Investigating Social Bias in Narrative Image Generation

2026-08-03

AI Summary

This study addresses the underexplored issue of social bias in text-to-image (T2I) models when generating narrative visual formats such as storyboards and comics. Extending the textual bias evaluation framework BBG to the image generation domain, this work presents the first systematic comparison of bias manifestations across six T2I models in three visual formats: photographs, storyboards, and comics. Through multimodal bias analysis and cross-format experiments, the study demonstrates that narrative visual formats significantly amplify social biases, revealing explicit mechanisms in event sequencing, character positioning, narrative outcomes, and textual elements. Empirical results show that proprietary models exhibit biased outputs in 25.9% of photographic generations on average, with rates increasing by 9.6 and 18.2 percentage points in storyboards and comics, respectively, confirming that narrative structures more readily expose latent societal biases.

Abstract

Text-to-image (T2I) generation models are increasingly embedded in applications such as media content creation and education, raising concerns about how their outputs may reproduce social biases. Prior work has shown that T2I models exhibit social biases, yet existing evaluations largely focus on a photo generation task. As a result, it remains unclear whether and how such biases manifest in more narrative visual formats, such as storyboards and comics, where characters and events are presented across multiple panels. In this work, we compare bias expression across photo, storyboard, and comic generation in six T2I models by adapting BBG, a text-based bias evaluation framework, to image generation. Our results show that proprietary models generate 25.9% biased outputs in photo generation on average, with biased outputs increasing by 9.6pp in storyboard generation and 18.2pp in comic generation. We also find that photos mainly encode biases through subtle visual cues, while storyboards and comics reveal them more explicitly through event sequencing, character positioning, narrative resolution, and textual elements. These findings show that biases that remain less visible in photo generation may surface in narrative visual formats, highlighting the importance of evaluating T2I systems with diverse visual formats beyond photo generation.

Problem

• social bias
• text-to-image generation
• narrative image generation
• storyboard
• comic

Innovation

• text-to-image generation
• social bias evaluation
• narrative image formats
• bias manifestation
• BBG adaptation