EmergencyBias: Bias in Text-to-Image Models under Emergency Scenarios

2026-08-01

AI Summary

This study investigates and evaluates demographic and behavioral biases in text-to-image generative models within emergency scenarios, introducing EmergencyBias—a unified conceptual framework that characterizes systemic disparities in role assignment, risk exposure, and intervention behaviors. The authors construct a comprehensive evaluation benchmark encompassing seven state-of-the-art models, six emergency contexts, and three sociodemographic dimensions. Through multidimensional prompt-controlled experiments and quantitative metrics, they analyze bias manifestations under both unspecified and controlled prompting conditions. To mitigate these biases, the work proposes ActionAlign, a lightweight calibration method that adjusts prompt embeddings to significantly reduce behavioral unfairness while preserving image generation quality, outperforming existing baselines.

Abstract

Bias in Text-to-Image (T2I) generation has become an important problem in multimedia content creation and communication. However, existing studies have primarily focused on relatively static and explicit forms of bias, such as disparities in the representation of gender, race, and geo-cultural attributes. Less attention has been paid to behavioral bias in how different groups are portrayed acting, reacting, and occupying social roles. Emergency scenarios provide a revealing setting for studying such bias because they require models to depict not only who is present, but also who is at risk, who intervenes, and how responsibility is allocated. In this paper, we define EmergencyBias, a form of bias in T2I generation under emergency scenarios that includes both demographic bias and behavioral bias. We construct an evaluation framework to systematically study EmergencyBias across seven leading T2I models, six representative emergency scenarios, and three demographic dimensions. Our experimental results reveal three main findings. First, under blank prompts without demographic specification, T2I models exhibit clear demographic bias in emergency scenarios, reflected in the distributions of portrayed individuals across gender, age, and skin tone. Second, under controlled prompts, behavioral bias in emergency responses remains systematically associated with demographic variation, with particularly pronounced disparities along gender and substantial differences across models. Third, we introduce ActionAlign, a lightweight prompt-embedding calibration method that outperforms a representative prompt-based baseline in reducing behavioral disparities while largely preserving image quality. Overall, our work identifies emergency scenarios as an important setting for bias evaluation in T2I models and offers a practical direction toward fairer visual generation in socially consequential contexts.

Problem

• EmergencyBias
• Text-to-Image generation
• behavioral bias
• demographic bias
• emergency scenarios

Innovation

• EmergencyBias
• behavioral bias
• text-to-image generation
• prompt embedding calibration
• ActionAlign