In the quantum world, correlation can imply causation.
A new experiment using particles of light shows that identifying a simple association between two variables is sufficient to determine whether one influences the other. This process to determine causality, described March 23 in Nature Physics, is surprisingly simple. In ordinary life, a correspondence between how two variables rise and fall — ice cream sales and number of drownings, for example — isn’t enough to conclude that one variable triggers the other. “Somehow this question of inferring a causal explanation from correlation is very different in the case of quantum mechanics,” says Ognyan Oreshkov, a quantum physicist at the Université libre de Bruxelles in Belgium.
The researchers observed the polarization of photons and used the association between measurements to determine whether the outcome of one measurement caused the other or both measurements had a common cause. The results don’t mean that, for instance, scientists will eventually plug clinical trial data into quantum computers to determine whether a drug effectively treats a disease. But Oreshkov says that the study is a first step toward exploring how causality emerges from quantum mechanics and manifests itself in the macroscopic world.
The murky relationship between correlation and causation arises in drug trials, genetic analyses and economics, among other fields. Two variables are correlated if learning the value of one provides insight into the value of another. For example, a recent survey revealed that women who watch cooking shows and cook at home are more likely to have a higher body mass index. Yet while it’s tempting to conclude that butter-laden Paula Deen recipes are making people fat, correlation does not imply causation. It’s possible that overweight people like to watch cooking shows. Or maybe a third variable is responsible for both watching cooking shows and high body mass index. A follow-up study, such as a randomized, controlled trial, is needed to infer cause and effect.
Correlation also plays a large role in quantum physics. Properties of a particle such as spin and polarization are a mystery until they are measured, but a phenomenon called quantum entanglement helps reduce some of that uncertainty. When two particles are entangled, measuring the value of one particle’s property informs the value of that same property for the particle’s entangled partner (SN Online: 3/16/15). In other words, entangled particles have correlated properties. Quantum physicists Katja Ried and Robert Spekkens at the Perimeter Institute for Theoretical Physics in Waterloo, Canada, and colleagues set up an experiment to explore whether quantum correlations alone could reveal underlying causation.
The researchers generated a pair of entangled photons and sent them through a circuit. A detector determined a polarization property (for instance, horizontal versus vertical) of one of the photons. A second detector then received either the same photon or the photon’s entangled partner and made another polarization measurement. The process was automated so that the researchers never knew which photon the second detector was measuring.
The researchers demonstrated that they could sift through the detector measurement data and find correlations that directly proved causality. They showed that matching measurements for every polarization property could be explained only by the second detector measuring the same photon as the first detector. In that case, there is a direct causal relationship between the measurements: Obtaining the first measurement ensures that the second measurement delivers the same result. Opposing measurements (such as one vertical, one horizontal) meant that the second detector must have measured the entangled partner. These measurements didn’t influence each other, but they had a common cause: the generation of the two entangled photons.
By analyzing the correlations between the two measurements, Ried and colleagues correctly determined how much of the correlation was due to a direct causal relationship versus a common cause for both measurements. It was the quantum equivalent of analyzing the results of the cooking show survey and determining whether the shows actually cause people to gain weight. “That’s a task that’s impossible with classical variables,” Ried says. “But it becomes easier when you use a quantum system.”