Interpretability

noun

say it: “in-tur-prih-tuh-BIL-ih-tee” /ɪnˌtɝprətəˈbɪləti/

How easily people can understand the reasons behind a model's decisions.

What is interpretability?

Interpretability is how easily a person can see why a model gave its answer. A small decision tree or a linear model with a few coefficients is easy to read; a forest of hundreds of trees isn't, and needs tools such as permutation importance to explain it.

In a sentence

  • “The bank chose a small tree for its interpretability: every refusal could be explained.”
  • “The random forest scored better but gave up interpretability.”

Level: Journeyman · More AI Basics terms at this level

Updated