Imprimir Republish

Knowledge production

Analysis of millions of papers and patents suggests science is becoming less disruptive

Study indicates limits on the ability of researchers to change the direction of fields of knowledge

Julia JaburIs science losing its ability to transform society? This fear, even if not expressed out loud, permeated the repercussions of an article published by three American social scientists in the journal Nature in January. Erin Leahey of the University of Arizona’s School of Sociology and Russell Funk and his doctoral student Michael Park, both from the Carlson School of Management at the University of Minnesota, analyzed more than 45 million articles published between 1945 and 2010 and 3.9 million patents filed in the USA between 1976 and 2010, concluding that over time, it became more difficult for scientists to make original discoveries capable of creating a new field of knowledge or reorienting the course of an existing research topic.

The trio used a metric called the CD index, the values of which range from -1 for the least disruptive studies to 1 for the most disruptive, and showed that the average value of the indicator fell gradually over time (see graph). The total decline over the studied periods was more than 90% for articles and more than 78% for patents. The absolute number of articles and patents considered transformative has actually remained at the same level over time. But since the total volume of scientific articles and patent applications has grown exponentially—it is estimated that around 3,000 articles are currently published per day—they are proportionally less likely to be classifiable as disruptive, leading the authors to conclude that the formidable effort undertaken to produce knowledge is yielding fewer dividends than investors hope.

The paper split innovative science into two categories. The first, known as disruptive, is characterized by “disrupting existing knowledge,” propelling science and technology in new directions, according to the article published in Nature. The second, defined as consolidating, “improves existing streams of knowledge, and therefore consolidates the status quo.” In order to classify discoveries into one of these categories, the team analyzed the bibliographies of the articles and patents. Consolidating studies tend to be cited in subsequent works alongside related articles that came before, suggesting that several cumulative advances have been made in the area. With disruptive papers, previous studies generally stop being cited, no longer seen as a reference for the new direction the field has taken. A healthy scientific ecosystem is one where there is a mix of disruptive discoveries and consolidating improvements,” Russell Funk told the Carlson School of Management website. “With incremental innovations being more common, it may take longer to make those key breakthroughs that push science forward more dramatically.”

One case of disruptive science mentioned in the article is widely known: the discovery of the double helix structure of DNA, presented in the journal Nature in 1953 by Francis Crick and James Watson (who won the Nobel Prize in Physiology or Medicine nine years later). Other examples have proven controversial. The three social scientists used six scientific contributions to illustrate the spectrum from consolidating to disruptive. A 1983 patent filed by Richard Axel, Saul Silverstein, and Michael Wigler of Columbia University, which described a new and efficient way to insert DNA into mammalian cells, ranked highest on the scale. At the other end, as an example of consolidating science, is a 1970 article by microbiologist David Baltimore, of the Massachusetts Institute of Technology (MIT) at the time, on the discovery of reverse transcriptase, an enzyme that catalyzes DNA formation from an RNA template.

In an opinion piece published on the news website STAT, geneticist George Church of Harvard Medical School and Juergen Eckhardt, vice president of Leaps by Bayer, an arm of the German multinational that aims to invest in groundbreaking research, highlighted paradoxes in the selected examples. The patent considered disruptive, they noted, did not represent a surprising discovery at the time, since at least two other groups had inserted DNA into cells before. The discovery of transcriptase, meanwhile, was “conceptually very surprising,” as it provided a way to study and commercialize RNA and earned David Baltimore a Nobel Prize in 1975.

Also categorized as “consolidator” was an article published in the journal Physical Review in November 1965 by Austrian-American chemist Walter Kohn, winner of the 1998 Nobel Prize in Chemistry, and his colleague Lu Jeu Sham, a postdoctoral researcher from the University of California, San Diego. The paper is a key reference for one of the most popular computational methods of quantum mechanics, used to describe electronic properties in solid-state physics, materials science, chemistry, and other fields. “It is one of the most cited articles in the history of physics,” explains Klaus Capelle, a physicist from the Federal University of ABC. “It changed the way we make calculations in the physics of condensed matter, quantum chemistry, materials science, and even in some areas of biology, pharmacology, and geophysics. If the CD index classifies this article as consolidating, there is something questionable about its methodology or its interpretation.”

The paper was probably identified as consolidating because whenever researchers cite it in subsequent articles, they also have to cite a study published a year earlier by Kohn and another postdoctoral colleague, Pierre Hohenberg. The first article described the basic theorem, while Kohn and Sham’s paper demonstrated how to perform calculations based on this theorem, presenting a mathematical algorithm for calculating the properties of solids. Materials used in computer processors or cell phones, for example, were developed based on this algorithm.

