Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints

Abstract

Language is increasingly being used to define rich visual recognition problems with supporting image collections sourced from the web. Structured prediction models are used in these tasks to take advantage of correlations between co-occurring labels and visual input but risk inadvertently encoding social biases found in web corpora. In this work, we study data and models associated with multilabel object classification and visual semantic role labeling. We find that (a) datasets for these tasks contain significant gender bias and (b) models trained on these datasets further amplify existing bias. For example, the activity cooking is over 33% more likely to involve females than males in a training set, and a trained model further amplifies the disparity to 68% at test time. We propose to inject corpus-level constraints for calibrating existing structured prediction models and design an algorithm based on Lagrangian relaxation for collective inference. Our method results in almost no performance loss for the underlying recognition task but decreases the magnitude of bias amplification by 47.5% and 40.5% for multilabel classification and visual semantic role labeling, respectively.

arXiv comment11 pages, published in EMNLP 2017

References in corpus (6)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Semantics derived automatically from language corpora contain human-like biases
• Microsoft COCO Captions: Data Collection and Evaluation Server
• VQA: Visual Question Answering
• A Tutorial on Dual Decomposition and Lagrangian Relaxation for Inference in Natural Language Processing
• A survey on measuring indirect discrimination in machine learning

Cited by in corpus (33)

• Shortcut Learning in Deep Neural Networks
• EDITS: Modeling and Mitigating Data Bias for Graph Neural Networks
• Towards Accountability for Machine Learning Datasets: Practices from Software Engineering and Infrastructure
• Gradient Starvation: A Learning Proclivity in Neural Networks
• Unshuffling Data for Improved Generalization
• Pre-trained Language Models in Biomedical Domain: A Systematic Survey
• Towards Explainable Neural-Symbolic Visual Reasoning
• Learning Gender-Neutral Word Embeddings
• Towards Interpretable Natural Language Understanding with Explanations as Latent Variables
• Auditing and Achieving Intersectional Fairness in Classification Problems
• Fairness-Aware Explainable Recommendation over Knowledge Graphs
• Understanding Bias in Machine Learning
• Group-Fair Online Allocation in Continuous Time
• Measuring Social Biases of Crowd Workers using Counterfactual Queries
• Exposing and Correcting the Gender Bias in Image Captioning Datasets and Models
• Generating Clues for Gender based Occupation De-biasing in Text
• Jigsaw-VAE: Towards Balancing Features in Variational Autoencoders
• Crank up the volume: preference bias amplification in collaborative recommendation
• Judging a Book by its Description : Analyzing Gender Stereotypes in the Man Bookers Prize Winning Fiction
• InfoFair: Information-Theoretic Intersectional Fairness
• Exploring Stereotypes and Biased Data with the Crowd
• Towards Reducing Bias in Gender Classification
• Maximal adversarial perturbations for obfuscation: Hiding certain attributes while preserving rest
• Learning Hierarchical Visual Representations in Deep Neural Networks Using Hierarchical Linguistic Labels
• Correcting Exposure Bias for Link Recommendation
• Reducing Overlearning through Disentangled Representations by Suppressing Unknown Tasks
• Addressing Fairness, Bias and Class Imbalance in Machine Learning: the FBI-loss
• Towards classification parity across cohorts
• P $\approx$ NP, at least in Visual Question Answering
• Generating Contradictory, Neutral, and Entailing Sentences
• TFW, DamnGina, Juvie, and Hotsie-Totsie: On the Linguistic and Social Aspects of Internet Slang
• Adversarial Examples Generation for Reducing Implicit Gender Bias in Pre-trained Models
• Proceedings of the CSCW 2021 Workshop -- Investigating and Mitigating Biases in Crowdsourced Data