AI Explainability and Model Risk Management

Dr.Anupam Datta speaks at the QuantUniversity Winter school 2021

This talk will provide an overview of foundational research in AI Explainability and describe how it enables key Model Risk Management tasks. We will cover axiomatic & causal counterfactual methods for explaining AI models (including the use of Shapley and Aumann-Shapley Values). We will then go on to discuss how Model Risk Management tasks, including assessing conceptual soundness, stability, and reliability of AI models can build on this foundation.

1. A. Datta, S. Sen, Y. Zick, Algorithmic Transparency via Quantitative Input Influence, in Proceedings of 37th IEEE Symposium on Security and Privacy, May 2016. [Paper]
2. K. Leino, S. Sen, A. Datta, M. Fredrikson, L. Li, Influence-Directed Explanations for Deep Convolutional Networks, in Proceedings of the IEEE International Test Conference, October 2018. [Preprint]
3. By Philippe Bracke, Anupam Datta, Carsten Jung and Shayak Sen, Bank of England Staff Working Paper No. 816; Machine learning explainability in finance: an application to default risk analysis. August 2019 (Paper)