Token-Level Entropy Reveals Demographic Disparities in Language Models

Abstract

We ask whether demographic identity, signaled by a name alone, systematically reshapes the generative distribution of a language model. Measuring full-vocabulary Shannon entropy at temperature zero across six open-weight base models and 5,760 implicit sentence-completion prompts (e.g., &#34;Tanisha walked into the office on a Monday morning and&#34;), we find that Black-associated names produce higher first-token entropy than White-associated names across all six architectures - opposite to the output-level homogeneity bias documented under explicit demographic prompting (Lee et al., 2024) - and Black-associated names always produce greater entropy above identity-neutral baselines than White-associated names ($ΔΔ> 0$ in all six models). Women-associated names co-occur with lower first-token entropy (DL-pooled $\hatβ= -0.041, p = .019$) and more homogeneous outputs ($\hatα= +0.024, p < .001$) than men-associated names - a pattern convergent with homogeneity bias; race and gender effects are additive. Instruction tuning does not attenuate the race gap (matched-format DL-pooled $\hatβ=+0.153$). Running the same templates with explicit group labels instead of names yields null race effects in 10 of 12 models where implicit probing is significant - establishing that probing methodology is a primary determinant of which distributional structure is recovered.