Fundamentally Unfair: Analysis of Gendered Discourse in U.S. Capital Trials

Abstract

Research shows that in capital trials - i.e., cases in which the defendant could face the death penalty if convicted - prosecutors invoke a range of gender stereotypes against women defendants to increase the likelihood of convictions and death sentences. The Supreme Court ruled in Andrew v. White (2025) that a prosecutor's extensive references to a woman defendant's sex life could be so unduly prejudicial as to render her trial "fundamentally unfair." In the wake of Andrew, evidence of gender bias could be used to advance clemency proceedings or appeals of death sentences. We thus identify an urgent need for systematic methods to measure how gender bias is expressed in the courtroom. As a step towards that goal, we collect a novel corpus of 137 men's and women's U.S. capital trial transcripts, including the trials of almost all women currently on death row. We leverage simple NLP techniques to conduct a quantitative analysis of gender-associated key terms used in these trials. We find that, even when conditioning on defendant demographics and crime characteristics, terms associated with Gender Roles, Sexualization, Appearance, and LGBTQ+ Identity are used more frequently in the trials of women. Through a supplemental qualitative analysis, we show that the presence of gender-associated key terms can indeed be an indicator of gender stereotyping. Ultimately, we show how simple and interpretable NLP techniques can be used as a tool to identify gender-associated key terms in women's capital trials. This is a first step towards determining the extent to which gender bias has rendered a trial fundamentally unfair.