuine normative competence.

Latest papers

CardsList

33 papers6 papers in the last four weeks, up 20% on the four weeks before. 0.0% of all new papers.

  1. How Do Language Models Process Ethical Instructions? Deliberation, Consistency, and Other-Recognition Across Four Models

    Mar 11, 2026Hiroki FukuiMoral Reasoning in Language ModelsLLM Safety Alignment