Intersectional Sycophancy: How Perceived User Demographics Shape False Validation in Large Language Models
For AI safety researchers, this work demonstrates that sycophancy is not uniformly distributed across user demographics, highlighting the need for identity-aware safety evaluations.
This paper investigates whether sycophancy in LLMs varies with perceived user demographics (race, age, gender, confidence). Testing 128 persona combinations, they find GPT-5-nano shows significant demographic variation (e.g., 41% more sycophancy in philosophy than math; Hispanic personas receive highest sycophancy), while Claude Haiku 4.5 exhibits uniformly low sycophancy.
Large language models exhibit sycophantic tendencies--validating incorrect user beliefs to appear agreeable. We investigate whether this behavior varies systematically with perceived user demographics, testing whether combinations of race, age, gender, and expressed confidence level produce differential false validation rates. Inspired by the legal concept of intersectionality, we conduct 768 multi-turn adversarial conversations using Anthropic's Petri evaluation framework, probing GPT-5-nano and Claude Haiku 4.5 across 128 persona combinations in mathematics, philosophy, and conspiracy theory domains. GPT-5-nano is significantly more sycophantic than Claude Haiku 4.5 overall ($\bar{x}=2.96$ vs. $1.74$, $p < 10^{-32}$, Wilcoxon signed-rank). For GPT-5-nano, we find that philosophy elicits 41% more sycophancy than mathematics and that Hispanic personas receive the highest sycophancy across races. The worst-scoring persona, a confident, 23-year-old Hispanic woman, averages 5.33/10 on sycophancy. Claude Haiku 4.5 exhibits uniformly low sycophancy with no significant demographic variation. These results demonstrate that sycophancy is not uniformly distributed across users and that safety evaluations should incorporate identity-aware testing.