Semantic Properties of cosine based bias scores for word embeddings
This work addresses a gap in bias detection methods for language models, providing insights for researchers, but it is incremental as it focuses on analyzing existing scores rather than introducing new ones.
The paper tackles the lack of comparative analysis for cosine-based bias scores in word embeddings by proposing geometric requirements for meaningful bias quantification and formally analyzing existing scores, showing that their limitations impact real-world applications.
Plenty of works have brought social biases in language models to attention and proposed methods to detect such biases. As a result, the literature contains a great deal of different bias tests and scores, each introduced with the premise to uncover yet more biases that other scores fail to detect. What severely lacks in the literature, however, are comparative studies that analyse such bias scores and help researchers to understand the benefits or limitations of the existing methods. In this work, we aim to close this gap for cosine based bias scores. By building on a geometric definition of bias, we propose requirements for bias scores to be considered meaningful for quantifying biases. Furthermore, we formally analyze cosine based scores from the literature with regard to these requirements. We underline these findings with experiments to show that the bias scores' limitations have an impact in the application case.