Deconstructing Bias: A Multifaceted Framework for Diagnosing Cultural and Compositional Inequities in Text-to-Image Generative Models

TL;DR

This paper introduces the Component Inclusion Score (CIS), a new metric to evaluate and diagnose cultural biases in text-to-image models, highlighting significant disparities and proposing interventions for fairer AI-generated imagery.

Contribution

It develops and applies CIS to benchmark cultural biases in T2I models, providing insights into data and architecture factors affecting fairness and inclusivity.

Findings

CIS reveals significant bias gaps between Western and non-Western prompts.

Data imbalance and model architecture influence cultural fairness in image generation.

Benchmarking shows potential interventions to improve inclusivity in T2I models.

Abstract

The transformative potential of text-to-image (T2I) models hinges on their ability to synthesize culturally diverse, photorealistic images from textual prompts. However, these models often perpetuate cultural biases embedded within their training data, leading to systemic misrepresentations. This paper benchmarks the Component Inclusion Score (CIS), a metric designed to evaluate the fidelity of image generation across cultural contexts. Through extensive analysis involving 2,400 images, we quantify biases in terms of compositional fragility and contextual misalignment, revealing significant performance gaps between Western and non-Western cultural prompts. Our findings underscore the impact of data imbalance, attention entropy, and embedding superposition on model fairness. By benchmarking models like Stable Diffusion with CIS, we provide insights into architectural and data-centric interventions for enhancing cultural inclusivity in AI-generated imagery. This work advances the field by offering a comprehensive tool for diagnosing and mitigating biases in T2I generation, advocating for more equitable AI systems.