Explainable/Interpretable AI

In this episode, we will continue with part two of a four-part series looking at Responsible AI (Listen to part one: Alternative Data in Risk Modeling).

One of the major challenges with effectively developing, deploying, and managing AI systems are often related to the "black box" nature of the model. Specifically, the complexity and non-linear nature of variables in some black-box AI models may be difficult to explain or understand. This includes explainability of the model logic as well as the individual decisions made by the model. In addition, the relative lack of transparency challenges model development and model validation teams to foresee unintended consequences from model usage, which could create an operational risk if the model is implemented in production.

Speakers

Iain Brown, Ph.D., Head of Data Science, SAS UK&I

Matthew Jones, Ph.D., Head of Retail Decision Modelling, Risk Community, Nationwide Building Society

Moderator:
Lisa Ponti, Ph.D., Vice President, Educational Outreach, Global Association of Risk Professionals (GARP)

-----------------------

Over the years, GARP and SAS have worked together to bring risk practitioners unique insights on a variety of topics related to financial risk and have partnered on this episode of our COVID podcast series.

About SAS

As a leader in analytics, SAS has more than 40 years of experience helping organizations solve their toughest problems. Our unrelenting commitment to innovation enables banks to modernize and sustain a competitive edge. SAS provides an integrated, enterprise-wide risk-management platform for managing risk in an organization, from strategic to reputational, operational, financial or compliance-related risk management. Learn more about how SAS is driving innovation and business value for risk and finance professionals at www.sas.com/risk.