By Ryan Ripsman
The Foundational Questions Institute, a non-profit organization dedicated to answering fundamental questions of physics and cosmology, runs a series of essay competitions. Each competition is centered around one question. Past competitions have asked questions like “what is the nature of time?” and “how should humanity steer the future?”. The contest receives submissions from world-renowned physicists, mathematicians, and philosophers and has prizes worth $40,000 in total. This year’s contest asks the questions, “How could science be different?” and “How could science be better?”
At first glance, it may seem like these are absurd questions. Science is the search for truth. It is a set of protocols that allows us to peek behind the curtain of our reality and uncover the rules governing our universe. But at its core, science is a human endeavour, and as such science is subject to human imperfections.
Before discussing ways science could be different or better, it is important to address how science is studied. There are two academic disciplines that study the nature of science: metascience and the philosophy of science. Both disciplines tackle different aspects of the field. The philosophy of science focuses on the theoretical aspects of science; it discusses how a
scientific theory is made and how incorrect theories are discarded. It discusses the scientific method and how the actual endeavour of science departs from this method. Metascience, on the other hand, focuses more on the practical aspects of science. Using statistical methods, metascience examines how science is conducted and what biases and inefficiencies arise from our current system of scientific analysis.
The philosophy of science has been studied since antiquity. While it is difficult to discuss a field as broad as this one in a single article, I think a good place to start is with one philosopher, Thomas Kuhn. Arguably the most influential philosopher of science of the 20th century, Thomas Kuhn wrote a book called The Structure of Scientific Revolutions. One of the most important concepts Kuhn introduced is the idea of a paradigm, which is a set of beliefs that underpins one’s worldview. Kuhn argues that scientific progress can be broken down into distinct phases. At first, scientists assume their paradigms are correct. Over time, they accumulate knowledge and build theories based on their paradigms. When phenomena are observed that conflict with the accepted paradigms, scientists develop more and more complicated theories to fit these phenomena into existing paradigms. Eventually, when enough results have been found that challenge the current paradigm, a crisis occurs, leading to a re-examination of the fundamental beliefs and a change of paradigms, called a paradigm shift. Kuhn argues that these paradigm shifts do not occur by convincing old scientists that their fundamental beliefs are wrong. Rather, Kuhn felt paradigm shifts could only occur when the old generation of scientists pass away and a new generation of scientists who hold different beliefs take their place.
For a concrete example of Kuhn’s principle of paradigms, consider the development of quantum mechanics. Quantum mechanics upended the way scientists see the universe and forced them to reconsider their fundamental assumptions. When the theory of quantum mechanics was first developed, many famous scientists rejected parts of quantum theory because it did not fit into their paradigm. In fact, even theorists who were essential to the development of quantum mechanics, like Einstein and Schrödinger, never fully accepted quantum mechanics. It was only when the previous generation of scientists passed away that
the paradigm shift was completed, and the quantum mechanical view of the universe was fully embraced.
While Kuhn’s paradigm shifts may be an unavoidable part of science, certain aspects of how science is conducted delay the emergence of new paradigms and keep outdated ones at the forefront of scientific study. For instance, an accepted part of scientific research is the idea of peer review. When scientists want to publish their scientific results, they are examined by a set of experts in the field, who decide if the results hold up to scientific scrutiny and are worth disseminating. Peer review is a critical part of scientific quality control and ensures that pseudoscience is not circulated. But it can also be argued that in certain cases, peer review can hold back new bold theories because they are in opposition to the paradigms held by the established scientists. Oftentimes, radical ideas are rejected by the scientists of the time. For instance, Copernicus’ idea of a solar system centered around the sun contradicted the views of most astronomers of the time. However, Copernicus was still able to publish his results, which led to the eventual acceptance of his theory. In modern times, when scientists’ research must pass the scrutiny of a jury of their peers before publication, it is questionable whether radical ideas like Copernicus’ heliocentric model of the solar system would ever see the light of day. In the future, peer review could be modified to ensure the quality of scientific output is maintained without slowing down the pace of progress.
Metascience recently emerged as a new scientific discipline in light of the discovery of the replication crisis. Starting in the early 2010s, it was found that many landmark psychological studies could not be replicated. From there, it quickly became clear that the problem was not restricted to just one scientific field. Replication is essential to science. To be sure a phenomenon is real and not just a fluke, other scientists need to be able to follow the same methods and obtain the same results. Metascientific analysis found that studies that cannot be replicated tend to be cited by the same number of articles as replicable research. This result indicates that irreplicable research influences and misleads other scientists, hurting the field of science as a whole.
Based on the metascientific analysis of the replication crisis, researchers have suggested rules to prevent the dissemination of irreplicable science. One common scientific practice that gives rise to the crisis is called p-hacking. The scientific method dictates that scientists should develop hypotheses based on established theories, and then perform experiments to determine if their hypothesis is correct. Sometimes scientists will perform an experiment to find evidence supporting a certain hypothesis and be disappointed to find that their hypothesis is not supported. The scientists will then start to examine alternative hypotheses, until they find a hypothesis which the data supports, a process called p-hacking. The issue with this practice is if you examine enough hypotheses, and try enough statistical tests, eventually you will find a statistically significant result by sheer chance. Studies that use this approach are often irreplicable. Researchers have suggested that the practice of preregistering experiments would prevent scientists from improperly reusing data. When an experiment is pre-registered, the scientist details a research plan before collecting data and commits to using the data to test only one hypothesis. It has been suggested that this approach could help reduce the number of irreplicable research studies.
Metascience has also been used to identify inequities in science. It has been shown that diseases predominantly affecting women have received less research attention and as a result, tend to be less understood than diseases affecting men. This discrepancy has, in part, been attributed to the historical scarcity of female medical researchers and the continued underfunding for research focused on women’s health. Beyond medicine, inequities exist in most fields of research. Multiple studies have shown that researchers from the global south have a much more difficult time publishing in prestigious journals and receiving the necessary resources for their research. As a result, issues that predominantly affect the global south receive less attention than issues that affect the global north.
When thinking about how science could be different, it is not enough to look at how science is currently done. It is also important to consider how advances in science and technology will change how science is conducted. A recent study has shown that artificial intelligence (AI) can be used to identify articles that are irreplicable. The study showed that the
AI was able to provide better predictions on whether an article will be replicable than human reviewers. Traditionally, peer review has been viewed as a “human” process conducted by other scientists. But, given the problems with peer review and irreplicable science discussed above, meta-scientists will need to consider whether moving peer review to an AI-based process will improve science.
AI also raises philosophical questions about the nature of science. Traditionally, constructing theories that explain a wide variety of phenomena has been viewed as central to the act of science. Theories are important because they allow us to gain a better understanding of nature and predict what will happen in new scenarios that have not been tested before. But what if we could develop methods for predicting different phenomena without actually developing theories? Deep learning, a new AI framework, is able to discover complicated patterns that would be difficult for humans to find on their own. However, deep learning systems have a much more difficult time explaining these patterns in a manner that humans can understand. As deep learning is integrated into science, we will need to decide what constitutes a scientific discovery. Should we “close the book” on scientific questions we can answer but not understand? We will need to decide what is more important: predicting or understanding.
As in previous years, this year’s Foundational Questions Institute essay competition is bound to attract entries from some great minds of our generation. If you think you number among them, this contest may be an opportunity for you to share your ideas with the world. But even if you don’t count yourself amongst these luminaries, I think it is important you continue questioning science. While our current approach to science has achieved great things, there are always benefits to improving it. And, even if you don’t want to be a part of the scientific enterprise, it is still important to think and question science, because questioning is an essential part of what makes us human.