aiqbench analysis
The paper argues that learned representations, such as variational autoencoders trained to reconstruct health data, can silently break causal estimates even when reconstruction looks accurate. It proves that no covariate-only training objective can certify a representation as sufficient for causal identification. In simulations, a predictive diagnostic caught a rare confounder that reconstruction loss missed, so readers building ML-based causal pipelines should validate for causal sufficiency rather than rely on reconstruction quality.
When Reconstruction Is Not Enough: Doubly Robust Counterfactual and Interventional Mediation Analysis with Learned Multimodal Representations
arXiv:2610.30220v1 Announce Type: new Abstract: Learned representations of high-dimensional multimodal health records are increasingly used to adjust for confounding, yet they are usually trained to reconstruct the data rather than preserve the information that causal identification requires. We study counterfactual outcomes under alternative treatments, including phenotype-specific effects and patient-level counterfactual risk profiles of the kind digital twins provide, and interventional direct and indirect effects under exposure-induced intermediate confounding. We give conditions under which a representation preserves identification and show that interventional mediation is more demanding than counterfactual comparison. We prove that no covariate-only training objective, whether reconstruction, contrastive, or physics-informed, can certify that a representation is sufficient, and that, in a linear structural model, a representation can lose arbitrarily little reconstruction accuracy yet induce arbitrarily large bias. We decompose the error of cross-fitted doubly robust estimators into a first-order representation bias, which neither double robustness nor cross-fitting removes, and a second-order remainder of nuisance-estimation errors, so coverage worsens as the sample size grows when the representation is insufficient. We propose a workflow combining modality-specific variational autoencoders trained outside the estimation sample, cross-fitted doubly robust estimators, locked phenotypes, select-then-confirm pathway analysis, portable representations, and diagnostics for representation insufficiency, exposure leakage, and sensitivity. In simulations, discarding a rare, moderately strong confounder gave indirect-effect intervals with 2% coverage at n = 3,000; reconstruction loss missed the problem, and a simple predictive diagnostic flagged it in every replicate.
Abstract:Learned representations of high-dimensional multimodal health records are increasingly used to adjust for confounding, yet they are usually trained to reconstruct the data rather than preserve the information that causal identification requires. We study counterfactual outcomes under alternative treatments, including phenotype-specific effects and patient-level counterfactual risk profiles of the kind digital twins provide, and interventional direct and indirect effects under exposure-induced intermediate confounding. We give conditions under which a representation preserves identification and show that interventional mediation is more demanding than counterfactual comparison. We prove that no covariate-only training objective, whether reconstruction, contrastive, or physics-informed, can certify that a representation is sufficient, and that, in a linear structural model, a representation can lose arbitrarily little reconstruction accuracy yet induce arbitrarily large bias. We decompose the error of cross-fitted doubly robust estimators into a first-order representation bias, which neither double robustness nor cross-fitting removes, and a second-order remainder of nuisance-estimation errors, so coverage worsens as the sample size grows when the representation is insufficient. We propose a workflow combining modality-specific variational autoencoders trained outside the estimation sample, cross-fitted doubly robust estimators, locked phenotypes, select-then-confirm pathway analysis, portable representations, and diagnostics for representation insufficiency, exposure leakage, and sensitivity. In simulations, discarding a rare, moderately strong confounder gave indirect-effect intervals with 2% coverage at n = 3,000; reconstruction loss missed the problem, and a simple predictive diagnostic flagged it in every replicate.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Methodology (stat.ME); Machine Learning (stat.ML)
arXiv:2610.30220 [cs.LG]
(or arXiv:2610.30220v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.30220 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 331 words
Related coverage
Reconstruct! Don't Encode: Self-Supervised Representation Reconstruction Loss for High-Intelligibility and Low-Latency Streaming Neural Audio Codec
arXiv · Sep 2, 2026
Momentum-Guided Semantic Forecasting (MoFore) for Self-Supervised Video Representation Learning
arXiv · Jun 16, 2026
A Geometric Theory of Decision Boundaries in Structured Markov Decision Processes
arXiv · Sep 17, 2026
Beyond Reconstruction Error: Analytical and Data-Driven Action Tokenization for Autoregressive Vision-Language-Action Models
arXiv · Sep 24, 2026