LIME (Local Interpretable Model-agnostic Explanations)

A tool that explains individual model predictions by approximating the model locally with an interpretable one, like linear regression.