The decline of disruptive science
We are publishing more than ever but we are now less innovative: why?
The data suggest something is changing, you don’t have quite the same intensity of breakthrough discoveries you once had.” — Russell Funk (source)
Over the years, there has been a definite increase in both funding and research (which is good news). But this begs the question: has an increase in funding and scientist numbers led to proportional increases in scientific discoveries?
On the one hand, there has always been a nostalgic feeling for the past (or the mythical golden age), on the other hand, several researchers have asked this question over the years. In 2018, an article in The Atlantic noted how Nobel Prizes in Physics were mainly awarded to research done decades earlier. Other research, shows how researchers are getting older when they make Nobel-valuable discoveries (37 years old on average versus 47 today). But more importantly, how do you quantify the breakthrough potential of a discovery?
Why is it important? Apart from deciding which scientist to hire, and which research to reward with funds and prizes, scientific research is the basis of everything. Modern medicine, artificial intelligence, and all modern technologies are applications of scientific discoveries.
Moreover, In recent years we have seen a huge increase in the publication of scientific articles, and the review process is not always linear. Not to mention, that recently published artificial intelligence models show that I will soon be using AI as an assistant in research, we must have a way to assess the quality and importance of discoveries
A new study published in Nature focused on these questions: how to define if a scientific article defines new directions of research and how to quantify it across time and fields.
First, some contributions improve existing streams of knowledge, and therefore consolidate the status quo. Second, some contributions disrupt existing knowledge, rendering it obsolete, and propelling science and technology in new directions. — Nature manuscript (source)
The authors defined a new measure they called the CD index. The intuition behind this measure is that if an article is disruptive, subsequent papers will no longer cite the predecessors (the references of that article). In other words, when an article changes the field, it makes the very articles it was citing obsolete. The CD index is ranging between -1 and 1, where a negative value means an article is a consolidation work and a positive value is a disruptive work.
An example, when Watson and Crick discovered that the real structure of DNA was a double helix they cited Linus Pauling’s idea (for whom the structure was a triple helix) in their article. Of course, Watson and Crick’s work (which also earned them a Nobel Prize) made Pauling’s work obsolete and no one cited it again. Or in Artificial intelligence after the publication of the Transformer, the use and citations of RNNs and LSTMs decreased significantly.
The authors conducted their analysis of 25 million articles (from 1945 to 2010 in the Web of Science) and nearly 4 million patents. In addition, they replicated their findings in 4 other datasets: JSTOR, the American Physical Society corpus, Microsoft Academic Graph, and PubMed (analyzing more than 40 million articles in total).
Across fields, we find that science and technology are becoming less disruptive. — original article
The results show a decline in both articles and patents. This affects not only social sciences (average CD dropping from 0.52 in 1945 to 0.04 in 2010), medicine and pharmaceutical sciences. Not only that, an obvious drop in technology as well (patents for ‘computers and communications’ saw an average CD dropping from 0.30 in 1980 to 0.06 in 2010)
The decline in disruptive science and technology is also observable using alternative indicators. -original article
The authors pointed out that disruptive articles can also be identified from an analysis of language. Articles with breakthroughs feature new words (which serve to indicate both new concepts and new paradigms). The authors thus noted a decline in unique words (a sharp decrease from 1950 until the 1970s and then a steady descent).
Moreover, earlier articles used words that evoked discovery or creation (‘produce,’ ‘determine’) while later articles prefer to use terms that represent incremental progress or improvement of an existing process (‘enhance,’ ‘improve,’ ‘associate’). In other words, the authors’ chosen language betrays the fact that they consider the work to be an improvement on previous discoveries or technologies (but not a new one).
These results suggest that the persistence of major breakthroughs — for example, measurement of gravity waves and COVID-19 vaccines — is not inconsistent with slowing innovative activity. In short, declining aggregate disruptiveness does not preclude individual highly disruptive works. — original article
The authors note that even though the percentage of disruptive articles has decreased, it does not mean that they have disappeared altogether (perchance!). Indeed, while the overall number of published articles has considerable increase over the years, the number of disruptive remained constant.
Clearly, it is important to understand the causes of the decline. Over the years, explanations have been possible:
- ‘low-hanging fruit’ is a theory that the easiest and most attainable discoveries and innovations have already been made.
- The increasing burden of knowledge, i.e., the fact that today a researcher needs much more training than in the past (being able to master a field detail requires much more study today).
- The cost of research has increased. Increased pressure to publish in the academy drives researchers to quantity plus quality (also discouraging riskier projects but which could lead to more disruptive discoveries).
- The trend toward ultra-specialization.
The authors meanwhile show that the cause of the decline in disruption was not a decline in the quality of science. According to them, the articles that led to the Nobel Prize progressively received a lower CD index (nor a change in publication or citation practices).
In addition, the authors tested whether the declining disruptiveness relates to the growth of knowledge. While some authors suggest that the growth of knowledge should foster discovery and invention, it is also true that in recent years researchers have encountered an increasing knowledge burden. Using regression analysis they showed that there is a positive relationship between the growth of knowledge and disruptiveness for papers. Unexpectedly, the authors observed a negative effect on patents.
The growth in publishing and patenting may lead scientists and inventors to focus on narrower slices of previous work.
The authors suggest that this abundance of knowledge is not exactly a good thing. After all, today there is a decline in the diversity of works cited (there is in fact an increase in the use of only 1% of most highly cited papers and patents), an increase by researchers in reusing the same citations, and an increase in self-citation.
The authors also note another surprising result:
The mean age of work cited, a common measure for the use of dated knowledge is increasing , suggesting that scientists and inventors may be struggling to keep up with the pace of knowledge expansion and instead relying on older, familiar work.
Parting thoughts
In conclusion, this work shows that there is a decline in disruptive science and technology in recent decades. The authors suggest that the cause is ‘to scientists’ and inventors’ reliance on a narrower set of existing knowledge.’
Other studies, discussed this decline in innovation in science (though none in this extensive manner). Others have also pointed to other factors such as a more competitive environment, higher costs, large groups, and consortia (another research suggests that consortia mainly produce incremental research).
In addition, this article shows that scientific research today is under scrutiny, and some traditional problems have been exacerbated in recent years. Today this discourse is increasingly important, an increasingly precarious and underpaid academic environment but also the increasingly extensive use of data and algorithms (sometimes without precise knowledge or lack of expertise), seems a recipe for disaster.
As the authors suggest perhaps it is time to rethink career planning and science policy. Also, scholars should be promoted toward potentially more disruptive projects, and prevented from being overwhelmed by the knowledge burden. Clearly, this would require a more institutional effort as well:
Universities may forgo the focus on quantity, and more strongly reward research quality, and perhaps more fully subsidize year-long sabbaticals. Federal agencies may invest in the riskier and longer-term individual awards that support careers and not simply specific projects, giving scholars the gift of time needed to step outside the fray, inoculate themselves from the publish or perish culture, and produce truly consequential work. — source
Scientific and academic research though sometimes devalued is still the basis of our technology and all its applications, its malaise is everyone’s problem. if the heart doesn’t beat the other organs perish as well.
If you have found it interesting:
You can look for my other articles, you can also subscribe to get notified when I publish articles, and you can also connect or reach me on LinkedIn. Thanks for your support!
Here is the link to my GitHub repository, where I am planning to collect code and many resources related to machine learning, artificial intelligence, and more.
