Explainable Machine Learning in Deployment

TL;DR

This paper investigates how organizations deploy explainable machine learning, revealing that current practices mainly serve internal debugging rather than end-user transparency, and proposes a framework to improve end-user explainability.

Contribution

It provides empirical insights into current explainability practices and introduces a framework to align explainability goals with end-user needs.

Findings

Most deployments are for machine learning engineers, not end users.

Current explainability techniques mainly support internal debugging.

A framework is proposed to improve explainability for end users.

Abstract

Explainable machine learning offers the potential to provide stakeholders with insights into model behavior by using various methods such as feature importance scores, counterfactual explanations, or influential training data. Yet there is little understanding of how organizations use these methods in practice. This study explores how organizations view and use explainability for stakeholder consumption. We find that, currently, the majority of deployments are not for end users affected by the model but rather for machine learning engineers, who use explainability to debug the model itself. There is thus a gap between explainability in practice and the goal of transparency, since explanations primarily serve internal stakeholders rather than external ones. Our study synthesizes the limitations of current explainability techniques that hamper their use for end users. To facilitate end user interaction, we develop a framework for establishing clear goals for explainability. We end by discussing concerns raised regarding explainability.