Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations

TLDR

Majority voting and averaging are common ways to resolve annotator disagreements, yet they can overlook systematic biases and the nuanced information captured by individual disagreements in subjective tasks. The study investigates whether multi‑annotator models can better handle disagreements than aggregating labels. A multi‑task approach predicts each annotator’s judgments as separate subtasks while sharing a common learned representation. Across seven binary classification tasks, this method matches or surpasses aggregated‑label baselines and yields uncertainty estimates that correlate better with annotation disagreement, aiding deployment decisions.

Abstract

Abstract Majority voting and averaging are common approaches used to resolve annotator disagreements and derive single ground truth labels from multiple annotations. However, annotators may systematically disagree with one another, often reflecting their individual biases and values, especially in the case of subjective tasks such as detecting affect, aggression, and hate speech. Annotator disagreements may capture important nuances in such tasks that are often ignored while aggregating annotations to a single ground truth. In order to address this, we investigate the efficacy of multi-annotator models. In particular, our multi-task based approach treats predicting each annotators’ judgements as separate subtasks, while sharing a common learned representation of the task. We show that this approach yields same or better performance than aggregating labels in the data prior to training across seven different binary classification tasks. Our approach also provides a way to estimate uncertainty in predictions, which we demonstrate better correlate with annotation disagreements than traditional methods. Being able to model uncertainty is especially useful in deployment scenarios where knowing when not to make a prediction is important.