Beyond explainability: Regulating machine learning in practice

Description

Machine learning (ML) is is becoming increasingly prevalent across industries, creating new types and new levels of risk. Managing this risk is quickly becoming the central challenge of major organizations, one that strains data science teams, legal personnel, and the C-suite alike. How organizations approach manage this risk will determine their ability to fully make use of ML and in some cases, their long-term success or failure.

However, the current approach to managing risk in ML is extremely limited and will only get organizations so far. Currently, attempts to address risk in ML are approached through the lens of explainability, with a focus on understanding the “black box” of ML models. But the future of ML risk management is going to be focused on much more than explaining the inner workings of these models.

Andrew Burt shares key lessons from past regulations focused on complex, opaque technology along with a proposal for new ways to effectively manage risk in ML.