How Science Thinks: Dynamic Hypergraph Models of Science's Unfolding Structure

• In Event: 361 - Section on Methodology Paper Session. Open Topic I

Abstract

Science can be viewed as a complex system. It is apparently complicated; shaped by strong interactions between diverse components; and displays emergent, often unexpected collective outcomes. This characterization of science is strikingly similar to the “actor-network” through which Bruno Latour has explored the politics of things, but networks provide a fruitful metaphor for other processes in science. Here, we explore how the complex network of science provides a substrate on which a scientist–and indeed science as a whole–thinks--whether linkages between scientists, topics, and methods inscribed by scientific articles can predict new ties in the network. Put another way: Does the structure of past research allow us to predict future research? Because experiments do not represent singular connections, our approach models science as a dynamic hypergraph. Using millions of scientific articles in MEDLINE, we validate this approach and show how science moves from problems posed and questions answered in one year to those examined in the next. Along the way, we discovery how science is “changing its mind”—we identify a historical shift in unfolding of biomedical science in the 1960s, before which, authors were the primary paths through which new things of all types became linked. In the 1960s, other things–chemicals, diseases, methods–became linked with sufficient density in the hypergraph that they became the primary reference structure through which discovery was organized. We also discovery the special role the methods play in linking authors and topics in biomedical science.

Authors

• James A. Evans, University of Chicago

• Jacob Gates Foster, University of California-Los Angeles

• Feng Shi, University of Chicago