Towards Procedural Fairness: Uncovering Biases in How a Toxic Language Classifier Uses Sentiment Information

TL;DR

This paper investigates how a toxic language classifier uses sentiment and identity terms, revealing biases and guiding future debiasing efforts to improve fairness in toxic language detection.

Contribution

It introduces a concept-based explanation framework to analyze the interaction between sentiment and identity features in toxic language classifiers, highlighting biases.

Findings

Sentiment information is sometimes overshadowed by identity term influence.

The classifier's sensitivity to sentiment varies across classes.

Results inform debiasing strategies for fairer toxic language models.

Abstract

Previous works on the fairness of toxic language classifiers compare the output of models with different identity terms as input features but do not consider the impact of other important concepts present in the context. Here, besides identity terms, we take into account high-level latent features learned by the classifier and investigate the interaction between these features and identity terms. For a multi-class toxic language classifier, we leverage a concept-based explanation framework to calculate the sensitivity of the model to the concept of sentiment, which has been used before as a salient feature for toxic language detection. Our results show that although for some classes, the classifier has learned the sentiment information as expected, this information is outweighed by the influence of identity terms as input features. This work is a step towards evaluating procedural fairness, where unfair processes lead to unfair outcomes. The produced knowledge can guide debiasing techniques to ensure that important concepts besides identity terms are well-represented in training datasets.