How Bias Binds: Measuring Hidden Associations for Bias Control in Text-to-Image Compositions

2025-11-10

AI Summary

In text-to-image generation, implicit bias amplification arises from semantic binding between objects and sensitive attributes within prompts; existing debiasing methods typically address single-object, isolated-attribute scenarios, neglecting the joint contextual effects of such bindings. This work introduces the “bias attachment score” — the first metric quantifying the activation strength of specific object–attribute bindings toward bias. Building upon this, we propose a context-aware, model-agnostic debiasing framework that explicitly decouples semantic bindings via prompt disentanglement and associative modeling—without requiring model retraining. Experiments demonstrate over 10% improvement in debiasing performance on composite prompt generation tasks. Our findings reveal the pivotal role of semantic binding in bias amplification and expose a fundamental limitation of current debiasing approaches in complex, context-rich settings.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Text-to-image generative models often exhibit bias related to sensitive attributes. However, current research tends to focus narrowly on single-object prompts with limited contextual diversity. In reality, each object or attribute within a prompt can contribute to bias. For example, the prompt"an assistant wearing a pink hat"may reflect female-inclined biases associated with a pink hat. The neglected joint effects of the semantic binding in the prompts cause significant failures in current debiasing approaches. This work initiates a preliminary investigation on how bias manifests under semantic binding, where contextual associations between objects and attributes influence generative outcomes. We demonstrate that the underlying bias distribution can be amplified based on these associations. Therefore, we introduce a bias adherence score that quantifies how specific object-attribute bindings activate bias. To delve deeper, we develop a training-free context-bias control framework to explore how token decoupling can facilitate the debiasing of semantic bindings. This framework achieves over 10% debiasing improvement in compositional generation tasks. Our analysis of bias scores across various attribute-object bindings and token decorrelation highlights a fundamental challenge: reducing bias without disrupting essential semantic relationships. These findings expose critical limitations in current debiasing approaches when applied to semantically bound contexts, underscoring the need to reassess prevailing bias mitigation strategies.

Problem

• Text-to-image models exhibit bias in object-attribute semantic bindings
• Current debiasing approaches fail with contextual attribute associations
• Reducing bias without disrupting semantic relationships remains challenging

Innovation

• Introduces bias adherence score for object-attribute bindings
• Develops training-free context-bias control framework
• Uses token decoupling to debias semantic relationships