When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks

TL;DR

This paper introduces a model for understanding annotator disagreement in subjective tasks like hate speech detection, emphasizing the importance of demographic context and opinion modeling to improve prediction accuracy and interpretability.

Contribution

It presents a novel approach to predict individual annotator ratings and target group opinions without linking to personal identifiers, enhancing privacy and accuracy.

Findings

Model improves annotator rating prediction by 22%.

Model increases variance prediction accuracy by 33%.

Using demographic info and online opinions suffices for accurate predictions.

Abstract

Though majority vote among annotators is typically used for ground truth labels in natural language processing, annotator disagreement in tasks such as hate speech detection may reflect differences in opinion across groups, not noise. Thus, a crucial problem in hate speech detection is determining whether a statement is offensive to the demographic group that it targets, when that group may constitute a small fraction of the annotator pool. We construct a model that predicts individual annotator ratings on potentially offensive text and combines this information with the predicted target group of the text to model the opinions of target group members. We show gains across a range of metrics, including raising performance over the baseline by 22% at predicting individual annotators' ratings and by 33% at predicting variance among annotators, which provides a metric for model uncertainty downstream. We find that annotator ratings can be predicted using their demographic information and opinions on online content, without the need to track identifying annotator IDs that link each annotator to their ratings. We also find that use of non-invasive survey questions on annotators' online experiences helps to maximize privacy and minimize unnecessary collection of demographic information when predicting annotators' opinions.