Why science has become less innovative
According to a recent study, science is less “disruptive” than it used to be. What’s changed?
The 20th century was a golden age for science. Albert Einstein published his general theory of relativity, Ernest Rutherford discovered the atomic nucleus and Alexander Fleming introduced us to penicillin. We also learned the universe was expanding, came up with the Big Bang theory and learned about DNA for the first time. All these discoveries were revolutionary. It’s hard for us to imagine science without them.
This century is still young and has already given us a few exciting findings — the detection of water on Mars, CRISPR and countless medical advancements. But when we look back on the explosion of scientific knowledge last century, research that shakes the foundations of science feels like it’s gotten rarer.
A paper published in Nature earlier this month suggests modern science really is less inclined to challenge itself. After analysing millions of papers and patents, the authors concluded that there had been a decline in “disruptive” research over the last century.
How do you go about measuring scientific ingenuity — and what factors have led to its deceleration?
The study
Michael Park and colleagues quantified how disruptive scientific papers and patents were using the consolidation-disruption (CD) index. A score of 1 represented a highly disruptive paper, while a score of -1 represented a paper that didn’t rock the boat at all.
The CD index is calculated by looking at papers that cite a certain study and checking how likely they are to also cite that study’s predecessors. If a study changes the direction of its field, subsequent research is less likely to cite earlier work. For example, after Watson and Crick unveiled the double helix structure of DNA, studies largely stopped referencing Linus Pauling’s triple helix model.
From 1945 to 2010, the average CD index declined, indicating that papers had become less disruptive. While the rate of decline was particularly steep for the social sciences and technology, all scientific fields considered in the study experienced a downward trend in their average CD index.
The language used in the titles of papers also appeared to have shifted. Recent papers were less likely to use verbs associated with creation (“produce”, “make”, “form”), discovery (“determine”, “report”) and perception (“measure”). Such language has been replaced with verbs that hint at the improvement (“improve”, “enhance”, “increase”), application (“use”, “include”) and assessment (“associate”, “relate”) of previous discoveries.
These observations suggest that over time, new studies have become less likely to push their fields in different directions.
It’s important to note that the absolute number of highly disruptive works has remained stable over time. But the fall in the proportion of research that can be considered disruptive still raises questions. Today’s scientists have more knowledge to build on than ever and technology their predecessors could only have dreamed of. Why isn’t progress accelerating?
Low-hanging fruit
Park and colleagues consider a few reasons why research may have become less disruptive on average.
One explanation is that the low-hanging fruit has just already been picked. Evidence suggests discoveries might be harder to come by nowadays. In 1965, Gordon Moore famously predicted the number of transistors on microchips would double every two years. But in 2017, Nicholas Bloom and collaborators found that it now takes 18 times as many researchers to double the number of transistors on a microchip as it did in the early 1970s.
It’s also tempting to wonder if there’s much left to discover. We know the Earth goes around the Sun, we know how genes are passed on, we know where new species come from — surely all the basic facts are covered now?
But we don’t know what we don’t know. Genetics wasn’t born until the mid-19th century. All the relevant discoveries we’ve made since then — DNA, RNA, the triplet code, genetic engineering — would have meant nothing to a Georgian. Who’s to say we’re not overlooking something just as significant?
It would be especially arrogant to presume we know it all when people from previous generations have thought the same thing. In 1894, Albert Michelson said of the physical sciences, “It seems probable that most of the grand underlying principles have been firmly established.” Quantum mechanics and Einstein’s theory of relativity had yet to be born.
Park and colleagues are sceptical of the low-hanging fruit explanation. The decline in disruptiveness is remarkably similar between different disciplines. “It seems unlikely fields would ‘consume’ their low-hanging fruit at similar rates or times,” they write. Furthermore, the consistency in the absolute number of disruptive studies suggests the age of scientific discovery is far from over.
Has research gotten worse?
It’s a distasteful question, but an obvious one: Is science simply not what it used to be?
The authors address this idea too. If the quality of published research really has just deteriorated, the decline in innovation should be less apparent when we only consider high-quality research. But when the analysis was restricted to studies published in the most discerning journals, or discoveries that had won a Nobel prize, the trend remained the same.
