Causal Preference Elicitation

TL;DR

This paper introduces a Bayesian framework for causal discovery that actively involves experts to improve the accuracy and efficiency of identifying causal structures, especially under limited query budgets.

Contribution

It presents a novel causal preference elicitation method that models expert judgments and uses information gain to efficiently query and recover causal graphs.

Findings

Faster posterior concentration on causal graphs

Improved recovery of directed effects

Effective with limited expert queries

Abstract

We propose causal preference elicitation, a Bayesian framework for expert-in-the-loop causal discovery that actively queries local edge relations to concentrate a posterior over directed acyclic graphs (DAGs). From any black-box observational posterior, we model noisy expert judgments with a three-way likelihood over edge existence and direction. Posterior inference uses a flexible particle approximation, and queries are selected by an efficient expected information gain criterion on the expert's categorical response. Experiments on synthetic graphs, protein signaling data, and a human gene perturbation benchmark show faster posterior concentration and improved recovery of directed effects under tight query budgets.