Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions

2026-05-10

AI Summary

This study addresses the problem of causal direction misidentification in time series—particularly sign reversal of causal effects—induced by confounding factors. The authors propose treating a physics-based simulator as a mechanistic implementation of the do-operator, generating interventional distributions through variable clamping and leveraging conditional flow matching to learn nonlinear intervention-conditioned distributions for identifying structural vector autoregressive (SVAR) models. They establish, for the first time, the equivalence between simulator-based clamping and causal interventions, provide a theoretical identifiability framework, and introduce an error decomposition analysis that reveals how insufficient simulation fidelity can induce causal sign reversal. The method accurately recovers causal signs across four scientific benchmarks and achieves an unbiased causal estimate with R² = 0.983 in an ultrafast laser physics case using a high-fidelity quantum solver, demonstrating its empirical validity.

Abstract

We propose SVAR-FM (Structural VAR with Flow Matching), a framework for time series causal discovery that treats a physics-based simulator as a mechanical realization of Pearl's do operator. Clamping a variable inside the simulator physically severs confounding paths, producing interventional data by construction. Conditional Flow Matching then learns the nonlinear interventional conditionals. Theoretically, we prove that the full structural VAR becomes identifiable under a coverage condition on the simulator-clampable variables, and derive an end-to-end error bound that decomposes into Monte Carlo, simulator fidelity, and Flow Matching terms. A sign-flip corollary predicts that when simulator accuracy falls below a threshold, the estimated causal effect reverses sign. Empirically, a benchmark across four scientific domains confirms that SVAR-FM recovers the correct causal sign where observational methods produce sign-reversed estimates due to confounding. A case study in ultrafast laser physics verifies the sign-flip prediction by physically varying the accuracy level of a first-principles quantum solver: the low-accuracy setting reverses the causal sign, while the high-accuracy setting recovers the correct direction (R-squared = 0.983, zero bias).

Problem

• causal discovery
• time series
• intervention
• confounding
• causal identifiability

Innovation

• interventional causal discovery
• structural VAR
• flow matching
• simulator as do-operator
• sign-flip phenomenon