Text and Content Analysis
Academic Research
-
Working Paper
Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation
Working Paper, 2026
Large Language Models (LLMs) are increasingly deployed to curate and rank human-created content, yet the nature and structure of their biases in these tasks remains poorly understood: which biases are robust across providers and platforms, and which can be mitigated through prompt design. We present a controlled simulation study mapping content selection biases across three major LLM providers (OpenAI, Anthropic, Google) on real social media datasets from Twitter/X, Bluesky, and Reddit, using six prompting strategies (\textit{general}, \textit{popular}, \textit{engaging}, \textit{informative}, \textit{controversial}, \textit{neutral}). Through 540,000 simulated top-10 selections from pools of 100 posts across 54 experimental conditions, we find that biases differ substantially in how structural and how prompt-sensitive they are. 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 reveal distinct trade-offs: GPT-4o Mini shows the most consistent behavior across prompts; Claude and Gemini exhibit high adaptivity in toxicity handling; Gemini shows the strongest negative sentiment preference. On Twitter/X, where author demographics can be inferred from profile bios, political leaning bias is the clearest demographic signal: left-leaning authors are systematically over-represented despite right-leaning authors forming the pool plurality in the dataset, and this pattern largely persists across prompts.
-
Working Paper
Synthetic personas distort the structure of human belief systems
Working Paper, 2026
Large language models (LLMs) are increasingly used as synthetic survey respondents, yet it is unclear whether their belief-system structure matches that of real publics. We compare 28 LLMs to the 2024 General Social Survey (GSS) using 52 attitude items and demographic persona traits. We estimate polychoric correlation matrices and propagate un-certainty in the GSS via bootstrap resampling with multiple imputation. Constraint is measured by the variance share explained by the first principal component and by effective dependence, a determinant-based measure of global linear dependence. Across models, LLM personas exhibit substantially higher constraint than humans; conditioning on persona traits reduces constraint far more for LLMs, indicating greater demographic mediation. Projection onto a shared GSS basis further shows overemphasis of the leading dimension and missing secondary structure. These results caution against treating LLM personas as a reliable foundation for synthetic survey data generation.