Understanding the ML governance framework
ML governance is complex as it deals with complex internal and regulatory policies. There are many stakeholders and technology systems involved in the full ML life cycle. Furthermore, the opaque nature of many ML models, data dependencies, ML privacy, and the stochastic behaviors of many ML algorithms make ML governance more challenging.
The governance body in an organization is responsible for establishing policies and the ML governance framework. To operationalize ML risk management, many organizations set up three lines of defense for their organizational structure:
- The first line of defense is owned by the business operations. This line of defense focuses on the development and use of ML models. The business operations are responsible for creating and retaining all data and model assumptions, model behavior, and model performance metrics in structured documents based on model classification and risk exposure. Models are tested...
