Mar 9, 2020

Want To Accelerate Scientific Breakthroughs? Reward Scientists For Disproving Their Own Theories

Adversarial collaboration is at the core of Templeton World Charity Foundation’s grantmaking strategy for neuroscience.

By Dawid Potgieter

When Albert Einstein first published his theory of special relativity in 1905, it shook physicists’ understanding of the universe to its core. Suddenly, there was a new mathematical theory that disagreed with the prevailing view of the universe under Newtonian physics. Something would have to be done before the field could progress further.

While Einstein’s theory was earth shattering, it was the testing of his theory against the old Newtonian theory, in a process now known as adversarial collaboration, that would change the way we conduct science forever. Now, Templeton World Charity Foundation (TWCF) is putting this same principle at the center of its grantmaking around neuroscience and the study of consciousness. 

Watch a new video featuring Lucia Melloni, Christof Koch & Daniel Kahneman on adversarial collaboration.

By 1915, Einstein’s theory of general relativity was substantially fleshed out, but there continued to be a fundamental tension between the two cosmological views: Einstein’s theory introduced concepts such as spacetime and gravitational time dilation that were incompatible with Sir Isaac Newton’s theories, which had been dominate for over two centuries. Where Newton’s model indicated that light should always travel in a straight line, for instance, Einstein argued that gravity would warp the fabric of space and cause light to bend its trajectory.

Once World War I ended and Europe returned to a semblance of normalcy, it was clear that the field of physics would remain stalled until there was a definitive conclusion about which theory was correct. On May 29, 1919, the British astronomer Sir Arthur Eddington — working from an experimental framework developed two years earlier by Astronomer Royal of Britain Sir Frank Watson Dyson — tested the two theories against each other. The experiment, which looked at whether light from distant stars was bent by the gravitation of the sun during a solar eclipse, was designed to effectively eliminate either Newton’s or Einstein’s theory. In the end, Einstein won.

The importance of this cannot be overstated. On one level, it opened a myriad of new pathways for inquiry in the fields of physics and cosmology. But on another, it demonstrated how competing theories could be pitted against one another in an experimental framework to achieve scientific advancement.


What is adversarial collaboration?

The Eddington experiment is perhaps the most high-profile example that helped lay the groundwork for adversarial collaboration. Scientists in other disciplines have since also made use of this practice. Adversarial collaboration in science refers to any exercise when people with views that are adversarial — in opposition to one another — work together to find out who is right or wrong.

In spite of the simplicity of the concept, adversarial collaboration in the formal sense is quite rare. It can also be very difficult because it requires a huge amount of commitment from both sides. Adversarial collaboration in psychology was popularized in the 1990s and 2000s by Daniel Kahneman, and more recently by Susan Fiske and others.

The research we funded under our Accelerating Research on Consciousness (ARC) initiative is the first example of a formal adversarial collaboration in neuroscience that makes use of best practices in open science and different neuroimaging methodologies.


Yet, while this approach has become fundamental to the practice of physics research, for instance, other fields, particularly the study of neuroscience and human consciousness, have lagged behind due to a lack of incentives to engage in adversarial collaboration and powerful negative incentives that discourage scientists from testing theories at the risk of being wrong.

There are currently a dozen or so competing, and mostly incompatible, theories of consciousness. No doubt more theories will be established over time. Neuroscientists and their funders face a dilemma now: Should we explore all of them at the same time? If not, how do we choose what theory to put our time, expertise and money behind? What if we can identify some of the incorrect theories sooner rather than later?

By analogy, a doctor trying to diagnose a patient might not want to run every test at once. That would be wasteful. Instead, the doctor may first rule out certain possibilities. Likewise, adversarial collaboration can be used to reduce the number of theories through falsification. Imagine if we could reduce the current number of theories of consciousness by half. That could increase the resources going towards testing each remaining theory and demonstrate a way of making continuous progress. This could vastly accelerate the progress of scientific discovery.

There is a problem though: The current trends in science funding encourage scientists to build careers defending their own pet theory at the expense of others. For many scientists, their ability to continue to fund their research depends on showing positive results, not disproving their own theories. Likewise, a proliferation of theories within a very limited funding pool can put pressure on researchers to prioritize safe experiments in the writing of grant proposals.

In an attempt to mitigate for these negative incentives and accelerate scientific breakthroughs, Templeton World Charity Foundation is putting the promotion of adversarial collaboration at the core of its grantmaking. We are beginning this effort by backing a $5 million adversarial collaboration between two leading theories of consciousness: “global workspace theory” and “integrated information theory.” To access these funds, leading proponents of each theory had to agree in advance on an experimental framework and the rules — so to speak — that will be used to decide whether or not a theory has been falsified.

In a very simplified summary, global workspace theory supposes that consciousness occurs when information is processed and broadcasted from the prefrontal cortex of the brain, while integrated information theory argues that consciousness derives from specific kinds of neural interactions centered closer to the back of the brain. (For those interested in a more in-depth examination of the two competing theories, I recommend this piece in Quanta magazine.) If the adversarial collaboration is successful, we will have eliminated one of these theories from contention and helped researchers narrow the focus of their work.

By backing experiments in this way, TWCF aims to accelerate scientific inquiry and improve the quality of science simultaneously by eliminating old theories and rewarding scientists for sharing data, collaborating and — potentially — disproving their own theories. With a bit of luck, the field of neuroscience can find its own Eddington experiment sooner rather than later.

Listen to a new podcast on adversarial collaboration, featuring Susan Fiske, Daniel Kahneman, and others.