Source-linked AI summary
A Century of Science: Globalization of Scientific Collaborations, Citations, and Innovations
Yuxiao Dong, Hao Ma, Zhihong Shen, Kuansan Wang
TL;DR
How science advances itself has been little understood. This study analyzes publications and their collaboration and citation records across 1900–2015, finding increasingly collaborative, visionary, and geographically diverse science.
Problem
How science advances itself is poorly understood, despite scientific publications recording that development over time.
Source: §1 · p. 1Method
Results
Science became more collaborative, visionary, and geographically diverse; over 90% of 2000s world-leading innovations came from collaborations, nearly four times the 1900s share.
Source: §3.4 · p. 5, §4.3 · p. 7, §5.4 · p. 9Takeaways & Limitations
The findings are presented as relevant to research funding policy, including encouragement and support for international collaborative projects.
Source: §1 · p. 2The study aggregates all fields; future work could examine disciplines and topics separately, as well as publication venues and links between economic development and scientific advances.
Source: §7 · p. 9
Abstract
from arXiv · showhide
Progress in science has advanced the development of human society across history, with dramatic revolutions shaped by information theory, genetic cloning, and artificial intelligence, among the many scientific achievements produced in the 20th century. However, the way that science advances itself is much less well-understood. In this work, we study the evolution of scientific development over the past century by presenting an anatomy of 89 million digitalized papers published between 1900 and 2015. We find that science has benefited from the shift from individual work to collaborative effort, with over 90% of the world-leading innovations generated by collaborations in this century, nearly four times higher than they were in the 1900s. We discover that rather than the frequent myopic- and self-referencing that was common in the early 20th century, modern scientists instead tend to look for literature further back and farther around. Finally, we also observe the globalization of scientific development from 1900 to 2015, including 25-fold and 7-fold increases in international collaborations and citations, respectively, as well as a dramatic decline in the dominant accumulation of citations by the US, the UK, and Germany, from ~95% to ~50% over the same period. Our discoveries are meant to serve as a starter for exploring the visionary ways in which science has developed throughout the past century, generating insight into and an impact upon the current scientific innovations and funding policies.
1 INTRODUCTION
The paper addresses limited knowledge about how science advances itself and uses digitized publications to study its evolution over the past century. It reports growing collaboration, broader referencing, and geographic diversification in scientific development.
1 INTRODUCTION
- Scientific publication volume doubled every 12 years from 1900 to 2015, while individual scientist productivity remained surprisingly constant.Source: §1 · p. 2
- In the 2000s, teams generated over 90% of world-leading innovations, nearly four times the 1900s share, as international collaboration increased 25-fold.The paper measures world-leading innovations as the top 1% most-cited papers.Source: §1 · p. 2
- Individual-author self-citations fell from 30% to 10%, while country self-citations declined from 90% to 30%.Scientists also increasingly cited older literature.Source: §1 · p. 2
- Scientific development diversified geographically after the US, UK, and Germany dominated early-century output, with advancement later emerging from countries in varied regions.The paper describes the US as the solitary leader for a period after World War II.Source: §1 · p. 2
- The authors suggest these findings can inform research funding policies, including support for international collaborative projects.Source: §1 · p. 2
2 DATA: MICROSOFT ACADEMIC GRAPH
The study draws on the Microsoft Academic Graph to examine collaboration, citations, and impact across 1900–2015. It analyzes a large connected author network and uses affiliation locations for geographic analyses.
2 DATA: MICROSOFT ACADEMIC GRAPH
- The dataset contains 89,559,406 papers, 795,032,662 citation links, and 1,236,796,098 author collaboration links from 53,326,061 authors.The study focuses on the largest connected component of the collaboration graph.Source: §2 · p. 2
- The authors state that their results hold whether or not authors without collaborators over their careers are included.Source: §2 · p. 2
- Geographic analyses use 21,695,143 papers whose authors’ affiliations could be associated with locations.An institution’s headquarters location is used when it has multiple locations.Source: §2 · pp. 2–3
- The analysis examines scientific collaborations, citations, and impact, with its questions summarized in Table 1.The table is described as a summary of questions answered from the data.Source: §2 · p. 3, Table 1 · p. 3
3 THE EVOLUTION OF COLLABORATION
From 1900 to 2015, scientific output grew while research shifted toward larger and more international collaborations. The section traces that shift across publication growth, researcher careers, and collaboration patterns around the world.
3.1 Te Growth of Science
- Publications doubled every 12 years from 1900 to 2015, while the number of publishing scholars doubled every 11 years; growth was interrupted around both world wars.The reported interruptions occurred around 1915 and 1940.Source: §3.1 · p. 3
3.2 Collaboration and Productivity
- Single-author papers fell from 80% in the 1900s to 20% in the 1990s, while individual productivity declined 50% between the 1900s and 2000s.Average yearly output remained roughly two papers per author; collaboration teams with hundreds or thousands of members began forming from 1975, largely enabled by electronic communications.Source: §3.2 · p. 3
- In 2015, single-author papers made up 15% of publications but 5% of top papers, whereas papers with over 10 authors had a relative rate of 4.Top papers are defined as the top 1% most-cited papers; the paper reports the rate as 20% / 5% for papers with over 10 authors.Source: Figure 2 · p. 3, §3.2 · p. 4
- Average collaboration duration remained 1.5 years and its maximum 3.5 years across generations, even as average career length declined; 40–50% of new scientists never published again after their first year.The paper defines career start and end by the earliest and latest publications attributed to a scientist in its records.Source: §3.2 · p. 4
- Scientists averaged 10 co-authored papers over a career, while average collaborators rose from 10 in the first half of the century to over 60 for careers starting in the 1980s.The late-1990s decline in collaborator counts reflects careers not yet completed by the dataset’s 2015 endpoint.Source: §3.2 · p. 4
3.3 Collaborations Across the World
- International collaborations rose from less than 1% of collaborations in 1900–1924 to 20% of publications in this century, with more countries joining the international stage.Early international links were dominated by the US and UK; by the 21st century the UK, Germany, and France had higher international collaboration rates than the global average.Source: Figure 3 · p. 4, §3.3 · p. 5
3.4 Summary
- Over 90% of world-leading innovations were generated by teams in the 2000s, nearly four times the 1900s share, while international collaborations increased 25-fold.The paper measures world-leading innovations as the top 1% most-cited papers.Source: §3.4 · p. 5
4 THE EVOLUTION OF CITATION
Citation behavior shifted toward older, less self-referential literature, while cross-border citation flows expanded and citation collection became less concentrated among historically dominant countries.
