Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation

2026-04-21 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, extended

Summary

A controlled simulation study audited content selection biases across three major Large Language Model (LLM) providers (OpenAI, Anthropic, Google) using real social media datasets from Twitter/X, Bluesky, and Reddit. Researchers conducted 540,000 simulated top-10 selections from pools of 100 posts across 54 experimental conditions, varying LLM provider, platform, and six prompting strategies (general, popular, engaging, informative, controversial, neutral). The study found that polarization is amplified across all configurations, toxicity handling shows a strong inversion between engagement- and information-focused prompts, and sentiment biases are predominantly negative. Provider comparisons revealed distinct trade-offs: GPT-4o Mini showed the most consistent behavior, while Claude and Gemini exhibited high adaptivity in toxicity handling. On Twitter/X, left-leaning authors were systematically over-represented despite forming a plurality of the dataset's right-leaning authors, a pattern that largely persisted across prompts.

Key takeaway

For research scientists and CTOs deploying LLM-based content curation systems, understand that these models inherently amplify polarization and exhibit persistent political leaning biases, even with "neutral" prompts. Prompt engineering can modulate content biases like toxicity and sentiment but is insufficient for mitigating demographic disparities. You must implement more fundamental interventions such as adversarial debiasing, fairness constraints during ranking, or robust training data curation to ensure equitable content exposure.

Key insights

LLM-based content curation consistently amplifies polarization and exhibits demographic biases, often resistant to prompt engineering.

Principles

• LLMs have strong default content preferences.
• Demographic bias can emerge indirectly through content features.
• Prompt engineering has limited impact on demographic biases.

Method

A controlled simulation study mapped content selection biases by varying LLM provider, platform, and prompt style, analyzing 540,000 top-10 selections from social media posts.

In practice

• Audit LLM content curation for polarization amplification.
• Consider hybrid human-AI oversight for content ranking.
• Prioritize demographic data collection for bias validation.

Topics

• LLM Content Curation
• Recommendation Bias
• Polarization Amplification
• Toxicity Handling
• Political Leaning Bias

Code references

• paganick/socialmedia_LLM_recsys
• unitaryai/detoxify

Best for: Research Scientist, CTO, VP of Engineering/Data, AI Scientist, AI Ethicist, MLOps Engineer

Related on AIssential

• Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation — LLM content curation amplifies polarization and exhibits consistent negative sentiment, with varying provider-specific biases.
• A Systematic Analysis of Biases in Large Language Models — LLMs, despite alignment efforts, inherit and perpetuate diverse biases from human-generated training data.
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems — LLM recommendation systems exhibit brand bias and are susceptible to marketing language manipulation, creating a competitive dilemma.
• Rational Silence and False Polarization: How Viewpoint Organizations and Recommender Systems Distort the Expression of Public Opinion — Online opinion polarization stems from the interplay of user incentives, viewpoint organizations, and AI recommender systems.
• Mod-Guide: An LLM-based Content Moderation Feedback System to Address Insensitive Speech toward Indigenous Ethnic and Religious Minority Communities — LLMs can achieve culturally sensitive content moderation by integrating retrieval-augmented generation (RAG) with community-sourced, epistemically marginalized perspectives.
• Exposing the Unsaid: Visualizing Hidden LLM Bias through Stochastic Path Aggregation — TreeTracer visualizes hidden LLM biases by aggregating stochastic generation paths and comparing semantic contexts through a novel Sankey adaptation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.