Research map / Probabilistic and scientific modeling / Causal discovery

Temporal Causal Reasoning with (Non-Recursive) Structural Equation Models

AAAI 2025 · Probabilistic and scientific modeling › Causal discovery · approach: Counterfactual identification and bounds · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Reinterprets structural equation models as time-series transformers of exogenous dynamics, introduces the temporal causal logic CPLTL, and shows polynomial-time model checking and a reduction of arbitrary-delay models to 1-step models.

Problem

Actual causality with structural equation models ignores temporal dynamics and usually restricts to recursive (acyclic) models, so causal claims about timing, duration, and feedback loops cannot be expressed.

Approach

Treat an SEM as a mechanism mapping a temporal context (a time series of exogenous values) into a computation (a time series of endogenous values), with time-sensitive interventions and LTL/PLTL operators, so cycles in the dependency graph become temporal feedback rather than technical problems.

Key results

• CPLTL model checking is shown to be solvable in polynomial time (in P), via periodic computation and a Markey-style past-operator lemma.
• Any arbitrary-delay temporal SEM can be simulated by an equivalent 1-step lagged TSEM of size O(|M|^2).
• Introduces rescalable temporal equivalence, e.g. a 1-second-step model is rescalably equivalent to a 1-minute-step model with coefficient k=60.

Limitations

• Only finite, deterministic models are considered; probabilistic and non-deterministic extensions are left open.
• Equivalence notions are defined but a bisimulation characterisation is left as open problem.
• Contemporaneous dependencies are excluded by design, and the framework does not extract temporal information from existing static models.

Experimental setup

Builds on: Halpern-Pearl actual causality LTL PLTL Markey periodic model checking

Closest papers by meaning

• Causal learning with the invariance principle (NeurIPS 2026 · similarity 0.88)
• Learning Structural Causal Models from Ordering: Identifiable Flow Models (AAAI 2025 · similarity 0.87)
• Unveiling Causal Reasoning in Large Language Models: Reality or Mirage? (NeurIPS 2024 · similarity 0.87)
• Causality can systematically address the monsters under the bench(marks) (NeurIPS 2026 · similarity 0.86)
• Exogenous Isomorphism for Counterfactual Identifiability (ICML 2025 · similarity 0.86)
• Systems with Switching Causal Relations: A Meta-Causal Perspective (ICLR 2025 · similarity 0.86)