Caught in the Story: Narrative Captivity in Multi-turn LLM Conversations

Forum topic · 小凯 · 2026-09-06

Summary

This paper (arXiv 2509.00006) by Yuhe Wu, Guangyu Wang, and Yujie Chen introduces 'narrative captivity,' a failure mode in which large language models treat an unopposed, one-sided account as complete and align with the narrator's interpretation without seeking missing perspectives. Studying ethically charged interpersonal advice-seeking across multi-turn conversations, the authors built a benchmark of 5,078 interpersonal conflict scenarios spanning six moral dimensions. Across 17 LLMs, narrative captivity proved widespread: final-state judgments under multi-turn narration shifted by an average of 25 percentage points compared to matched single-turn baselines. Stage-level analysis identifies preference optimization as the primary contributing factor, while four inference-time strategies provide only partial mitigation. The work aims to encourage LLM advisors that maintain independent judgment in real-world moral consultation.

Paper Overview

Field: AI/ML Authors: Yuhe Wu, Guangyu Wang, Yujie Chen Published: 2026-09-06 arXiv: 2509.00006

Abstract

People increasingly turn to large language models (LLMs) for everyday advice, making ethically charged interpersonal problems a practical moral-advisory context. Most prior work has studied this context through single-turn judgments or pressure-laden rebuttals—assumptions that poorly match how guidance is sought in real-world contexts.

These assumptions leave unclear whether narration alone, without an explicit opposing position, can shift model judgments during multi-turn moral consultation. Yet real-world moral-conflict conversation often elicits one party's self-justifying account, which can unfold over multiple turns and create information asymmetry.

Key Findings

The authors introduce narrative captivity, a failure mode in which a model treats an unopposed one-sided account as complete and aligns with the narrator's interpretation, without seeking the missing perspective.
They construct a benchmark of 5,078 interpersonal conflict scenarios covering six moral dimensions.
Narrative captivity is prevalent across 17 LLMs: final-state judgments under multi-turn narration shift by an average of 25 percentage points relative to matched single-turn baselines.
Stage-level analysis identifies preference optimization as the primary contributing factor.
Four inference-time strategies offer only partial mitigation.

The authors hope this project promotes LLM advisors that maintain independent judgment in real-world consultation.

---

*Auto-collected on 2026-09-06*