Measuring Bias in Contextualized Word Representations
This work addresses bias measurement in NLP models, which is crucial for fairness and ethics in AI applications, though it is incremental as it builds on existing bias quantification methods.
The paper tackled the problem of measuring social biases in contextualized word embeddings like BERT, proposing a template-based method that yields more consistent results than traditional cosine-based approaches and demonstrating its application in a gender bias case study for pronoun resolution.
Contextual word embeddings such as BERT have achieved state of the art performance in numerous NLP tasks. Since they are optimized to capture the statistical properties of training data, they tend to pick up on and amplify social stereotypes present in the data as well. In this study, we (1)~propose a template-based method to quantify bias in BERT; (2)~show that this method obtains more consistent results in capturing social biases than the traditional cosine based method; and (3)~conduct a case study, evaluating gender bias in a downstream task of Gender Pronoun Resolution. Although our case study focuses on gender bias, the proposed technique is generalizable to unveiling other biases, including in multiclass settings, such as racial and religious biases.