Scientific findings are always based on previous studies. But sometimes a discovery is so groundbreaking that it opens up entirely new avenues of research. Such disruptive work drives scientific progress the most. Identifying them can therefore help create optimal conditions for further breakthroughs. But how do you measure innovative strength? Researchers have now developed a new approach.
The more important a study is, the more often it is cited by others. Following this logic, citations typically serve as a measure of the impact of a research paper. But sometimes this method falls short. If only direct citations are included, it cannot be reflected if a study has opened up a completely new field of research in which the original study is no longer necessarily mentioned in every publication based on it. “Scientific influence spreads through both direct and indirect citation pathways, and discoveries are often scattered across multiple articles,” explains a team led by Munjung Kim from Indiana University in Bloomington. “That makes it difficult to measure disruptiveness.”
Comparison of past and future
Together with her colleagues, Kim has now developed a new approach to classify the impact of scientific discoveries. To do this, the researchers classified more than 55 million studies and patents into a high-dimensional grid that uses artificial intelligence to record all direct and indirect connections to further studies and shows which previous publications a work is based on and which new research work it has inspired.
“The core idea of our approach is that we can imagine two different vectors for any work: one for the past and one for the future,” explain Kim and her colleagues. “When a contribution significantly reshapes the connection between past and future or initiates a new line of research, these two contexts diverge.” The greater the distance between the past and future vectors, the greater the changes the respective work has triggered. “Our measure – the Embedding Disruptiveness Measure (EDM) – provides continuous, high-resolution insight into how scientific contributions are reshaping the relationship between traditional knowledge and new research directions,” say the researchers.
Scientific influence made measurable
To test their new measure, the researchers checked where numerous Nobel Prize-winning research and other outstanding works were classified. In fact, the EDM identified the relevant contributions more reliably than previous methods. This was particularly true for discoveries made independently by multiple research groups at the same time – a case in which other metrics often fail.
An example of such a simultaneous discovery is the Higgs mechanism, which explains how elementary particles gain their mass. It was discovered both by the eponymous physicist Peter Higgs and, independently of him, by François Englert and Robert Brout. Because Higgs cites these two in his paper, common disruption metrics incorrectly rate his paper as uninnovative. The EDM, on the other hand, correctly indicates that Higgs, Englert and Brout together initiated a scientific breakthrough.
“By more robustly identifying disruptive innovations and concurrent discoveries, our method enables more precise attribution of transformative contributions while providing insights into the mechanisms that drive scientific breakthroughs,” the research team writes. From the perspective of co-author Sadamori Kojaku from Binghamton University in New York, this could have significant implications for science policy. If it is known under which circumstances scientific breakthroughs have taken place in the past, appropriate conditions can be specifically created in the future to promote further groundbreaking research. “These insights are helpful for prioritizing funding,” he says. “We now have the quantitative metrics to examine in which research phase disruptive work occurs and is most important.”
Source: Munjung Kim (Indiana University, Bloomington, USA) et al., Science Advances; doi: 10.1126/sciadv.adx3420