New Method Reveals Complex Causation

Summary: A new method developed by researchers allows scientists to identify unique, redundant and synergistic causality, providing a clearer view of what affects complex systems. This method, known as SURD, has implications in fields ranging from climate science to aerospace engineering.

Traditional methods often confuse variables that are not the true cause, but SURD minimizes errors by accurately parsing causality. This tool has the potential to assist in the design of optimized systems by more precisely identifying causal factors.

Researchers demonstrated the utility of SURD by studying turbulence and revealing previously hidden interactions between airflow variables. Their work highlights the benefits of SURD for more accurate causal analysis in complex domains.

Basic Information:

• SURD (Synergistic-Unique-Redundant Decomposition) clarifies causality by isolating unique, redundant and synergistic factors.
• By distinguishing between causal types, the method minimizes false positives and aids precise analysis.
• Tested in multiple scenarios, SURD consistently provided accurate causal predictions where other methods failed.

Source: caltech

Cause and effect. We understand this concept from a young age. Pull the string of the tug toy and the toy follows. Naturally, as the system grows, the number of variables increases and noise comes into play, things get more complicated. Eventually, it can become almost impossible to tell whether a variable causes an effect or is simply associated or related to it.

Consider an example from climate science. Experts who study major atmospheric circulation patterns and their effects on global weather want to know how these systems might change as climates warm.

Many variables come into play here: ocean and air temperatures and pressures, ocean currents and depths, and even the details of the earth’s rotation over time. But which variables cause which measured effects?

This is where information theory comes into play as a framework for formulating causality. Adrián Lozano-Durán, an associate professor of aeronautics and astronautics at Caltech, and members of his group at both Caltech and MIT have developed a method that can be used to determine causality even in such complex systems.

The new mathematical tool can reveal the contributions of each variable in a system to a measured effect, both individually and, more importantly, in combination.

In a paper published today, November 1, in the journal, the team describes their new method, called synergistic-unique-redundant causality decomposition (SURD). Nature Communication.

The new model can be used in any situation where scientists are trying to determine the true cause or causes of a measured effect. This could be anything from what triggered the stock market crash in 2008, to the contribution of various risk factors in heart failure, to how ocean variables affect the population of certain fish species, to what mechanical properties are responsible for a system’s failure. material.

“Causal inference is multidisciplinary and has the potential to advance advances in many fields,” says Álvaro Martínez-Sánchez, lead author of the new paper and a graduate student at MIT in Lozano-Durán’s group.

For Lozano-Durán’s group, SURD will be the most useful tool in the design of aerospace systems. For example, by determining which variable increases an aircraft’s drag, the method can help engineers optimize the vehicle’s design.

“Previous methods will only tell you how much causality comes from a variable,” Lozano-Durán explains.

“What is unique about our method is its ability to capture the full picture of everything that causes an effect.”

The new method also prevents misidentification of causalities. This is largely because it goes beyond measuring the impact produced by each variable independently. In addition to what the authors call “unique causality,” the method also incorporates two new categories of causality, redundant and synergistic causality.

Redundant causality occurs when more than one variable produces a measured effect, but not all variables are needed to achieve the same result. For example, a student may get good grades in class because he is very smart or hard-working. Both can result in a good grade, but only one is necessary. Two variables are redundant.

Synergistic causality involves multiple variables that must work together to create an effect. Not every variable will give the same result on its own. For example, a patient takes medicine A but cannot get rid of his disease.

Likewise, when he takes drug B, he sees no improvement. However, when he takes both medications, he recovers completely. Drugs A and B are synergistic.

SURD mathematically separates the contributions of each variable in a system into unique, redundant, and synergistic causation components.

The sum of all these contributions must satisfy a conservation of information equation that can then be used to reveal the existence of hidden causality, that is, variables that cannot be measured or are thought to be unimportant. (If latent causality turns out to be too large, researchers know they need to re-evaluate the variables they included in their analysis.)

To test the new method, Lozano-Durán’s team analyzed 16 validation cases using SURD; these were scenarios with known solutions that would otherwise pose significant challenges to researchers trying to determine causality.

“Our method will consistently give you a meaningful answer in all of these cases,” says Gonzalo Arranz, a postdoctoral researcher at the Graduate Aerospace Laboratories at Caltech who is also an author of the paper.

“Other methods confuse causalities that should not be confused, and sometimes they are. “For example, they get a false positive that identifies a causality that doesn’t exist.”

In the paper, the team used SURD to study the formation of turbulence during airflow around a wall. In this case, air flows slower at lower altitudes close to the wall and faster at higher altitudes.

Previously, some theories about what was happening in this scenario had suggested that the high-altitude flow affected what was happening near the wall and not the other way around. Other theories suggest the opposite; suggests that airflow near the wall affects what happens at higher altitudes.

“We analyzed the two signals with SURD to understand in which way the interactions occur,” says Lozano-Durán.

“The causality appears to be due to distant velocity. Additionally, there is also a synergy where signals interact to create another type of causality. This decomposition, or dissection of causality, is what is unique about our method.”

News about this mathematical modeling and causality research

Writer: Kim Fesenmaier
Source: caltech
Contact: Kimm Fesenmaier – CalTech
Picture: Image courtesy of Neuroscience News

Original Research: Open access.
“Breaking down causality into synergistic, unique, and redundant components” By: Adrián Lozano-Durán et al. Nature Communication

---

Abstract

Breaking down causality into synergistic, unique, and redundant components

Causation lies at the heart of scientific research and forms the fundamental basis for understanding interactions between variables in physical systems.

Despite its central role, existing methods for causal inference face significant challenges due to nonlinear dependencies, stochastic interactions, self-causation, collider effects, and the effects of exogenous factors, among others.

While existing methods can effectively address some of these challenges, no single approach has been able to successfully integrate all of these aspects.

Here, we address these challenges with SURD: Synergistic-Unique-Redundant Causation Decomposition. SURD measures causality as the increase in redundant, unique, and synergistic information gained about future events from past observations.

The formulation is non-invasive and can be applied to both computational and experimental studies even when samples are small.

We benchmark SURD in scenarios that pose significant challenges to causal inference and show that it offers a more reliable measure of causality than previous methods.