Characterizing Sources of Uncertainty to Proxy Calibration and Disambiguate Annotator and Data Bias

TL;DR

This paper demonstrates that quantifying epistemic and aleatoric uncertainty in models enhances interpretability, reveals annotator disagreement, identifies biased data, and improves calibration and performance in complex tasks like emotion recognition.

Contribution

The work introduces a simple modification of Monte Carlo dropout to measure uncertainties, linking them to human disagreement and data bias, advancing interpretability and fairness.

Findings

Aleatoric uncertainty correlates with human disagreement (r≈0.3).

Uncertainty measures can identify difficult and subjective samples.

Total uncertainty serves as a surrogate for model calibration.

Abstract

Supporting model interpretability for complex phenomena where annotators can legitimately disagree, such as emotion recognition, is a challenging machine learning task. In this work, we show that explicitly quantifying the uncertainty in such settings has interpretability benefits. We use a simple modification of a classical network inference using Monte Carlo dropout to give measures of epistemic and aleatoric uncertainty. We identify a significant correlation between aleatoric uncertainty and human annotator disagreement ($r\approx.3$). Additionally, we demonstrate how difficult and subjective training samples can be identified using aleatoric uncertainty and how epistemic uncertainty can reveal data bias that could result in unfair predictions. We identify the total uncertainty as a suitable surrogate for model calibration, i.e. the degree we can trust model's predicted confidence. In addition to explainability benefits, we observe modest performance boosts from incorporating model uncertainty.