Necessity and Sufficiency for Explaining Text Classifiers: A Case Study in Hate Speech Detection
This work addresses the need for more informative explanations in text classification, particularly for detecting and mitigating bias in hate speech detection systems, though it is incremental as it builds on existing feature attribution methods.
The authors tackled the problem of explaining text classifiers by introducing a feature attribution method that provides necessity and sufficiency scores for tokens, applied to hate speech detection, and showed that these scores expose classifier biases against marginalized groups by correlating with false positive errors.
We present a novel feature attribution method for explaining text classifiers, and analyze it in the context of hate speech detection. Although feature attribution models usually provide a single importance score for each token, we instead provide two complementary and theoretically-grounded scores -- necessity and sufficiency -- resulting in more informative explanations. We propose a transparent method that calculates these values by generating explicit perturbations of the input text, allowing the importance scores themselves to be explainable. We employ our method to explain the predictions of different hate speech detection models on the same set of curated examples from a test suite, and show that different values of necessity and sufficiency for identity terms correspond to different kinds of false positive errors, exposing sources of classifier bias against marginalized groups.