Studying science quantitavely has often taken the form of studying publications, such as citation counts, or identifying author networks. But now, Samuel Arbesman and Nicholas Christakis (2011) argue that there are two fairly recent developments that would enable a new approach to the study of scientific discoveries:

1) Vast computational resources and storage capacity, and

2) Automated science.

This new appraoch would offer potential for a new field that concerns itself with the study of scientific discoveries. In the words of the authors:

These innovations offer the potential for a new type of scientometrics: quantitatively examining scientific discoveries themselves. This study of discoveries, rather than simply of scientific publications, offers the opportunity to understand science at a deeper level. We term this discovery-based approach to scientometrics as eurekometrics.

So, rather than focusing on the properties of scientific publications, this new field of eurekometrics would look at the properties of the discoveries themselves. This allows examination of both material (for example, size or complexity) and human (for example, cost or duration) properties of a discovery. Examples of this approach can be seen in figure 1. Based on these measurements, a crude proxy for difficulty of discovery could be developed.




Figure 1: A physical property of discovery over time. A) Mean diameter (in km) of minor planets discovered. B) Mean size (in g) of newly discovered mammalian species. C) Mean inverse atomic weight of newly discovered chemical elements.
(Source: Arbesman and Christakis, 2011)


The authors see great potential for this new field of eurekometrics, and mention three ways in which this potential could be implemented:


1) Large amounts of discovery data is being gathered. Findings, rather than publicatons, will constitute the primary output of this vast databases, and these can be studied through eurekometrics.


2) In areas such as automated drug discovery, automated chemical synthesis pathway discovery and other fields of automated science, a quantifiable analysis of what is discovered can elucidate some things, such as, for example, whether there is a relationship between human effort and the object of inquiry.


3) Citizen science projects (such as Galaxy Zoo, Foldit, and many more) do not only create huge amounts of data, but can also help in understanding how collaborative efforts may add to future discoveries, and which properties the discoveries most suitable for this type of research possess.


The authors, however, do recognize certain limits of eurekometrics. In their own words:

Despite the great strides in automated discovery and digitization of data that is currently occurring, however, there are limits to eurekometrics. The most important limitation is how to determine what constitutes a “discovery.” Quantifying what constitutes a discovery is never an easy proposition: Is each publication a discovery? Or do only certain ones rise to meet that definition? Furthermore, even if we can list discoveries, it needn't necessarily be possible to quantify their properties. For example, while it's possible to quantify the properties of minor planets and extrasolar planets, it is not nearly as easy to quantify the properties of methodological innovations made in computational fields.

Arbesman and Christakis (2011) conclude:

Scientometrics has for too long focused on understanding scientific progress at the level of the publication. Eurekometrics will allow us to understand the pace and determinants of scientific discovery in a way that simply examining the patterns in publications will not. For the first time, we will be able to explore how the properties of nature yield to human science.

Reference


Arbesman, S. and Christakis, N. (2011). Eurekometrics: Analyzing the Nature of Discovery. PLoS Computational Biology. 7(6): e1002072. doi:10.1371/journal.pcbi.1002072 (Click here)