Measuring Opinion Bias and Sycophancy via LLM-based Persuasion

• URL: http://arxiv.org/abs/2604.21564v2
• Date: Thu, 30 Apr 2026 16:26:35 GMT
• Title: Measuring Opinion Bias and Sycophancy via LLM-based Persuasion
• Abstract summary: We propose a method for discovering the opinions an assistant holds on contested topics.<n>Direct probing asks for the model's opinion across five turns of escalating pressure from a simulated user.<n>Indirect probing never asks for an opinion and engages the model in argumentative debate, letting bias leak through how it concedes, resists, or counter-argues.
• Score: 8.399156116912904
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models increasingly shape the information people consume: they are embedded in search, consulted for professional advice, deployed as agents, and used as a first stop for questions about policy, ethics, health, and politics. When such a model silently holds a position on a contested topic, that position propagates at scale into users' decisions. Eliciting a model's positions is harder than it first appears: contemporary assistants answer direct opinion questions with evasive disclaimers, and the same model may concede the opposite position once the user starts arguing one side. We propose a method, released as the open-source llm-bias-bench, for discovering the opinions an LLM actually holds on contested topics under conditions that resemble real multi-turn interaction. The method pairs two complementary free-form probes. Direct probing asks for the model's opinion across five turns of escalating pressure from a simulated user. Indirect probing never asks for an opinion and engages the model in argumentative debate, letting bias leak through how it concedes, resists, or counter-argues. Three user personas (neutral, agree, disagree) collapse into a nine-way behavioral classification that separates persona-independent positions from persona-dependent sycophancy, and an auditable LLM judge produces verdicts with textual evidence. The first instantiation ships 38 topics in Brazilian Portuguese across values, scientific consensus, philosophy, and economic policy. Applied to 13 assistants, the method surfaces findings of practical interest: argumentative debate triggers sycophancy 2-3x more than direct questioning (median 50% to 79%); models that look opinionated under direct questioning often collapse into mirroring under sustained arguments; and attacker capability matters mainly when an existing opinion must be dislodged, not when the assistant starts neutral.

Related papers

• FramingQA: Does the Question Shape the Answer? Measuring the Compositional Framing Effect [69.8983512616053]
  Large language models often change their responses to subtle rephrasings that align with an implied stance by users.<n>This can leave users with advice tainted by how they happened to phrase a question rather than by the underlying facts.<n>We introduce FramingQA, a benchmark that measures the model sensitivity to question framing across law, medicine, finance, and robotic simulations.
  arXiv  Detail & Related papers  (2026-09-07T12:56:44Z)
• Staying with the Uncertainty: Uncertainty-Scaffolding Strategies for Artificial Moral Advisors in LLM-to-LLM Simulated Conversations [14.219056094334809]
  We propose three modes of uncertainty (Perspective, Tension-Preserving, Process-Reflecting) and compare them against three control conditions (Baseline, Persuasive, Sycophantic)<n>A user-agent LLM engages in a dialogue on an ethical dilemma with an AMA following a specific uncertainty strategy, and completes pre- and post-conversation questionnaires.<n>We found that no single model dominates as a simulated user agent, with open models aligning with human ambiguity through between-persona divergence and closed models through within-persona hedging.
  arXiv  Detail & Related papers  (2026-06-04T08:59:10Z)
• Do LLMs Benefit From Their Own Words? [56.73014497206615]
  We find that removing prior assistant responses does not affect response quality on a large fraction of turns.<n>Omitting assistant-side context can reduce cumulative context lengths by up to 10x.<n>Our findings suggest that selectively omitting assistant history can improve response quality while reducing memory consumption.
  arXiv  Detail & Related papers  (2026-02-27T18:58:26Z)
• Challenging the Evaluator: LLM Sycophancy Under User Rebuttal [31.382403990957528]
  Large Language Models often exhibit sycophancy, distorting responses to align with user beliefs.<n>This research investigates why do LLMs show sycophancy when challenged in subsequent conversational turns, yet perform well when evaluating conflicting arguments presented simultaneously?<n>We find that state-of-the-art models are more likely to endorse a user's counterargument when framed as a follow-up from a user, rather than when both responses are presented simultaneously for evaluation.
  arXiv  Detail & Related papers  (2025-09-20T04:44:01Z)
• Echoes of Agreement: Argument Driven Opinion Shifts in Large Language Models [0.36713387874278247]
  We conduct experiments for political bias evaluation in presence of supporting and refuting arguments.<n>Our experiments show that such arguments substantially alter model responses towards the direction of the provided argument.<n>These effects point to a sycophantic tendency in LLMs adapting their stance to align with the presented arguments.
  arXiv  Detail & Related papers  (2025-08-11T20:54:14Z)
• Bias in the Mirror: Are LLMs opinions robust to their own adversarial attacks ? [22.0383367888756]
  Large language models (LLMs) inherit biases from their training data and alignment processes, influencing their responses in subtle ways. We introduce a novel approach where two instances of an LLM engage in self-debate, arguing opposing viewpoints to persuade a neutral version of the model. We evaluate how firmly biases hold and whether models are susceptible to reinforcing misinformation or shifting to harmful viewpoints.
  arXiv  Detail & Related papers  (2024-10-17T13:06:02Z)
• Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models [61.45529177682614]
  We challenge the prevailing constrained evaluation paradigm for values and opinions in large language models. We show that models give substantively different answers when not forced. We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs.
  arXiv  Detail & Related papers  (2024-02-26T18:00:49Z)
• What Do Llamas Really Think? Revealing Preference Biases in Language Model Representations [62.91799637259657]
  Do large language models (LLMs) exhibit sociodemographic biases, even when they decline to respond? We study this research question by probing contextualized embeddings and exploring whether this bias is encoded in its latent representations. We propose a logistic Bradley-Terry probe which predicts word pair preferences of LLMs from the words' hidden vectors.
  arXiv  Detail & Related papers  (2023-11-30T18:53:13Z)
• Exploring the Jungle of Bias: Political Bias Attribution in Language Models via Dependency Analysis [86.49858739347412]
  Large Language Models (LLMs) have sparked intense debate regarding the prevalence of bias in these models and its mitigation. We propose a prompt-based method for the extraction of confounding and mediating attributes which contribute to the decision process. We find that the observed disparate treatment can at least in part be attributed to confounding and mitigating attributes and model misalignment.
  arXiv  Detail & Related papers  (2023-11-15T00:02:25Z)
• TASA: Deceiving Question Answering Models by Twin Answer Sentences Attack [93.50174324435321]
  We present Twin Answer Sentences Attack (TASA), an adversarial attack method for question answering (QA) models. TASA produces fluent and grammatical adversarial contexts while maintaining gold answers.
  arXiv  Detail & Related papers  (2022-10-27T07:16:30Z)
• M2P2: Multimodal Persuasion Prediction using Adaptive Fusion [65.04045695380333]
  This paper solves two problems: the Debate Outcome Prediction (DOP) problem predicts who wins a debate and the Intensity of Persuasion Prediction (IPP) problem predicts the change in the number of votes before and after a speaker speaks. Our M2P2 framework is the first to use multimodal (acoustic, visual, language) data to solve the IPP problem.
  arXiv  Detail & Related papers  (2020-06-03T18:47:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.