Between 2003 and 2022, the number of published articles in AI grew faster relative to the number of articles in computer science,^{​See table SPBS-22 in National Science Board, National Science Foundation. 2023. Publications Output: U.S. Trends and International Comparisons. Science and Engineering Indicators 2024. NSB-2023-33. Available at https://ncses.nsf.gov/pubs/nsb202333.} due in part to the newness of the AI field compared with the more established field of computer science. AI articles worldwide grew by 1,100% during this period, reaching 123,402 articles in 2022,^{​See NSB-2023-33, table SPBS-99.} or 4% of all S&E publications globally,^{​See NSB-2023-33, figure PBS-3.} compared with 290% growth in computer science articles.^{​See NSB-2023-33, table SPBS-22.} From 2017 to 2022, the six countries with the highest overall publication outputs^{​See NSB-2023-33, figure PBS-3.} were also the countries with the highest AI research output (China, India, the United States, Japan, the United Kingdom, and Germany) (figure 1). In 2022, the top two producers of AI research articles were China (42,524 articles, or 35% of total AI publication output) and India (22,557, or 18%), followed by the United States (12,642, or 10%). Germany, Japan, and the United Kingdom published similar numbers of publications, ranging between 3,700 and 4,700 articles (3%–4%).

From 2017 to 2022, 37% of U.S. research papers on AI were the result of international collaboration, placing the United States between the five other top producers of AI research papers, with the United Kingdom (61%) and Germany (40%) producing a higher rate of internationally collaborative research and with Japan (25%), China (17%), and India (10%) producing a lower rate (figure 2). Rates of international collaboration for the United States were slightly lower for AI research papers than for all S&E research papers (37% versus 39%). Likewise, across the other five top producers of AI research papers, rates of international collaboration were lower for AI research papers than for all S&E research papers. Compared with other countries, China had the greatest proportion of AI papers that were domestic collaborations (41%). Across the six top-producing countries, the rate of articles produced by a single institution were more common in AI research than in all S&E research (42% versus 26%).

Overall, scientific research has become increasingly collaborative over time (Gazni, Sugimoto, and Didegah 2012; Wuchty, Jones, and Uzzi 2007). Although the rate of international collaboration in AI publications has been smaller than the rate of international collaboration across all S&E fields over the past 5 years, international collaboration in AI articles has gradually increased overall between 2003 and 2022. By country, international collaborations in AI increased in Japan (from 15% to 28%), the United States (from 24% to 39%), Germany (from 37% to 42%), and the United Kingdom (from 36% to 66%) (figure 3). Over this same time period, India and China did not show an increasing trend, despite some fluctuation. For example, after China exhibited a period of increased international collaboration in AI research, from 7% in 2009 to 23% in 2015, the rate has since decreased to 16% in 2022.

The proportion of single institution publications in AI decreased over time in the United States, from 48% in 2003 to 31% in 2022 (figure 4). Despite this decrease, the proportion of U.S. single institution publications remained higher in AI research than in all S&E research, which decreased from 36% to 20% over the same time period. Over time, the rate of domestic collaboration in AI between U.S. institutions remained relatively stable from 2003 to 2022, ranging between 25% and 30%. In China, the proportion of single institution publications in AI decreased from 59% to 38% between 2003 and 2022, albeit with more fluctuation. China’s proportion of single institution publications both in AI papers and among all S&E fields were similar until 2007, after which the proportion of single institution papers in AI research became higher, while the overall proportion of single institution papers in all S&E research continued to decrease.

In 2020, COVID-19 was identified as a national priority (https://www.whitehouse.gov/priorities/), and this shifting priority in research may have impacted collaboration patterns for this research area in 2020. In the same year, 35% of the United States’ published research on COVID-19 involved international collaborations, which was lower than the rates in the United Kingdom (55%), Germany (52%), and Japan (45%) but was higher than the rates in China (27%) and India (28%) (figure 5). The overall rates of international collaboration in the United Kingdom and Germany were higher for all S&E research than for COVID-19 research (65% and 55%, respectively).

Although each of the top producing countries had a lower rate of international collaborations in AI research than in S&E research, the results were mixed for COVID-19. As the number of articles in AI has increased, the rate of international collaboration also increased. For COVID-19 collaborations in 2020, only some of the top producing countries had lower rates of international collaboration in AI research than in all S&E research.

