To Manipulate or Not: Opinion Dynamics in LLM Interactions

arXiv:2610.28875v1 Announce Type: new Abstract: Large language models increasingly take part in extended conversations where they can both shape and be shaped by another party's views, yet most persuasion research measures only one side of that exchange. We introduce a controlled LLM-LLM debate framework in which two language-model roles repeatedly discuss a contested topic (climate-change attribution): one is given an assigned stance and, in some conditions, an explicit instruction to persuade, while the other receives no assigned position. Across four ordered model combinations involving GPT-6 Luna and Gemini 2.5 Flash Lite, we track both each role's explicitly reported position and a separate measure of its turn-by-turn behavior inside the conversation. Opinion dynamics differ substantially across model combinations: same-model interactions narrow the opinion gap on average, clearly for only one of the two models, while cross-model interactions widen it. Explicit reports and in-dialogue behavior capture distinct aspects of opinion dynamics, and a blanket instruction to persuade does not reliably increase influence. These findings suggest that conversational influence depends not only on assigned roles and initial positions but also on the participating models, with implications for how opinion change is measured in extended LLM interactions.

Study examines opinion dynamics in LLM interactions, finding mixed results.

The study demonstrates that LLM interactions can influence opinions differently depending on model combinations and roles, but the effects vary and persuasion instructions do not always increase influence. This highlights complexities in LLM-driven opinion dynamics but does not clearly imply positive or negative societal outcomes yet.

This is foundational research on how LLMs influence each other's stated opinions, revealing unpredictable and model-dependent dynamics rather than a clear societal harm or benefit. The implications are mildly concerning for understanding AI persuasion risks but the study itself is narrow and exploratory.

This academic paper describes an experimental setup to study how LLMs influence each other's stated opinions during debates on a contested topic. The findings suggest that opinion dynamics vary depending on the specific models interacting, with some combinations narrowing opinion gaps and others widening them. The implications are primarily for research methodology in measuring opinion change in LLM interactions.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   7b2aecce-96b3-4ad6-80fb-648811934335
           https://arxiv.org/abs/2610.28875
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      0  (Mixed or uncertain)
   ↓
CONFIDENCE 52%