Observational causality by states and interaction type for scientific discovery

2025/05/16 by Álvaro Martínez-Sánchez, Martínez-Sánchez, Álvaro, Adrián Lozano-Durán +1 · 1 citation
Arts and Humanities · Biochemistry, Genetics and Molecular Biology · Physics and Astronomy · #Data Analysis #FOS: Physical sciences #Fluid Dynamics (physics.flu-dyn) #Gene Regulatory Network Analysis #Model Reduction and Neural Networks #Philosophy and History of Science #Statistics and Probability (physics.data-an)

paper · pdf · doi:10.48550/arxiv.2505.10878

openalex publication_date 2025/05/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Causality plays a central role in understanding interactions between variables in complex systems. These systems often exhibit state-dependent causal relationships, where both the strength and direction of causality vary with the value of the interacting variables. In this work, we introduce a state-aware causal inference method that quantifies causality in terms of information gain about future states. The effectiveness of the proposed approach stems from two key features: its ability to characterize causal influence as a function of system state, and its capacity to distinguish between redundant and synergistic interactions. The method is validated across a range of benchmark cases in which the direction and strength of causality evolve in a prescribed manner with the state of the system. We further demonstrate the applicability of our approach in two real scenarios: the interaction between motions across scales in a turbulent boundary layer, and the Walker circulation phenomenon in tropical Pacific climate dynamics. Our results show that, without accounting for state-dependent causality as well as redundant and synergistic effects, traditional approaches to causal inference may lead to incomplete or misleading conclusions.

Citations

• Causal Representation Learning in Temporal Data via Single-Parent Decoding
• Disentangling high order effects in the transfer entropy
• Finite Scalar Quantization: VQ-VAE Made Simple
• A Database for Reduced-Complexity Modeling of Fluid Flows
• Information-theoretic formulation of dynamical systems: causality, modeling, and control
• High-recall causal discovery for autocorrelated time series with latent confounders
• Neural Discrete Representation Learning
• Information flow and causality as rigorous notions ab initio
• Quantifying synergistic mutual information
• Causal Mechanisms in the Social Sciences
• Nonnegative Decomposition of Multivariate Information
• On Information and Sufficiency

Cited by

• Cause-and-effect approach to turbulence forecasting

Related

• A simple test for causality in complex systems
• Detecting causality from time series in a machine learning framework
• Causal Discovery from Conditionally Stationary Time Series
• Causal inference for process understanding in Earth sciences
• Bridging Prediction and Attribution: Identifying Forward and Backward Causal Influence Ranges Using Assimilative Causal Inference