Assessing the Local Interpretability of Machine Learning Models

TL;DR

This study evaluates human understanding of machine learning models' interpretability through user tasks, confirming that simpler models like decision trees are more interpretable than neural networks, using a new interpretability metric.

Contribution

The paper introduces a user study testing interpretability definitions and proposes runtime operation count as a metric for local interpretability of models.

Findings

Participants perform better on simpler models

Operation count correlates with interpretability

Decision trees and logistic regression are more interpretable than neural networks

Abstract

The increasing adoption of machine learning tools has led to calls for accountability via model interpretability. But what does it mean for a machine learning model to be interpretable by humans, and how can this be assessed? We focus on two definitions of interpretability that have been introduced in the machine learning literature: simulatability (a user's ability to run a model on a given input) and "what if" local explainability (a user's ability to correctly determine a model's prediction under local changes to the input, given knowledge of the model's original prediction). Through a user study with 1,000 participants, we test whether humans perform well on tasks that mimic the definitions of simulatability and "what if" local explainability on models that are typically considered locally interpretable. To track the relative interpretability of models, we employ a simple metric, the runtime operation count on the simulatability task. We find evidence that as the number of operations increases, participant accuracy on the local interpretability tasks decreases. In addition, this evidence is consistent with the common intuition that decision trees and logistic regression models are interpretable and are more interpretable than neural networks.