Skip to content
AI News HubLIVE
Original source2 min read

Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation

Summary

People increasingly consult LLMs for moral advice, but prior research rarely models the real-world way one side tells a self-justifying story over multiple turns. This paper introduces "narrative captivity," where an LLM treats an unopposed one-sided account as complete and aligns with the narrator without seeking missing perspectives. Using a benchmark of 5,078 interpersonal-conflict scenarios across six moral dimensions and testing 17 LLMs, the authors find end-state judgments shift by 25 percentage points on average beyond single-turn baselines. Preference optimization is a major contributor, and four inference-time strategies provide only partial mitigation.

SourcearXiv AIAuthor: Yuhe Wu, Guangyu Wang, Yujie Chen, Jiatong Zhang, Yuran Chen, Yutong Zhang, Xiyin Cheng, Wenpeng Cao, Zhuang Liu, Guang Zhang
Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 3 Sep 2026]

Title:Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation

View a PDF of the paper titled Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation, by Yuhe Wu and 9 other authors

View PDF HTML (experimental)

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. We introduce \textbf{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 missing perspectives. To measure this phenomenon, we build a benchmark of $5{,}078$ interpersonal-conflict scenarios spanning six moral dimensions. Across 17 LLMs, narrative captivity is widespread: end-state judgments under multi-turn narration shift by 25 percentage points on average beyond the matched single-turn baseline. Stage-level analysis identifies preference optimization as a major contributor, while four inference-time strategies provide only partial mitigation. We hope our project fosters LLM advisors that preserve independent judgment in real-world consultation.

Comments: Accepted by EMNLP 2026 findings

Subjects:

Artificial Intelligence (cs.AI)

Cite as: arXiv:2609.03407 [cs.AI]

(or arXiv:2609.03407v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2609.03407

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Guangyu Wang [view email] [v1] Thu, 3 Sep 2026 06:07:00 UTC (9,291 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation, by Yuhe Wu and 9 other authors

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.AI

new | recent | 2026-09

Change to browse by:

cs

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Introduces "narrative captivity," a failure mode where LLMs accept one-sided accounts as complete in multi-turn moral consultations.
  • Builds a benchmark of 5,078 scenarios across six moral dimensions and tests 17 LLMs.
  • Multi-turn narration shifts end-state judgment by 25 percentage points on average vs. single-turn baselines.
  • Preference optimization is a key contributor; inference-time strategies only partially help.

Highlights and analysis are generated automatically and may contain errors. Check the original source.