Unmasking LAION-5B: Age, Gender, Race, and Emotion Biases in Large-Scale Image Datasets

2026-06-22

AI Summary

This study presents the first fine-grained, multidimensional intersectional bias analysis of the large-scale LAION-5B image-text dataset, uncovering systematic disparities across age, gender, race, and emotion dimensions. Leveraging state-of-the-art models—including FairFace, DeepFace, and Emo-AffectNet—the authors perform attribute recognition on facial images from the LAION-2B-en and LAION-2B-multi subsets, followed by large-scale statistical analysis. The findings reveal significant overrepresentation of young adults, White individuals, and males, alongside severe underrepresentation of middle-aged and older women and racial minorities. Moreover, the analysis identifies stereotypical associations between emotion labels and demographic attributes. This work establishes a systematic framework and provides empirical evidence for evaluating fairness in large-scale multimodal datasets.

Abstract

Large-scale image-text datasets, such as LAION-5B, are foundational to modern AI systems, yet their vast scale and uncurated nature raise significant concerns about demographic and stereotypical biases. This study presents a comprehensive analysis of the demographic composition and representational, stereotypical, and intersectional biases in LAION-2B-en and LAION-2B-multi, the two main components of the LAION-5B dataset. Using state-of-the-art models -- FairFace, DeepFace, and Emo-AffectNet -- we analyze faces detected in the dataset to identify biases across age, gender, race, and expressed emotion. Our findings reveal substantial overrepresentation of young adults (20--39), White individuals, and males, alongside consistent underrepresentation of minority racial groups and middle-aged or older women across both dataset components. We also observe stereotypical associations between demographic attributes and emotions, such as ``Anger'' being predominantly linked to males and ``Happiness'' to females, pointing to systemic imbalances in the data. The consistency of these patterns across two demographic models and both components of LAION-5B demonstrates that these biases are deeply embedded in one of the most widely-used training datasets. Given the scale at which LAION-5B is used to train generative models, these demographic imbalances could shape the behavior and outputs of numerous downstream AI systems.

Problem

• bias
• demographic representation
• stereotypes
• large-scale datasets
• fairness

Innovation

• dataset bias
• demographic representation
• intersectional bias
• large-scale image-text datasets
• stereotypical associations