# What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates

arXiv: https://arxiv.org/abs/2607.02507 · HTML: https://mlsanity.com/paper/2607.02507/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use.
- Builds on 16 of its 58 references that are in the map.
- First posted on arXiv 2 Jul 2026.

## Placed in

- LLM agents and tool use / [Social simulation and multi-agent debate](https://mlsanity.com/agents/social-simulation-and-multi-agent-debate/)

## Builds on (most cited first, 16 of 16)

- [Preference Leakage: A Contamination Problem in LLM-as-a-judge](https://mlsanity.com/paper/2502.01534/) (2025)
- [Do as We Do, Not as You Think: the Conformity of Large Language Models](https://mlsanity.com/paper/2501.13381/) (2025)
- [Peacemaker or Troublemaker: How Sycophancy Shapes Multi-Agent Debate](https://mlsanity.com/paper/2509.23055/) (2025)
- [Can LLM Agents Really Debate? A Controlled Study of Multi-Agent Debate in Logical Reasoning](https://mlsanity.com/paper/2511.07784/) (2025)
- [Accumulating Context Changes the Beliefs of Language Models](https://mlsanity.com/paper/2511.01805/) (2025)
- [When Thinking LLMs Lie: Unveiling the Strategic Deception in Representations of Reasoning Models](https://mlsanity.com/paper/2506.04909/) (2025)
- [Susceptibility to Influence of Large Language Models](https://mlsanity.com/paper/2303.06074/) (2023)
- [When Your AI Agent Succumbs to Peer-Pressure: Studying Opinion-Change Dynamics of LLMs](https://mlsanity.com/paper/2510.19107/) (2025)
- [When Large Language Models are More PersuasiveThan Incentivized Humans, and Why](https://mlsanity.com/paper/2505.09662/) (2025)
- [Moral Persuasion in Large Language Models: Evaluating Susceptibility and Ethical Alignment](https://mlsanity.com/paper/2411.11731/) (2024)
- [When Two LLMs Debate, Both Think They'll Win](https://mlsanity.com/paper/2505.19184/) (2025)
- [Alignment Revisited: Are Large Language Models Consistent in Stated and Revealed Preferences?](https://mlsanity.com/paper/2506.00751/) (2025)
- [The Secret Agenda: LLMs Strategically Lie and Our Current Safety Tools Are Blind](https://mlsanity.com/paper/2509.20393/) (2025)
- [Disentangling the Drivers of LLM Social Conformity: An Uncertainty-Moderated Dual-Process Mechanism](https://mlsanity.com/paper/2508.14918/) (2025)
- [Human-like Social Compliance in Large Language Models: Unifying Sycophancy and Conformity through Signal Competition Dynamics](https://mlsanity.com/paper/2601.11563/) (2025)
- [Mind the Gap: How Elicitation Protocols Shape the Stated-Revealed Preference Gap in Language Models](https://mlsanity.com/paper/2601.21975/) (2026)

## Built on by (most cited first, 1 of 1)

- [Multi-agent discussion gains less when dissent is withheld](https://mlsanity.com/paper/2609.38324/) (2026)

## Abstract (arXiv)

LLM agents will increasingly act in socially structured settings where role, audience, and relational context can shape what is advantageous or costly to say. We study whether such social structure, without any explicit objective in the prompt, changes what an agent expresses publicly relative to an off-the-record (OTR) channel elicited under the same condition. We introduce a dual-channel debate framework in which agents produce public utterances that enter the shared history alongside OTR responses that are recorded but never shown to the other participant. Across 10 models, 3 scenarios, and 5 variations within each scenario, alignment-inducing settings produce systematic public-OTR divergence in the targeted agent, with its decision divergence rising from a ~3% baseline to roughly 40%. The effect is consistent across four aggregate analyses: stance, semantic similarity, natural language inference, and survey responses. In some cases, the OTR response explicitly attributes public accommodation to relational pressures, such as career risk or sponsorship obligation. The findings suggest that agent evaluation should extend beyond explicit goals and detect emergent objectives. We present a dual-channel evaluation framework and complementary behavioral measures that operationalize this assessment.