LLM silicon surrogates for arts participation surveys exhibit positive liking bias, lose taste relationality, and fail to preserve known social space alignments.
Algorithmic Fidelity of Large Language Models in Generating Synthetic G erman Public Opinions: A Case Study
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CL 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
Using DML and mediation analysis on AllSides articles, only fine-tuned GPT-4o-mini exhibits significant sentiment-to-ideology causal effects at community level, unlike human annotators, indicating spurious coupling learned during fine-tuning.
citing papers explorer
-
Not-quite-human tastes: the stylized omnivorousness of LLM survey surrogates
LLM silicon surrogates for arts participation surveys exhibit positive liking bias, lose taste relationality, and fail to preserve known social space alignments.
-
Does Topic Sentiment Cause Perceived Ideology? Comparing Human and LLM Annotations in Political News Articles
Using DML and mediation analysis on AllSides articles, only fine-tuned GPT-4o-mini exhibits significant sentiment-to-ideology causal effects at community level, unlike human annotators, indicating spurious coupling learned during fine-tuning.