Evaluating Debiasing Techniques for Intersectional Biases
This addresses fairness issues in NLP for marginalized groups by moving beyond single-attribute debiasing, though it is incremental in extending existing techniques to intersectional contexts.
The paper tackles the problem of evaluating debiasing techniques for intersectional biases in NLP models, where existing methods often focus on binary attributes in isolation, and finds that new approaches like bias-constrained models and extended iterative nullspace projection can handle multiple protected attributes.
Bias is pervasive in NLP models, motivating the development of automatic debiasing techniques. Evaluation of NLP debiasing methods has largely been limited to binary attributes in isolation, e.g., debiasing with respect to binary gender or race, however many corpora involve multiple such attributes, possibly with higher cardinality. In this paper we argue that a truly fair model must consider `gerrymandering' groups which comprise not only single attributes, but also intersectional groups. We evaluate a form of bias-constrained model which is new to NLP, as well an extension of the iterative nullspace projection technique which can handle multiple protected attributes.