Modeling and Discovering Direct Causes for Predictive Models

TL;DR

This paper presents a causal modeling framework for identifying direct causes in predictive models, along with algorithms and an independence rule to improve discovery efficiency, supported by theoretical and empirical validation.

Contribution

It introduces a new causal framework and algorithms for discovering direct causes in predictive models, with an innovative independence rule to enhance speed.

Findings

Algorithms are sound and complete under certain assumptions.

The independence rule accelerates cause discovery.

Empirical results validate theoretical claims.

Abstract

We introduce a causal modeling framework that captures the input-output behavior of predictive models (e.g., machine learning models). The framework enables us to identify features that directly cause the predictions, which has broad implications for data collection and model evaluation. We then present sound and complete algorithms for discovering direct causes (from data) under some assumptions. Furthermore, we propose a novel independence rule that can be integrated with the algorithms to accelerate the discovery process, as we demonstrate both theoretically and empirically.