Amortizing Causal Sensitivity Analysis via Prior Data-Fitted Networks

2026-05-11

AI Summary

Traditional causal sensitivity analysis methods require instance-wise computation, making them inefficient when adapting to changes in datasets, causal queries, sensitivity parameters, or treatment variables. This work proposes the first foundation model for causal sensitivity analysis, leveraging amortized inference through a prior-data-fitted network and in-context learning to dramatically improve generalization efficiency. The approach introduces Lagrangian scalarization to automatically generate boundary training labels, circumventing the need for model-dependent analytical derivations while recovering Pareto frontier solutions. Experimental results demonstrate that, across diverse settings, the proposed method achieves speedups of several orders of magnitude over conventional approaches during test-time inference.

Abstract

Causal sensitivity analysis aims to provide bounds for causal effect estimates in the presence of unobserved confounding. However, existing methods for causal sensitivity analysis are per-instance procedures, meaning that changes to the dataset, causal query, sensitivity level, or treatment require new computation. Here, we instead present an in-context learning approach. Specifically, we propose an amortized approach to causal sensitivity analysis based on prior-data fitted networks. A key challenge is that the sensitivity bounds are not directly available when sampling training data. To address this, we develop a general prior-data construction that is applicable across the class of generalized treatment sensitivity models. Our construction involves a Lagrangian scalarization of the objective to generate training labels for the bounds through a tradeoff between causal effect min/max-imization and sensitivity model violation, which avoids model-specific analytical derivations. We further show that, under standard convexity and linearity conditions, our objective recovers the full Pareto frontier of solutions. Empirically, we demonstrate our amortized approach across various datasets, causal queries, and sensitivity levels, where our approach achieves a test-time computation that is orders of magnitude faster than per-instance methods. To the best of our knowledge, ours is the first foundation model for in-context learning for causal sensitivity analysis.

Problem

• causal sensitivity analysis
• unobserved confounding
• amortized inference
• in-context learning
• computational efficiency

Innovation

• amortized inference
• causal sensitivity analysis
• in-context learning
• prior-data fitted networks
• Pareto frontier