4.1 Reference Characterization
- From 1900 to 2015, reference lists grew from two to three papers on average to 30, while their growth was superlinear.The authors suggest the increase partly reflects the accumulation of available literature and possibly the greater complexity and interdisciplinarity of modern science.Source: §4.1 · p. 6
4.2 Citing Across Countries
- The US’s citation lead grew after World War II, while Japan, France, China, and many other countries gained citation participation later in the century.China and 100-plus other countries contributed sizeable reference shares but collected noticeably smaller citation shares.Source: §4.2 · pp. 6–7
- The US collected about twice as many citations as it sent out, while Russia’s receive-to-send rate rose and the UK’s declined over time.The US and UK remained above the equal-receive-and-send line across five periods, while Russia remained below it.Source: §4.2 · p. 7
- Cross-country citations rose from 10% in the first quarter of the 20th century to 70% in this century.Early in the century, the US, Germany, and the UK created 95% and collected 97% of citations; by the 21st century, those shares had fallen to 46% and 58%, respectively.Source: Figure 6 · p. 6, §4.2 · p. 7
4.3 Summary
- Reference age doubled and self-citations declined as science moved toward older literature and citing others’ work.The average reference age rose from about 6 years to 11 years after 1975; self-citations fell from over 30% in the 1900s to 10–15% in the 2000s.Source: §4.1 · p. 6, Figure 5 · p. 6, §4.3 · p. 7
5 THE EVOLUTION OF IMPACT
Citation impact increased over time, while leading research institutions and award recipients became geographically more diverse, though the US remained prominent.
5.1 Inflation, Dominance of Top Research
- Papers published before 1950 averaged fewer than two citations within 20 years, compared with four for papers before 1975 and ten for those before 1995.The authors caution that citation inflation affects comparisons of publications from different periods and scientists at different career stages.Source: §5.1 · p. 7
- Since mid-century, top 1% papers shifted from collecting more citations quickly to collecting a larger share over longer periods.Overall, about 25% of citations received by papers published in the same year went to that year’s top 1%; polarization was sharpest in the 1900s at 40–50%.Source: §5.1 · pp. 7–8
5.2 Distribution of Top Institutions
- The top-200 landscape diversified over the century, with China, South Korea, Singapore, Brazil, and additional European locations joining; meanwhile, the citation-count gap between the top and 200th institutions declined across five periods.Source: §5.2 · p. 8
5.3 Distinguished Award Winners
- The US dominated the Nobel science prizes after World War II and led the Economics Nobel, Fields Medal, and Turing Award histories described through 2016.For the Turing Award, 50 of 64 recipients from 1966–2015 were from the US.Source: Figure 8 · p. 8, §5.3 · p. 9
5.4 Summary
- By the 21st century, 40% of the top 200 most-cited research institutions were outside the US, UK, and Germany, up from 4% in the early 20th century.The authors describe a shift from early dominance by the Northeastern US, the UK, and Germany, through shared leadership with Continental Europe, to the rise of Asia and other regions.Source: §5.2 · p. 8, §5.4 · p. 9
6 RELATED WORK
Prior science-of-science research addressed impact metrics, prediction, characterization, and network analyses; this study distinguishes itself through its century-scale, all-fields geographic scope.
Prior science-of-science research
- Science-of-science research has developed metrics for impact and studied citation prediction, impact distributions, author credit, and collaboration and citation networks.Examples include impact factor, h-index, citation-count prediction, and network analyses of collaboration, citation, and team organization.Source: §6 · p. 9
Scholarly systems
- Online scholarly systems such as AMiner, Cite-SeerX, Google Scholar, and Microsoft Academic Services were developed to serve researchers.Source: §6 · p. 9
Distinction from related work
- Unlike studies focused on static or short-term patterns, this work examines all scientific fields over 1900–2015 and adds a planetary-scale geographic perspective.The authors contrast their work with a Physics study and studies of geographically conditioned institutional links using papers from 2003–2010 and 1996–2013.Source: §6 · p. 9
7 CONCLUSION
The paper presents a planetary-scale account of scientific development from 1900 through 2015, finding greater collaboration, deeper referencing, and geographic diversification.
7 CONCLUSION
- From the first quarter of the 20th century onward, science increasingly involved collaboration, drew on older literature, and diversified geographically.The average number of collaborators per scientist increased at least fivefold, cited-literature age rose 90–100%, and scientific leadership became more geographically diverse.Source: §7 · p. 9
- The authors suggest that knowledge about science’s evolution may help scientists improve the discipline and contribute to a better world.Source: §7 · p. 9
- The analysis aggregates all fields, leaving discipline- and topic-specific leadership, publication-venue evolution, and links between economic development and science for future study.Source: §7 · p. 9