Ch 3: Paradigm Shifts or Cumulative Progress — Summary / Overview

Understanding how powerful tool innovations unlock breakthroughs leads us to fundamental questions about how science evolves: do we abandon some major discoveries—including breakthrough tools? Is science best seen as a series of sudden paradigm shifts—or rather as deeply cumulative? In the landmark book The structure of scientific revolutions, Thomas Kuhn developed the most influential and cited explanation of science yet proposed. For Kuhn, science does not progress cumulatively but is driven by abrupt paradigm shifts: radical upheavals—like the Copernican revolution—that overturn what we believe and can redefine entire fields. Since then, this core question about scientific progress is widely debated, but no consensus and comprehensive analysis yet exist. By studying this question systematically, we uncover a remarkable pattern: three key measures of scientific progress—science's major discoveries, methods and fields—each illustrate the deeply cumulative evolution of science rather than sudden ruptures. First, no major scientific methods or tools we use across fields have been entirely abandoned or subject to paradigm shifts—from statistics and microscopes to centrifuges. Next, no major scientific fields have been entirely replaced—from genetics and computer science to chemistry. Instead, we refine and extend them over time, often over centuries, and they represent vast bodies of cumulative knowledge. Finally, scientific discoveries are also highly cumulative: only 1% of over 750 major discoveries have ultimately been abandoned. Our methods and tools provide the strongest and most striking evidence of science's deep continuity. So we offer a new answer to this foundational question in science and the philosophy and history of science by drawing on methods from statistics and empirical sciences. This brings us to questions about the classic scientific method of testing hypotheses by observing and experimenting—a principle that has remained unchanged for centuries. Is it compatible with the cumulative nature of science? Do science's major discoveries follow it? Or do we need a new understanding of scientific methodology? The 78 T HE ENGINE OF SCIENTIFIC DISCOVERY

Does science mainly go through revolutionary paradigm shifts that fundamentally challenge and overturn what we believe? Or is science mainly a deeply cumulative system, where our current methods and discoveries connect back to our past methods and discoveries and extend forward to enable future methods and discoveries? The question is important because it shapes how we approach scientific research and science education and policy. Should we focus on conducting bold, disruptive science or on refining existing science? Two dominant ways have shaped how researchers study the evolution and progress of science. One way is taken by the highly influential historian of science Thomas Kuhn who offered the most well-known explanation of science to date. Using select case studies, he explored how individual theories may go through extraordinary paradigm shifts: fundamental changes in the theories of a field. Many still find this classic view of how science unfolds compelling— and many still research his view of non-cumulative science. Kuhn's seminal book still dominates debates on scientific progress, accruing thousands of citations each year (Google Scholar). The other way to study scientific progress is taken by scientists using big data to analyse patterns in publications. These scientists argue that scientific articles—on the whole—may be becoming less revolutionary over time. They highlight that bold, high-risk innovation is becoming rare as researchers focus more on established knowledge. These scientists suggest that smaller teams are more likely to generate disruptive ideas than larger teams. They also stress that the sheer volume of publications in today's scientific system can lead us to overlook new transformative ideas. Researchers adopting big data trace science through article citations—the most widely used measure of the impact of discoveries. But this approach does not uncover other important factors to understanding scientific progress. Citations are driven in part by path dependency—researchers with limited time often just cite more cited research—that can impede the spread of new groundbreaking ideas. Focusing on already well-known established research (impact) can hinder cuttingedge breakthroughs (novelty). Citations do not capture the large impact of most major discoveries throughout history, simply because we only began to track citations widely in the second half of the 20th century. Citations also do not capture scientific fields or our powerful tools well. Most studies—using citations or not—explore a sample of major breakthroughs or publications. Some do suggest that science has cumulative aspects by relying on case studies. But no comprehensive analysis yet exists that answers the foundational question: is science—namely science's major discoveries, methods and fields—fundamentally cumulative or revolutionary? Here we aim to help tackle this long-standing debate. Here we move beyond commonly studying just a sample of discoveries or theories often within a field (as Kuhn did mainly in physics) or using a sample of article citations. Instead, we systematically analyse the nature of scientific progress through science's major discoveries across fields. In probing the evolution of science, we