When AI Generates Covariates: Causal Typing and Estimand Drift in Sequential Experiments
Abstract
AI-generated covariates from notes, conversations, images, and wearable streams can change the causal question when their roles are left unspecified. A generated feature may represent a treatment version, pre-action state, history, design variable, mediator, outcome proxy, observation process, or intercurrent event; these roles are not interchangeable. We formulate a causal type discipline for sequential experiments: a versioned representation map, a causal role classifier, a claim-status filter, and an estimand lock. The lock fixes a standardized proximal effect before generated covariates enter the analysis. Under audit correctness and standard identification assumptions, admissible role assignments preserve this estimand. We apply the established conditional-covariance characterization of compression bias to substitution of generated representations for design-relevant states. A standardized decomposition separates compression, conditional-law, and standardization drift. Further results cover mediator adjustment, post-action leakage, marker-intervention conflation, outcome-guided discovery, and state-measurement error. Cluster-level orthogonal estimators distinguish empirical and superpopulation targets under repeated sessions and missing outcomes. Simulations show that refinement helps when it retains design-relevant information, whereas design erasure, leakage, and same-data marker selection can produce bias or undercoverage. The framework places causal semantics and claim status before confirmatory inference with generated representations.
Related
- Estimand-aware and donor-aware triangulation of predefined gene-set signals across human tendon transcriptomic datasets
- Built-in Selection Bias in Proportional Hazards Models with Omitted Covariates: Simulation Evidence and Alternative Approaches
- Hebbian learning generates individuality in spatial decision-making
- Stabilizing selection generates selection against introgressed DNA
- Evidence-Calibrated Reanalysis of an Otosclerosis Locus Generates a Testable Myeloid State Hypothesis
BibTeX
@misc{fujita2026when,
title = {{When AI Generates Covariates: Causal Typing and Estimand Drift in Sequential Experiments}},
author = {Takes Fujita and Nobutaka Hattori},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.17772},
eprint = {2609.17772},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.17772},
}
RIS
TY - UNPB
TI - When AI Generates Covariates: Causal Typing and Estimand Drift in Sequential Experiments
AU - Takes Fujita
AU - Nobutaka Hattori
PY - 2026
DA - 2026/09/15
PB - arXiv
DO - 10.48550/arxiv.2609.17772
UR - https://arxiv.org/abs/2609.17772
AB - AI-generated covariates from notes, conversations, images, and wearable streams can change the causal question when their roles are left unspecified. A generated feature may represent a treatment version, pre-action state, history, design variable, mediator, outcome proxy, observation process, or intercurrent event; these roles are not interchangeable. We formulate a causal type discipline for sequential experiments: a versioned representation map, a causal role classifier, a claim-status filter, and an estimand lock. The lock fixes a standardized proximal effect before generated covariates enter the analysis. Under audit correctness and standard identification assumptions, admissible role assignments preserve this estimand. We apply the established conditional-covariance characterization of compression bias to substitution of generated representations for design-relevant states. A standardized decomposition separates compression, conditional-law, and standardization drift. Further results cover mediator adjustment, post-action leakage, marker-intervention conflation, outcome-guided discovery, and state-measurement error. Cluster-level orthogonal estimators distinguish empirical and superpopulation targets under repeated sessions and missing outcomes. Simulations show that refinement helps when it retains design-relevant information, whereas design erasure, leakage, and same-data marker selection can produce bias or undercoverage. The framework places causal semantics and claim status before confirmatory inference with generated representations.
ER -