Responsible model deployment via model-agnostic uncertainty learning.

Reliably predicting potential failure risks of machine learning (ML) systems when deployed with production data is a crucial aspect of trustworthy AI. This paper introduces the , a novel post-hoc for estimating failure risks and predictive uncertainties of black-box classification model. In addition to providing a , the decomposes the uncertainty estimates into aleatoric and epistemic uncertainty components, thus giving informative insights into the sources of uncertainty inducing the failures. Consequently, can distinguish between failures caused by data variability, data shifts and model limitations and provide useful guidance on appropriate risk mitigation actions (e.g., collecting more data to counter data shift). Extensive experiments on various families of black-box classification models and on real-world and synthetic datasets covering common ML failure scenarios show that the reliably predicts deployment-time failure risks in all the scenarios, and outperforms strong baselines.