Community-Aligned Behavior Under Uncertainty: Evidence of Epistemic Stance Transfer in LLMs

2025-11-14

AI Summary

This study investigates whether large language models (LLMs), after alignment to a specific online community, generalize behavior patterns reflecting that community’s cognitive stance—rather than merely reproducing superficial correlations from training data. Method: We propose the “cognitive stance transfer” evaluation framework, integrating targeted factual knowledge deletion, multi-dimensional behavioral probing, and comparative analysis against community-specific corpora, empirically tested on Russian–Ukrainian military discourse and U.S. partisan Twitter data. Contribution/Results: Even after systematic removal of key factual knowledge, aligned models consistently reproduce community-specific uncertainty-handling strategies—including attribution biases, affective modulation, and argumentative structures—demonstrating that alignment encodes deep, persistent behavioral priors. This work provides the first empirical evidence that LLM community alignment induces transferable, entrenched cognitive stances, establishing a novel paradigm for studying alignment risks and socio-cognitive modeling.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

When large language models (LLMs) are aligned to a specific online community, do they exhibit generalizable behavioral patterns that mirror that community's attitudes and responses to new uncertainty, or are they simply recalling patterns from training data? We introduce a framework to test epistemic stance transfer: targeted deletion of event knowledge, validated with multiple probes, followed by evaluation of whether models still reproduce the community's organic response patterns under ignorance. Using Russian--Ukrainian military discourse and U.S. partisan Twitter data, we find that even after aggressive fact removal, aligned LLMs maintain stable, community-specific behavioral patterns for handling uncertainty. These results provide evidence that alignment encodes structured, generalizable behaviors beyond surface mimicry. Our framework offers a systematic way to detect behavioral biases that persist under ignorance, advancing efforts toward safer and more transparent LLM deployments.

Problem

• Testing whether aligned LLMs transfer epistemic stances beyond training data
• Evaluating community-specific behavioral patterns under knowledge uncertainty
• Detecting persistent behavioral biases in LLMs after factual knowledge removal

Innovation

• Targeted deletion of event knowledge
• Validated with multiple probing methods
• Evaluated community response patterns under ignorance