# Interpreting Black Box Models via Hypothesis Testing

## TL;DR

This paper introduces a hypothesis testing framework for interpreting black box models, providing controlled error rates and scalable methods that identify significant features in complex models like vision and language systems.

## Contribution

It reframes interpretability as a multiple hypothesis testing problem and proposes two methods, one with provable false discovery control and an approximate scalable alternative.

## Key findings

- High power in simulation tests compared to existing methods
- Effective feature identification in vision and language models
- Explanations are intuitive and easy to interpret

## Abstract

In science and medicine, model interpretations may be reported as discoveries of natural phenomena or used to guide patient treatments. In such high-stakes tasks, false discoveries may lead investigators astray. These applications would therefore benefit from control over the finite-sample error rate of interpretations. We reframe black box model interpretability as a multiple hypothesis testing problem. The task is to discover "important" features by testing whether the model prediction is significantly different from what would be expected if the features were replaced with uninformative counterfactuals. We propose two testing methods: one that provably controls the false discovery rate but which is not yet feasible for large-scale applications, and an approximate testing method which can be applied to real-world data sets. In simulation, both tests have high power relative to existing interpretability methods. When applied to state-of-the-art vision and language models, the framework selects features that intuitively explain model predictions. The resulting explanations have the additional advantage that they are themselves easy to interpret.