Interpretable machine learning for healthcare (Part 1)

Interpretable machine learning for healthcare (Part 1)

This will be a multi part series about explaining AI models predictions’ in healthcare.

Part 1 is a short introduction to the field of interpretable machine learning.

Chapter 1: Explainable Machine Learning is essential but not sufficiant for adoption of AI in healthcare.

yup, that’s right. Explain, make sure they understand before they use the model.

Chapter 2: The illusion of Interpretability Vs Accuracy tradeoff

AI models are ofter split into two categories:

• Intrinsincly interpretable models: These models are self explaining. For exemple, you can explain predictions of a a decision tree by looking at the tree structure.

• Non interpretable models. These models are not able to be explained by the model itself

There is a common beliference in the field of AI that non interpretable models are needed for better accuracy, and that using interpretable models will stand in the way of performing better. There is no scientific base to support this claim.

Chapter 2: