Goal-Oriented Sequential Bayesian Experimental Design for Causal Learning

2025-07-09

AI Summary

This work addresses the inefficiency and overreliance on full-model inference in causal experimental design. We propose GO-CBED, a goal-oriented sequential Bayesian experimental design framework that directly maximizes the expected information gain (EIG) about user-specified causal quantities—such as a particular interventional effect—rather than inferring the complete causal structure. Our key contributions are threefold: (i) the first integration of non-myopic sequential optimization with goal-directed causal queries; (ii) a differentiable variational lower bound estimator for end-to-end EIG approximation; and (iii) joint optimization of a Transformer-based policy network and a normalizing-flow-based variational posterior, enabling both real-time decision-making and high-fidelity information evaluation. Empirical evaluation on synthetic and semi-synthetic gene regulatory networks demonstrates that GO-CBED significantly outperforms existing baselines under limited experimental budgets and complex underlying mechanisms.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• Maximizes information gain on user-specified causal quantities
• Optimizes non-myopic sequential interventions for targeted learning
• Addresses intractable EIG computation via variational estimation

Innovation

• Goal-oriented Bayesian framework for causal learning
• Variational lower bound estimator for EIG computation
• Transformer-based policy network for real-time decisions