655: AI ROI: How to get a profitable return on an AI-project investment — with Keith McCormick

Transparency

Model transparency is a crucial aspect of AI development, balancing complexity and interpretability. Keith McCormick emphasizes that transparency extends beyond mathematical clarity to include the entire development process, enabling clients to be informed partners 1. He discusses the use of tools like LIME and SHAP to provide post-hoc explanations of model behavior, though he acknowledges the limitations of these methods 2. McCormick also references Cynthia Rudin, who argues against the necessity of complex models for accuracy, suggesting simpler models can be equally effective 3.

It's not just the transparency into the math, it's the transparency into the process so that the client can be a full, educated partner in the development of the solution.

--- Keith McCormick

This perspective highlights the importance of clear communication and understanding in AI projects.

   

Healthcare

In healthcare, explainable AI (XAI) plays a pivotal role in predicting patient outcomes, such as 30-day readmissions. Keith McCormick illustrates how models can predict the likelihood of a patient returning with the same diagnosis, which aids in preventive measures 4. He notes that while some may view this as simplistic, integrating risk scores with models can enhance decision-making, as seen in insurance fraud detection projects 5.

If I have a clear mission statement, I can build the model.

--- Keith McCormick

This approach underscores the necessity of clear objectives and the potential of XAI to provide actionable insights in healthcare.