Hernan Chaimovich of USP’s Institute of Chemistry believes some of the criticism of the Nature article results from a misunderstanding. “It highlights the importance of innovative science, whether disruptive or consolidating,” he says. “Innovative science that consolidates a field can have as much impact as a disruptive paper.” For FAPESP’s scientific director Luiz Eugênio Mello, the discussion sparked by the article is positive. “Any quality analysis based on 45 million samples evidently has some merit and the criticism of certain exceptions does not detract from the overall benefit,” he says. “It is worth noting that several examples of ‘nondisruptive’ contributions represent fundamental revolutions. If there was indeed a relative decrease in disruptive science, it should be remembered that never before has humanity developed a vaccine—or 10 of them for that matter—in less than a year,” says Mello, referring to COVID-19 vaccines. The unprecedented new messenger RNA vaccines for the novel coronavirus were classified in the article as consolidating science, since they resulted from a gradual development.

Questions over the methodology were not enough to quell the debate sparked by the article, which raises concerns about the productivity of science in previous studies. Naercio Menezes Filho, an economics professor and head of the Brazilian Center for Early Childhood Research (CPAPI) at INSPER, wrote an article in the newspaper Valor Econômico comparing the conclusions of the Nature article with those of a paper published in American Economic Review in 2020 by four economists, including John Van Reenen, who was his doctoral advisor at University College London in the 1990s.

Titled “Are Ideas Getting Harder to Find?,” the study concluded that the number of researchers in the USA needed to be greatly increased to maintain the same level of productivity. “More and more people are being allocated to research, whether in companies or academia, but it is more difficult than ever to find new ideas to be applied,” says Menezes. The authors show that between 1960 and 2015, productivity in agriculture grew at a rate of 1.5% per year, while the number of agricultural researchers increased by 5%.

Several hypotheses for the lack of impetus were discussed. The problem may be related to how difficult it is for scientists to stay on top of all the knowledge generated in their fields due to the growth in the scientific literature and a tendency to focus on specialist topics that yield incremental innovations. There are other variables. The rising number of active researchers—who are required to constantly show what they are doing—has made science much more competitive. Scientists often spread their results across several scientific articles, instead of disclosing the main findings in one manuscript of greater weight.

According to Paulo Artaxo of USP’s Institute of Physics, the way science is conducted has changed radically in recent decades. “100 years ago, science was carried out in just a few laboratories in the developed world. Today, it is done in 200 countries. The democratization of science is a really good thing, and thousands or even millions of researchers are carrying out science today,” he says. There are collateral effects, however, says Artaxo. “No one can keep up with the thousands of papers published every day. We are only able to read the most visible ones. The publishing industry makes a lot of money from this approach to science. Hundreds of new journals appear every month. It is only natural that there is less novel content than there used to be,” says Artaxo. “Anyway, I wouldn’t say that disruptive science has declined, rather the total volume of scientific production has grown.”

Another possibility is that investment in lines of research at the cutting edge may be falling short of what is needed. “Researchers often do not want to risk dedicating themselves to a very innovative and long-term area of investigation because they cannot go too long without publishing articles—that would harm their chances of achieving new positions or promotions,” says Menezes Filho. Funding agencies have always had to carefully divide their investments between basic and applied science, or to use more recent terminology, fairly balance funding between research guided by the curiosity of scientists and research commissioned to solve major social challenges (see Pesquisa FAPESP issue nº 246). “Together with science spreading across a larger number of countries and people, there is also a greater diversity of formats, with discovery science, implementation science, and problem-solving science,” says Luiz Eugênio Mello. The foundation offers funding opportunities both for short-term projects, such as research grants, and long-term projects, such as five-year thematic projects and Research, Innovation, and Dissemination Centers (RIDCs), which are funded for up to 11 years, allowing scientists to invest in long-term studies without the pressure to publish quick results. In 2022, FAPESP launched a new initiative, the Generation Program, which is expected to have a total budget of R$56 million for 35 research proposals in its first round. Through this program, young researchers without employment up to six years after obtaining their PhD can request up to R$1.6 million for independent projects that aim to advance the boundaries of knowledge.

While it is possible that the number of original ideas is dwindling, it seems counterintuitive to say that science is losing its ability to change society, whether through disruptive or consolidating research. In the STAT article, Eckhardt of Bayer and Church of Harvard listed several highly promising discoveries on the horizon, such as immunotherapies for cancer, advances in genome editing, planet colonization, and aging reversal techniques.

For Hernan Chaimovich, the methodology proposed in the Nature article represents an interesting attempt to measure the innovative capacity of science. “But you have to be careful saying that disruptive findings are in decline. What we can say is that investment in science has increased substantially in recent years and this has significantly increased research of an incremental nature,” he says. Beyond the numbers, he points out, it seems that science is still producing discoveries capable of changing the course of humanity. “There are important studies based on smaller disruptions and there are also major disruptions in progress, such as the merging of artificial intelligence and quantum computing. The transformative potential of this is enormous, and as in previous examples, it is not possible to predict or imagine what it will represent in the future.”

Scientific article
PARK, M. et al. Papers and patents are becoming less disruptive over time. Nature. vol. 613, pp. 138–44. jan. 5, 2023.