Publication data are derived from a large database of publication records that were developed for Science and Engineering Indicators 2024, Publications Output: U.S. Trends and International Comparisons (NSB-2023-33), from the Scopus database by Elsevier. The publication counts and coauthorship information presented are derived from information about research articles and conference papers (hereafter referred to collectively as articles) published in conference proceedings and peer-reviewed scientific and technical journals. Elsevier selects journals and conference proceedings for the Scopus database based on evaluation by an international group of subject-matter experts (see NSB-2023-33, Technical Appendix), and the National Center for Science and Engineering Statistics (NCSES) undertakes additional filtering of the Scopus data to ensure that the statistics presented in Science and Engineering Indicators measure original and high-quality research publications (Science-Metrix 2023). Although the listed affiliation is generally reflective of the locations where research was conducted, authors may have honorary affiliations, have moved, or have experienced other circumstances preventing their affiliations from being an exact corollary to the research environment.

The subset of AI articles was determined by All Science Journal Classification subject matter classification. Global coronavirus publication output data for 2020 were extracted from two different sources. The COVID-19 Open Research Dataset (CORD-19) was created through a partnership between the Office of Science and Technology Policy, the Allen Institute for Artificial Intelligence, the Chan Zuckerberg Initiative, Microsoft Research, Kaggle, and the National Library of Medicine at the National Institutes of Health, coordinated by Georgetown University’s Center for Security and Emerging Technology. CORD-19 is a highly inclusive, noncurated database. The other coronavirus publication output data source was the Scopus database, which permits more refined analysis because it includes more fields (e.g., instructional country of each author). (See NSB-2021-4, Technical Appendix).

Carvalho DS, Felipe LL, Albuquerque PC, Zicker F, Fonseca BDP. 2023. Leadership and International Collaboration on COVID-19 Research: Reducing the North–South Divide? Scientometrics 128:4689–705. Available at https://doi.org/10.1007/s11192-023-04754-x.

Chinchilla-Rodríguez Z, Sugimoto CR, Larivière V. 2019. Follow the Leader: On the Relationship between Leadership and Scholarly Impact in International Collaborations. PLOS ONE 14:e0218309. Available at https://doi.org/10.1371/journal.pone.0218309.

Gazni A, Sugimoto CR, Didegah F. 2012. Mapping World Scientific Collaboration: Authors, Institutions, and Countries. Journal of the American Society for Information Science and Technology 63:323–35. Available at https://doi.org/10.1002/asi.21688.

Katz JS, Hicks D. 1997. How Much Is a Collaboration Worth? A Calibrated Bibliometric Model. Scientometrics 40:541–54. Available at https://doi.org/10.1007/BF02459299.

Lee Y-N, Walsh JP, Wang J. 2015. Creativity in Scientific Teams: Unpacking Novelty and Impact. Research Policy 44:684–97. Available at https://doi.org/10.1016/j.respol.2014.10.007.

Leydesdorff L, Wagner CS. 2008. International Collaboration in Science and the Formation of a Core Group. Journal of Informetrics 2:317–25. Available at https://doi.org/10.1016/j.joi.2008.07.003.

Science-Metrix. 2023. Bibliometric Indicators for the Science and Engineering Indicators 2024. Technical Documentation. Available at https://science-metrix.com/bibliometrics-indicators-for-the-science-and-engineering-indicators-2024-technical-documentation/. Accessed 26 August 2023.

Wagner CS, Brahmakulam IT, Jackson BA, Wong A, Yoda T. 2001. Science and Technology Collaboration: Building Capacity in Developing Countries? Santa Monica, CA: RAND Corporation. Available at https://www.rand.org/pubs/monograph_reports/MR1357z0.html.

Wuchty S, Jones BF, Uzzi B. 2007. The Increasing Dominance of Teams in Production of Knowledge. Science 316:1036. Available at https://doi.org/10.1126/science.1136099.

Boothby C, Schneider B; National Center for Science and Engineering Statistics (NCSES). 2024. International Collaboration in Selected Critical and Emerging Fields: COVID-19 and Artificial Intelligence. NSF 24-323. Alexandria, VA: National Science Foundation. Available at https://ncses.nsf.gov/pubs/nsf24323.

Clara Boothby
ORISE Fellow
NCSES
E-mail: cboothby@nsf.gov

Benjamin Schneider
Interdisciplinary Science Analyst
NCSES
Tel: 703.292.8828
E-mail: beschnei@nsf.gov

National Center for Science and Engineering Statistics
Directorate for Social, Behavioral and Economic Sciences
National Science Foundation
2415 Eisenhower Avenue, Suite W14200
Alexandria, VA 22314
Tel: (703) 292-8780
FIRS: (800) 877-8339
TDD: (800) 281-8749
E-mail: ncsesweb@nsf.gov