An orchestra without a conductor plays a complex piece. The listener hears the overall music but wants to understand who leads the melody and who merely echoes it in unison. This is roughly the challenge tackled by the researchers behind the new study: in any complex system, whether climate, brain, or finance, many processes influence each other simultaneously, and telling apart direct connections from collective effects is incredibly difficult.
Until now, scientists had two extreme approaches. The first simply ignored all other processes and looked only at the pair: how strongly one variable predicts another. The second accounted for all other variables at once but lost the distinction between those that truly matter and those that merely create noise. The authors propose a third path: instead of considering all subsets, they look only for those that maximally weaken or maximally strengthen the information transfer between the two processes of interest. The difference between these extreme cases and the ordinary pairwise value reveals how much of the data flow is due to the pure pair and how much to collective effects.
Here is how it works. Take two processes, for example sea surface temperature in the tropical Pacific (the El Niño index) and the atmospheric pressure difference between Tahiti and Darwin (the Southern Oscillation). First, measure how much information about the future pressure is contained in the past temperature, ignoring all other climate indices. This gives the pairwise value. Then, search through all possible subsets of other indices and find the one that makes this information transfer minimal. This means the selected variables duplicated part of the signal, creating redundancy. Next, find the subset that makes the transfer maximal. This shows that some information appears only when several variables are considered together, giving rise to synergy. The difference between the pairwise value and the minimum gives the redundant component, the difference between the maximum and the pairwise value gives the synergistic component, and the minimum itself is taken as the unique contribution attributable to the direct link between the two processes.
On a toy model with four processes, where two sources converge at a common node that then influences the target, the method performed flawlessly. The pairwise transfer from one source to the target was 0.32, the minimum was zero, and the maximum was 1.25. The decomposition yielded: unique contribution zero, redundant 0.32, synergistic 0.93. This matches the model's structure exactly: there is no direct path, but information passes through the common node and through the joint action of the two sources. On real climate data, the result was less dramatic but no less important. The pairwise transfer from El Niño to the Southern Oscillation was 0.078, the minimum 0.074, and the maximum 0.094. The unique contribution was 0.074, redundant 0.004, and synergistic 0.016. In other words, about 20% of the maximum information flow is due to collective effects that are invisible when looking only at the pair.
This method is valuable not only for climatology. It can be applied to any time series where direct connections need to be separated from mediated ones: in neuroscience, to understand how different brain regions interact; in economics, to identify which markets truly influence each other and which merely react to common news; in social networks, to distinguish real influence from herd effects. And unlike many other higher-order approaches, it does not require enormous computational resources because it uses a greedy algorithm: variables are added one by one until adding more no longer significantly changes the result. This makes the method applicable to large networks with hundreds or thousands of nodes.
The orchestra without a conductor continues to play, but now we have a way to read the score. We can not only hear the overall music but also understand which instruments lead the melody, which merely echo it, and which create a harmony that exists in no single part. And this understanding paves the way for more accurate causality analysis in complex systems, where collective effects are not an exotic exception but account for a substantial share of the information flow.
El Niño and the Southern Oscillation are so tightly linked that they are combined into the acronym ENSO, but the proposed method shows that even in such a pair, about a fifth of the information is transmitted not directly but through other climate indices, such as the temperature of the Northern Tropical Atlantic or the Southern Tropical Atlantic.