Trust Me, I Know the Way: Predictive Uncertainty in the Presence of Shortcut Learning

TL;DR

This paper investigates how shortcut learning affects the measurement of epistemic uncertainty in neural networks, proposing a reconciliation of conflicting viewpoints on uncertainty quantification methods.

Contribution

It clarifies the role of shortcuts in epistemic uncertainty and reconciles different perspectives on entropy-based uncertainty decomposition in neural networks.

Findings

Shortcuts significantly influence epistemic uncertainty as disagreement.

Entropy decomposition can reflect model ignorance or disagreement depending on learning context.

Understanding shortcuts helps improve uncertainty estimation in neural networks.

Abstract

The correct way to quantify predictive uncertainty in neural networks remains a topic of active discussion. In particular, it is unclear whether the state-of-the art entropy decomposition leads to a meaningful representation of model, or epistemic, uncertainty (EU) in the light of a debate that pits ignorance against disagreement perspectives. We aim to reconcile the conflicting viewpoints by arguing that both are valid but arise from different learning situations. Notably, we show that the presence of shortcuts is decisive for EU manifesting as disagreement.