Goal-Oriented Sequential Bayesian Experimental Design for Causal Learning

ICLR 2026 Conference Submission14524 Authors

Keywords: Causal Reasoning, Causal Discovery; Bayesian Optimal Experimental Design; Goal-Oriented Design ; Amortized Variational Inference; Expected Information Gain; Mutual Information Lower Bounds; Adaptive Experiments

TL;DR: A goal-oriented, non-myopic Bayesian framework for sequential causal experimental design that targets specific queries rather than full models. It outperforms existing methods across various causal tasks, especially with complex causal mechanisms.

Abstract: We present GO-CBED, a goal-oriented Bayesian framework for sequential causal experimental design. Unlike conventional approaches that select interventions aimed at inferring the full causal model, GO-CBED directly maximizes the expected information gain (EIG) on user-specified causal quantities of interest, enabling more targeted and efficient experimentation. The framework is both non-myopic, optimizing over entire intervention sequences, and goal-oriented, targeting only model aspects relevant to the causal query. To address the intractability of exact EIG computation, we introduce a variational lower bound estimator, optimized jointly through a transformer-based policy network and normalizing flow-based variational posteriors. The resulting policy enables real-time decision-making via an amortized network. We demonstrate that GO-CBED consistently outperforms existing baselines across various causal reasoning and discovery tasks---including synthetic structural causal models and semi-synthetic gene regulatory networks---particularly in settings with limited experimental budgets and complex causal mechanisms. Our results highlight the benefits of aligning experimental design objectives with specific research goals and of forward-looking sequential planning.

Primary Area: causal reasoning

Submission Number: 14524