Stance Drift: How AI-mediated Communication Distorts Our Message

2026-10-03

AI Summary

This study addresses the distortion and polarization drift of user stances in large language model (LLM)-mediated communication. We construct a two-stage “generate-extract” evaluation pipeline and introduce a novel quantification framework, Stance Preservation Rate (SPR), grounded in a probabilistic state transition model to systematically reveal stance drift mechanisms in AI-mediated scenarios. Mitigation strategies, including reflective prompting, are further incorporated for optimization. Experimental results demonstrate that SPR consistently falls below 0.7 under default configurations, confirming a significant fidelity gap; however, optimization elevates GPT-5.4’s SPR to 0.775. This work establishes a new paradigm for evaluating and enhancing stance consistency in AI-mediated communication.

Abstract

Large language models (LLMs) increasingly mediate human communication, from drafting emails to summarizing scientific reports, yet whether they faithfully preserve a speaker's position remains largely untested. We model AI-mediated communication as a two-step generation-extraction pipeline: one LLM produces an argument from a specified stance, and a second LLM extracts the stance from that argument. We represent the pipeline as a probabilistic state transition over five Likert-type stance categories and define the stance preservation rate (SPR) as the average probability that the extracted stance matches the initial stance. Across 112 debate propositions, none of the nine LLMs tested exceeded an SPR of 0.7 under the default configuration. Three drift patterns accounted for most of the drift: polarization, deviation from neutrality, and flipping. Among the mitigation strategies tested, including in-context learning, multiple extraction with shuffled options, assertion, and reflection, only adding medium reasoning effort to a reflection prompt for GPT-5.4 substantially improved the SPR, to 0.775, yet polarization remained the largest pattern, with 0.119 of the transition mass. An exploratory comparison with human extraction on a single proposition suggests that drift arises at both the generation and the extraction stage. These results point to a fidelity gap in AI-mediated communication, with implications for journalism, policy deliberation, scientific communication, and other domains where opinion-laden messages pass through language models.

Problem

• Stance Drift
• AI-mediated Communication
• Large Language Models
• Stance Preservation
• Fidelity Gap

Innovation

• Stance Drift
• AI-mediated Communication
• Large Language Models
• Probabilistic State Transition
• Stance Preservation Rate