uniquely shift the focus here to also study an overlooked aspect: our R scientific methods and tools. This enables us to draw completely new insights about the cumulative nature of science's major discoveries and methods across fields. Introducing this novel method perspective, we find an extraordinary continuity in the methods that we refine over time and enable generating new theories, breakthroughs and fields. A shift from analysing at the individual to the aggregate level helped transform our understanding in numerous fields. It is how Boltzmann and Maxwell developed statistical mechanics, and how Austin Bradford Hill's randomised controlled trials paved the way for cumulative meta-analyses that reshaped biomedical, agricultural and behavioural sciences. We will explore specific fields. In genetics for example, a set of major advances was only possible using our cumulative methods including microscopes, x-ray methods and electrophoresis we continually upgrade. These sparked a series of transformative discoveries, including heredity, DNA sequencing and mapping the human genome, that keep expanding genetics, improving our health and reducing diseases we face. In computer science, a set of major advances was only achieved using our cumulative methods including mathematical and statistical methods and transistors we constantly improve. These led to a series of discoveries, including the Turing machine, information theory (often called the Magna Carta of the digital age) and microchips, that constantly extend computer science and enable the computers, smartphones, internet and artificial intelligence that we use and define our modern society. In the field of electricity, a set of major advances was only feasible using our cumulative methods including galvanometers, batteries and electric generators we continually enhance. These triggered a series of discoveries, including electromagnetism, the theory of electromagnetic radiation and alternating current, that collectively enable the world of electric motors and power plants we rely on daily. The story of genes, computers and electricity is a remarkable story of cumulative evolution—a story of how we stand on the shoulders and methods of giants. And they are representative of major fields across science. So even if new breakthroughs can at times challenge and disrupt research, science is still overall cumulative. Box 3.1 Thomas Kuhn's explanation of science—driven by revolutionary paradigm shifts Kuhn published the most widely read and best-selling book on scientific progress to date, The structure of scientific revolutions, in 1962. Written while at the University of California, Berkeley, this pioneering book shaped popular understanding of the history of science. The history of science can be seen as a cycle in which established ideas and facts are doubted, new problems and evidence then lead to new revolutionary ideas and facts (and replace the established ones), which eventually over time are also doubted once problems and anomalies with them become apparent, and the cycle begins again. For Kuhn, the process of science reflects revolutionary paradigm shifts, in which we reject current assumptions and theories and adopt entirely new ones. He argues that a paradigm shift 'is far from a cumulative process … Rather it is a reconstruction of the field from new fundamentals' and 'cumulative acquisition of novelty is not only rare in fact but 80 T HE ENGINE OF SCIENTIFIC DISCOVERY im probable in principle' . 'Scientific revolutions are thus disruptive episodes of fundamental reconfigurations, through which scientific knowledge develops in a noncumulative way' . Revolutionary science is not cumulative for Kuhn because scientific revolutions (major breakthroughs) replace our current views and theories entirely—and here we focus on this central hypothesis (not on everyday normal science). A cumulative (or disruptive) advance in science can lead to a cumulative (or disruptive) advance in technology; and the same applies from technology to science. Kuhn's central and bold thesis is: not just some but 'All significant breakthroughs are break-"withs" old ways of thinking'. But there is a paradox here— changes in theories are the result of a new breakthrough and not what brings about the new breakthrough. Changes in theories cannot explain how science's breakthroughs arise in the first place. This is the key gap and what we want to understand. With a degree in physics and a PhD in history of science, this laid the background for Kuhn's hypothesis of paradigm shifts. He studied select theories especially in physics up to the early 20th century. Yet he still made sweeping claims about all of science. By focusing on the iconic examples of radical changes in theories of physical reality from Aristotle to Newton to Einstein that span over two millennia, these cases may seem to partly support his hypothesis. For Kuhn, the stark shift from the Ptolemaic earth-centred theory of the universe to the Copernican sun-centred theory characterises the classic paradigm change—and he focused much research on it. Yet by examining the paradigm shift hypothesis with over 750 major discoveries, we uncover that the dramatic shift from Ptolemy's model (developed in the year 150) to Copernicus' model (developed in 1543) presents one of the few exceptional cases when we abandoned a central model— though it was in early science. And it was cutting-edge new methods and tools that supported and confirmed Copernicus' model.