New Framework Detects Causality in Complex Systems Without Experiments

A research team led by Professor Wei Lin of Fudan University, Professor Luonan Chen of the Chinese Academy of Sciences, and Professor Kazuyuki Aihara of the University of Tokyo has introduced a theoretical framework designed to identify genuine cause-and-effect relationships in complex systems without performing any physical interventions. Reporting in the journal The Innovation, the researchers presented a method called interventional dynamical causality, or IntDC, which allows scientists to distinguish true causal links from mere correlations using only observational time-series data. This approach addresses a fundamental limitation in fields such as climatology, neuroscience, and economics, where deliberate experimentation is often prohibitively expensive, technically infeasible, or simply impossible due to the nature of the systems involved.

The core challenge in these disciplines is that traditional statistical tools often confuse correlation with causation. When a third, unseen factor drives two variables simultaneously, standard methods may incorrectly report that one variable causes the other. The new framework distinguishes between two levels of causality: constructive dynamical causality, which asks if the dynamics of one variable depend on another, and interventional dynamical causality, which applies the stricter standard of modern causal inference. Under this stricter standard, a variable X is considered a cause of Y only if an external intervention on X would result in a change to Y. The IntDC framework is specifically designed to meet this higher standard without requiring the actual performance of such interventions.

To achieve this, the team developed a computational criterion known as interventional embedding entropy, or IEE. This method is built upon the concept of delay embedding, a classic idea from dynamical systems theory which states that the full state of a complex system can be reconstructed from the past behavior of just one of its measurements. The key insight behind IEE is that if variable X exerts an intervention-level influence on variable Y, the observational record of Y already contains the imprint of X’s perturbations. By reconstructing the system’s state space from these delay embeddings, IEE quantifies how much intervened causal information flows between variables, allowing researchers to recover causal structures that are free from confounding effects.

The method was first validated through benchmark simulations involving classic logistic dynamical systems with a common driver. In these scenarios, many traditional approaches produced spurious links between variables, incorrectly suggesting a causal relationship where none existed. In contrast, IEE correctly rejected these false positives. The researchers also tested the method’s robustness across various conditions, including different data lengths, noise levels, embedding dimensions, and parameter choices. The framework proved effective across logistic dynamics, coupled Hénon maps, and chaotic neural networks, demonstrating its reliability in nonlinear and nonseparable systems.

Beyond simulations, the team applied the IEE method to real-world data from the life sciences. Using calcium-imaging recordings from the nematode Caenorhabditis elegans, the researchers reconstructed functional neural connectomes and quantified the causal influence between different brain regions. This application demonstrated the method’s ability to map complex biological networks without manipulating the organisms. The framework also proved useful in epidemiology, where it was applied to data from Japan’s COVID-19 outbreak. The analysis identified transmission relationships between regions, offering insights into how the virus spread through the population based solely on observed data.

In the field of circadian biology, the method was used to rebuild the regulatory network surrounding key clock genes. The analysis pinpointed critical regulators within the network, highlighting the potential of the framework to elucidate intricate biological mechanisms. These applications span diverse non-intervention complex systems, showcasing the versatility of the IntDC framework. The method offers three practical advantages: it measures intervention-level causality in systems that cannot be manipulated, it quantifies causal strength in a comparable and rankable way, and it supports the reconstruction of causal regulatory networks directly from data.

The authors note that as large-scale data and artificial intelligence continue to reshape how science handles complex systems, the combination of dynamical causal theory with machine learning will likely further sharpen data-driven causal discovery. This development turns observation alone into a window on cause and effect, providing a powerful tool for researchers who cannot conduct traditional experiments. The study, titled “Deciphering interventional dynamical causality from non-intervention complex systems,” was published in The Innovation in October 2026. The authors declare no competing interests, and the work represents a significant step forward in the field of computational causal inference.

Subject of Research: Scientists can now uncover true cause and effect in complex systems — without touching them

Article Title: Scientists can now uncover true cause and effect in complex systems — without touching them

Article References: Scientists can now uncover true cause and effect in complex systems — without touching them. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: Scientists, uncover, true, cause, effect, complex, systems, without, touching, them, scientific research