Understanding the Origins of Bias in Word Embeddings

Abstract

The power of machine learning systems not only promises great technical progress, but risks societal harm. As a recent example, researchers have shown that popular word embedding algorithms exhibit stereotypical biases, such as gender bias. The widespread use of these algorithms in machine learning systems, from automated translation services to curriculum vitae scanners, can amplify stereotypes in important contexts. Although methods have been developed to measure these biases and alter word embeddings to mitigate their biased representations, there is a lack of understanding in how word embedding bias depends on the training data. In this work, we develop a technique for understanding the origins of bias in word embeddings. Given a word embedding trained on a corpus, our method identifies how perturbing the corpus will affect the bias of the resulting embedding. This can be used to trace the origins of word embedding bias back to the original training documents. Using our method, one can investigate trends in the bias of the underlying corpus and identify subsets of documents whose removal would most reduce bias. We demonstrate our techniques on both a New York Times and Wikipedia corpus and find that our influence function-based approximations are very accurate.

Cited by in corpus (18)

• Image Representations Learned With Unsupervised Pre-Training Contain Human-like Biases
• On the Accuracy of Influence Functions for Measuring Group Effects
• FairFil: Contrastive Neural Debiasing Method for Pretrained Text Encoders
• Two Simple Ways to Learn Individual Fairness Metrics from Data
• Man is to Person as Woman is to Location: Measuring Gender Bias in Named Entity Recognition
• Evaluating Gender Bias in Natural Language Inference
• Multi-Stage Influence Function
• Rissanen Data Analysis: Examining Dataset Characteristics via Description Length
• Fairness in Missing Data Imputation
• Generating Interactive Worlds with Text
• FastIF: Scalable Influence Functions for Efficient Model Interpretation and Debugging
• Investigating Robustness and Interpretability of Link Prediction via Adversarial Modifications
• Mitigating Gender Bias in Captioning Systems
• A Source-Criticism Debiasing Method for GloVe Embeddings
• Obstructing Classification via Projection
• An unsupervised framework for tracing textual sources of moral change
• Low Frequency Names Exhibit Bias and Overfitting in Contextualizing Language Models
• Going Beyond T-SNE: Exposing \texttt{whatlies} in Text Embeddings