Decomposing Epistemic Uncertainty for Causal Decision Making

2026-01-30

AI Summary

This work addresses the challenge of causal effect identification in the presence of unobserved confounding, where existing methods struggle to distinguish whether uncertainty arises from finite-sample noise or fundamental non-identifiability. The authors propose a novel framework that constructs a confidence set over the observed data distribution and computes the intersection of causal effect bounds across all compatible structural causal models within this set. This approach uniquely decomposes uncertainty into statistical uncertainty—reducible with more data—and intrinsic non-identifiability—requiring additional variables for resolution. Leveraging neural causal models, the method employs min-max and max-min optimization strategies to derive the tightest possible bounds. Experiments on both synthetic and real-world datasets demonstrate that the framework accurately determines whether collecting more samples is beneficial, thereby guiding decisions on further variable measurement or the adoption of randomized trials.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Causal inference from observational data provides strong evidence for the best action in decision-making without performing expensive randomized trials. The effect of an action is usually not identifiable under unobserved confounding, even with an infinite amount of data. Recent work uses neural networks to obtain practical bounds to such causal effects, which is often an intractable problem. However, these approaches may overfit to the dataset and be overconfident in their causal effect estimates. Moreover, there is currently no systematic approach to disentangle how much of the width of causal effect bounds is due to fundamental non-identifiability versus how much is due to finite-sample limitations. We propose a novel framework to address this problem by considering a confidence set around the empirical observational distribution and obtaining the intersection of causal effect bounds for all distributions in this confidence set. This allows us to distinguish the part of the interval that can be reduced by collecting more samples, which we call sample uncertainty, from the part that can only be reduced by observing more variables, such as latent confounders or instrumental variables, but not with more data, which we call non-ID uncertainty. The upper and lower bounds to this intersection are obtained by solving min-max and max-min problems with neural causal models by searching over all distributions that the dataset might have been sampled from, and all SCMs that entail the corresponding distribution. We demonstrate via extensive experiments on synthetic and real-world datasets that our algorithm can determine when collecting more samples will not help determine the best action. This can guide practitioners to collect more variables or lean towards a randomized study for best action identification.

Problem

• epistemic uncertainty
• causal inference
• unobserved confounding
• non-identifiability
• observational data

Innovation

• epistemic uncertainty decomposition
• causal effect bounds
• non-identifiability
• neural causal models
• sample vs non-ID uncertainty