Towards Clear Expectations for Uncertainty Estimation

TL;DR

This paper highlights the need for standardized evaluation protocols for Uncertainty Quantification in Machine Learning, proposing five downstream tasks to better assess the practical utility of UQ methods.

Contribution

It introduces a new perspective by defining five downstream tasks to clarify UQ requirements and questions the effectiveness of current state-of-the-art methods through empirical evaluation.

Findings

No statistical superiority of advanced UQ methods over simple baselines

Current evaluation protocols may not reflect real-world utility

Calls for standardized, relevant metrics for UQ assessment

Abstract

If Uncertainty Quantification (UQ) is crucial to achieve trustworthy Machine Learning (ML), most UQ methods suffer from disparate and inconsistent evaluation protocols. We claim this inconsistency results from the unclear requirements the community expects from UQ. This opinion paper offers a new perspective by specifying those requirements through five downstream tasks where we expect uncertainty scores to have substantial predictive power. We design these downstream tasks carefully to reflect real-life usage of ML models. On an example benchmark of 7 classification datasets, we did not observe statistical superiority of state-of-the-art intrinsic UQ methods against simple baselines. We believe that our findings question the very rationale of why we quantify uncertainty and call for a standardized protocol for UQ evaluation based on metrics proven to be relevant for the ML practitioner.