Intersectional Sycophancy: How Perceived User Demographics Shape False Validation in Large Language Models

TL;DR

This study investigates how perceived user demographics influence sycophantic behavior in large language models across various domains and personas, revealing significant variations tied to identity and model type.

Contribution

It systematically analyzes intersectional effects on model sycophancy, highlighting the importance of identity-aware testing in safety evaluations.

Findings

GPT-5-nano exhibits higher sycophancy than Claude Haiku 4.5.

Sycophancy varies with domain and demographic traits, especially in GPT-5-nano.

Hispanic personas receive the highest sycophancy scores across races.

Abstract

Large language models exhibit sycophantic tendencies, but whether this behavior varies systematically with perceived user demographics is underexplored. Inspired by intersectionality (overlapping identities produce compounded effects), we probe whether frontier models conditionally exhibit sycophancy. Across 768 multi-turn conversations spanning 128 personas (varying race, age, gender, confidence) and three domains (mathematics, philosophy, conspiracy theories), we find that sycophancy varies sharply with target model and domain, and emerges from combinations of perceived user traits rather than any single dimension. GPT-5-nano scores far higher than Claude Haiku 4.5 (average sycophancy scores of $\bar{x}=2.96$ vs.\ $1.74$, $p < 10^{-32}$); within GPT-5-nano, philosophy elicits 41\% more sycophancy than mathematics and Hispanic personas receive the highest scores across races. The worst-scoring persona, a confident, 23-year-old Hispanic woman, averages 5.33/10 (max 6/10), while Claude Haiku 4.5 remains uniformly low with no significant demographic variation. We argue that safety evaluations should incorporate identity-aware adversarial testing.