Knowledge MaxxingInterpretability
Interpretability is how directly a person can understand a model's internal logic itself, rather than only receiving a separately generated explanation of its output.
Prerequisite: Explainability
What matters
A highly interpretable model, like a small decision tree, can be understood directly by inspection, while a complex model may need a separate explanation method layered on top instead.
Example
A simple linear model's coefficients can be read directly to understand its logic, making it more interpretable than a large neural network needing external explanation tools.
Common misconception
Interpretability is not simply a nicer word for explainability; interpretability refers to understanding the model itself, while explainability can be added externally to any model.
Recall question
How does an interpretable model differ from one that merely has explanations generated for it?