Understanding what drives change in complex systems — from natural ecosystems to engineered designs — is vital to advancing science and technology, yet such causal chains are often shrouded by countless interconnected factors.
For engineers, climate scientists, and data analysts, pinpointing these links can mean the difference between accurately forecasting trends or missing critical insights entirely.
Now, an innovative algorithm from a team of MIT engineers may help reveal the hidden dynamics within intricate systems. Using methods based on information theory, the algorithm has broad applications, from assessing climate patterns to projecting population growth to optimizing aircraft efficiency.
The algorithm, developed by engineers from MIT’s Department of Aeronautics and Astronautics (AeroAstro), analyzes complex datasets to map causal relationships between variables over time. This ‘causality map’ enables scientists to see how specific factors influence each other and to assess the nuances of those relationships.
For instance, the algorithm could determine whether two variables exhibit a synergistic effect, where they only produce an outcome when paired together, or a redundant effect, where two variables have the same impact.
“The significance of our method lies in its versatility across disciplines,” explains Álvaro Martínez-Sánchez, a graduate student in AeroAstro. “It can be applied to better understand the evolution of species in an ecosystem, the communication of neurons in the brain, and the interplay of climatological variables between regions, to name a few examples.”
The tool has notable implications in climate science, population studies, and engineering, where scientists and designers need a reliable way to analyze the myriad forces driving trends in their fields. Engineers anticipate its potential for enhancing aerospace design by clarifying the relationship between aircraft features and efficiency.
“We hope by embedding causality into models, it will help us better understand the relationship between design variables of an aircraft and how it relates to efficiency,” says Adrián Lozano-Durán, an associate professor in AeroAstro.
MIT postdoctoral researcher Gonzalo Arranz, who collaborated on the study, echoes this sentiment, emphasizing how the algorithm may reveal previously undetectable links between variables with weak intensities that still hold substantial significance.
Applying information theory to causality
The team turned to information theory, a branch of applied mathematics that studies the transmission of data in networks, a concept initially developed by the late MIT professor emeritus Claude Shannon. Using information theory, the team constructed their algorithm to treat variables as nodes within a network, where messages transfer information across the system.
“We treat the system as a network, and variables transfer information to each other in a way that can be measured,” Lozano-Durán explains. “If one variable is sending messages to another, that implies it must have some influence. That’s the idea of using information propagation to measure causality.”
A primary advantage of MIT’s algorithm, dubbed SURD (Synergistic-Unique-Redundant Decomposition), is its ability to evaluate multiple variables in tandem. Unlike traditional causal inference methods, which examine one pair of variables at a time, SURD simultaneously assesses a web of relationships. This approach not only yields a more holistic causality map but also enables the algorithm to identify specific types of interactions that most other methods overlook.
In contrast to methods that rely primarily on the intensity of interactions between variables, SURD can detect subtler relationships that may play a critical role in the system’s behavior.
Gauging unmeasurable influences
One of SURD’s unique features is its ability to estimate ‘causal leakage’, or the influence of factors that have not yet been accounted for in a dataset. This allows researchers to detect potential gaps in their analysis and explore whether additional variables might improve their understanding of a system’s behavior.
“Part of our method detects if there’s something missing,” Lozano-Durán states. “We don’t know what is missing, but we know we need to include more variables to explain what is happening.”
To validate their approach, the team applied SURD to multiple benchmark datasets commonly used in causal inference studies. The algorithm accurately identified known causal links across diverse cases, from predator-prey relationships in marine ecosystems to meteorological interactions involving temperature and pressure across geographic regions.
These tests confirmed SURD’s ability to detect causality in scenarios where other methods might only capture part of the picture.
Beyond enhancing climate science and engineering, the MIT team envisions their algorithm as a powerful resource for many disciplines. “SURD has the potential to drive progress across multiple scientific and engineering fields, such as climate research, neuroscience, economics, epidemiology, social sciences, and fluid dynamics, among others,” says Martínez-Sánchez.
The team has made SURD available for public use, inviting researchers and analysts across disciplines to test the algorithm on their own datasets. The MIT engineers hope that, by fostering collaboration, SURD can contribute to novel insights in scientific and engineering inquiries where causality remains elusive.
Journal Reference:
Martínez-Sánchez, Á., Arranz, G. & Lozano-Durán, A., ‘Decomposing causality into its synergistic, unique, and redundant components’, Nature Communications 15, 9296 (2024). DOI: 10.1038/s41467-024-53373-4
Article Source:
Press Release/Material by Massachusetts Institute of Technology (MIT)
Featured image credit: Michal Jarmoluk | Pixabay