Quantifying the value of basic science

Sept. 7, 2017, 3:04 a.m.

Maggie-Lam

In this era of constrained research funding, the value of basic research to the public is sometimes debated. Arguments are often advanced anecdotally: on the one hand well known example on the unpredictable later applicability of basic research abound, from CRISPR to modern innovations on smart phones, which built upon research in the 60s on transtistors which built upon quantum mechanisms from 100 years ago.

On the other hand, the value of basic research is doubted by a spectrum of individuals from the public to well meaning medical professionals who lament that not more research is oriented towards direct translation into health care. Despite strong arguments on both sides, actual data on the value of basic research have been scarce. A recent paper by Ahmadpoor and Jones provided a data science driven view into the value of basic research in terms of their connectivity to patented invention.

Ahmadpoor and Jones queried the citation of 4.8 million patented innovations issued by the US Patent and Trademarks Office (USPTO) and citation network within 32 millions archived within the Web of Science articles. A patent directly citing a journal article and the journal article they cited are both given a citation distance of 1. Articles citing the article are given a distance of 2, and so on, in the construction of a directed citation graph. The distances of journals and patents from the paper-patent interface are then reported.

Remarkably, up to 80% of published papers eventually connect to a patented invention, which is a strong testament to the value of basic research. Moreover, the modal distribution of research linkage to a patent is 3, suggesting most research may only lead to practical and patentable invention after rounds of refinement and continued research. Moreover there is often a substantial temporal delay between article publications nd patent citation, ranging from an average of 7 years for a directly cited article (distance of 1) to 20 years for a more distantly connected paper (distance of 6).

As expected, there is a huge discrepancy in the citation distance between fields, with computer science and nanotechnology research articles often directly linked to a patent, followed by biomedical sciences (average distance of 2-3), then more “basic” fields such as astronomy (avg. distance of 4) and mathematics (avg. distance of 5).

On the other hand, patents that can be categorized as multicellular organisms such as biomedical research are often directly linked to published article. However, some perhaps unexpected categories such as “chairs and seats” and “pipes and conduits" are also ultimately connected to basic scientific research articles.

There are some limitations with the study, e.g., only patents filed in the US are queried where patent law differs from rest of the world in prior disclosure requirements. The authors also claim that paper impact decays smoothly as they become more disconnected from patents, but it is unclear how much of this may be influenced by network topology and influence of review articles, which anecdotally may favor citation of already highly cited papers.

Article link: Science 2017 (357) 6351.

Image Credit - chuttersnap