Gender Bias in Neural Natural Language Processing

Abstract

We examine whether neural natural language processing (NLP) systems reflect historical biases in training data. We define a general benchmark to quantify gender bias in a variety of neural NLP tasks. Our empirical evaluation with state-of-the-art neural coreference resolution and textbook RNN-based language models trained on benchmark datasets finds significant gender bias in how models view occupations. We then mitigate bias with CDA: a generic methodology for corpus augmentation via causal interventions that breaks associations between gendered and gender-neutral words. We empirically show that CDA effectively decreases gender bias while preserving accuracy. We also explore the space of mitigation strategies with CDA, a prior approach to word embedding debiasing (WED), and their compositions. We show that CDA outperforms WED, drastically so when word embeddings are trained. For pre-trained embeddings, the two methods can be effectively composed. We also find that as training proceeds on the original data set with gradient descent the gender bias grows as the loss reduces, indicating that the optimization encourages bias; CDA mitigates this behavior.

References in corpus (7)

• Efficient Estimation of Word Representations in Vector Space
• Semantics derived automatically from language corpora contain human-like biases
• Recurrent Neural Network Regularization
• Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
• Exploring the Limits of Language Modeling
• Pointer Sentinel Mixture Models
• Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods

Cited by in corpus (16)

• Learning the Difference that Makes a Difference with Counterfactually-Augmented Data
• Unmasking Contextual Stereotypes: Measuring and Mitigating BERT's Gender Bias
• Mitigating Gender Bias in Natural Language Processing: Literature Review
• Explaining Deep Learning Models using Causal Inference
• Directional Bias Amplification
• Does Data Augmentation Improve Generalization in NLP?
• The Woman Worked as a Babysitter: On Biases in Language Generation
• Towards Controllable Biases in Language Generation
• Reducing Gender Bias in Word-Level Language Models with a Gender-Equalizing Loss Function
• Explaining The Efficacy of Counterfactually Augmented Data
• Do Fine-tuned Commonsense Language Models Really Generalize?
• Differentially Private Representation for NLP: Formal Guarantee and An Empirical Study on Privacy and Fairness
• Gender in Danger? Evaluating Speech Translation Technology on the MuST-SHE Corpus
• Investigating Societal Biases in a Poetry Composition System
• Event-Related Bias Removal for Real-time Disaster Events
• Counterfactual Maximum Likelihood Estimation for Training Deep Networks