The narrowing of focus
Park and collaborators’ primary explanation for the decline in disruptive science is a narrowing of the use of previous research. The diversity of work cited has declined. Scientists and inventors keep referencing the same papers, and these papers tend to be topically similar. Though there are more papers out there to peruse than ever, the “effective” stock of knowledge has narrowed.
Essentially, a lot of science is being produced without getting used. Since previous knowledge informs future progress, this narrows the scope of potential discoveries.
The quantity of new research being published keeps increasing. Contrary to what you might think, the torrent of fresh knowledge may actually be detrimental to scientific progress. If too many studies are being published in your field, keeping up with new research becomes overwhelming. You can’t read everything, so you’ll probably focus on reading the papers that everyone else is citing. Obscure but innovative studies can easily pass you by.
As knowledge accumulates, those who wish to make new discoveries face a higher educational burden. You can compensate for this burden by narrowing your expertise, but innovation often comes from the novel union of separate ideas. Darwin was smart enough to come up with natural selection, but to understand evolution, his theories had to be married up with those of Gregor Mendel, the pea plant-growing monk who explained how traits could be transmitted from one generation to another.
Due to this narrowing of expertise, modern scientists work in larger teams than ever before. But if you’re searching for new ideas, too many researchers spoil the broth. Wu and colleagues found that larger teams tend to focus on popular developments and are less likely to produce disruptive work.
Publish or perish
When Nature summarised Park and colleagues’ findings with a bit of a shrug, evolutionary biologist José Cerca had some strong words:
Anyone who’s interested in research will have encountered the phrase “publish or perish”. Academic institutions frequently use the amount of papers a researcher publishes as a measure of their competency. To advance your career, you’ve got to keep churning out new papers.
The desperation of researchers to publish is evident. Online advertisements offer the chance to buy authorship on research papers. Alternatively, you can purchase a fake manuscript from a paper mill.
If you’re scrambling to get as much research published as possible, you’re less likely to pursue risky avenues of thought that could lead to disruptive discoveries. Your institution may not even allow you to carry out unconventional research. Psychologist Julie Nagel describes the distress of a researcher whose tenure was denied because his work was “too creative”.
Additionally, overvaluing publication decreases the general quality of research. Studies may be duplicated or split into tiny pieces (“salami slicing”), making the literature even harder for scientists to sort through.
Scientific integrity also suffers. While only 1–3% of scientists have committed outright fraud, 75% have engaged in questionable research practices. Negative findings are important — if a drug is ineffective, we’re better off knowing — but they’re less attractive to publishers.
If you’re desperate to get published, you might employ dubious techniques to squeeze positive results out of your study. You could exclude data that contradicts your hypothesis, fiddle with your statistics or claim you expected an unexpected finding from the start.
As Cerca mentions, this mindset is also toxic for researchers, and unhappy people don’t tend to do their best work. If you’re overworked and miserable, you’ll probably lack the time and energy to slow down, read widely and explore new ideas.
Encouraging disruptive science
How can we increase the proportion of disruptive research?
Park and colleagues suggest that research needs to slow down. Universities ought to reward research quality over quantity, and scientists should be given time to keep up with new developments. Moving away from the publish or perish mindset and encouraging researchers to read widely could foster an environment more conducive to the publication of ground-breaking science.
But we should also remember that not all research needs to be disruptive. For example, you might want to conduct a clinical trial of a medical treatment that should work in theory but hasn’t yet been tested.
“There are super important discoveries that aren’t disruptive but advance a field or make a technology viable,” says Russell Funk, a co-author of the Nature paper. “You really need both.”
We’re experiencing a decline in disruptive science, but we’re also experiencing a replication crisis. It’s important to check whether other scientists’ findings are valid by attempting to reproduce their results. We can’t take it for granted that all research will stand up to scrutiny. More than 70% of scientists have tried and failed to reproduce other researchers’ results, and more than half have failed to reproduce their own. Science isn’t at its best when everyone is at the cutting edge.
Yet the cause of — and the solution to — the lack of disruptive studies and the lack of reproducibility studies may be the same. Reproducibility studies are unpopular because they’re unattractive to publishers. As Stuart Ritchie explains in his book Science Fictions, some journals actually have a policy against publishing studies that repeat an experiment, even if they show the results of the original experiment to be flawed.
To accelerate scientific progress, academic institutions must prioritise the quality of a researcher’s contributions to their field, not the size of their publication record.
