Decision-Aligned Evaluation of Uncertainty Quantification
For practitioners and researchers using uncertainty quantification in decision-making, this work identifies flaws in current evaluation protocols and provides a principled, decision-aligned alternative.
The paper introduces decision-alignment, a criterion for evaluating uncertainty quantification metrics based on their alignment with downstream decision utility. It shows that common metrics like negative log-likelihood and expected calibration error are often misaligned, and proposes prior-weighted utility metrics that consistently align with realized decision utility across benchmarks and real-world case studies.
Uncertainty estimates in machine learning are typically evaluated using generic metrics such as the negative log-likelihood and expected calibration error, yet good performance on such metrics does not necessarily imply high utility in downstream decisions. We introduce decision-alignment, a criterion that reveals which evaluation metrics meaningfully align with downstream utilities. Applying this framework, we show that many widely used uncertainty metrics are either misaligned with common decision problems or encode pathological prior beliefs about the downstream task. We then propose prior-weighted utility metrics, a special class of proper scoring rules that provides decision-aligned uncertainty evaluation. Across benchmark experiments and real-world case studies, our metrics consistently align with realized decision utility, while conventional metrics do not. Our results surface flaws in the current UQ evaluation protocol and offer a principled extension of existing metrics toward decision-relevant UQ evaluation.