Interpretability Can Be Actionable

TL;DR

This paper advocates for evaluating interpretability of neural networks based on actionability, emphasizing practical impact through concrete decisions and interventions.

Contribution

It introduces a framework for actionable interpretability, defining evaluation criteria focused on real-world utility and addressing barriers to practical impact.

Findings

Proposes actionability as a key interpretability criterion

Defines two dimensions: concreteness and validation

Identifies five domains for leverage in interpretability

Abstract

Interpretability aims to explain the behavior of deep neural networks. Despite rapid growth, there is mounting concern that much of this work has not translated into practical impact, raising questions about its relevance and utility. This position paper argues that the central missing ingredient is not new methods, but evaluation criteria: interpretability should be evaluated by actionability--the extent to which insights enable concrete decisions and interventions beyond interpretability research itself. We define actionable interpretability along two dimensions--concreteness and validation--and analyze the barriers currently preventing real-world impact. To address these barriers, we identify five domains where interpretability offers unique leverage and present a framework for actionable interpretability with evaluation criteria aligned with practical outcomes. Our goal is not to downplay exploratory research, but to establish actionability as a core objective of interpretability research.