Skip to main content
Log in

The generation mechanism of research leadership in international collaboration based on GERGM: a case from the field of artificial intelligence

  • Published:
Scientometrics Aims and scope Submit manuscript

Abstract

Conducting an in-depth analysis of 235,746 research papers in the field of artificial intelligence spanning from 2001 to 2020, this study quantified the extent of research leadership in international collaborations by discerning the country of the corresponding author. To comprehensively investigate both endogenous and exogenous effects, we employed the Generalized Exponential Random Graph Model, an advanced methodology adept at characterizing network structures with real-valued edges. This research elucidates the pivotal role of intrinsic structural factors influenced by edge dependencies and evaluates their impact on research leadership in international collaborations. Specifically, our findings reveal a positive and significant effect of the mutual effect and the transitivity effect. Furthermore, language and geography no longer play a significant role in generating international research collaborations between two countries. Additionally, scientific productivity also holds an important position in generating research leadership. However, R&D expenditures no longer facilitate the establishment of leadership for international research collaboration.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save

Springer+ Basic
$34.99 /Month
  • Get 10 units per month
  • Download Article/Chapter or eBook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. The CEPII (founded in 1978) is the leading French center for research and expertise on the world economy. To conduct empirical work on the world economy, CEPII has developed and maintains a database that is accessible to the academic community and the public on its website. This database includes macroeconomic data, production and specialization indicators, international trade data, and country geographical and language distance data. For more detailed information, please visit this website: http://www.cepii.fr/CEPII/en/bdd_modele/bdd_modele.asp.

  2. MCMC MLE is a statistical technique used for estimating the parameters of complex models, especially when traditional maximum likelihood estimation is impractical due to computational limitations. It combines the concepts of Markov chains and Monte Carlo simulations to approximate the likelihood function, enabling efficient parameter estimation in models with intricate dependencies. This method is pivotal in our study to accurately estimate the parameters of GERGM, ensuring the validity of the results.

References

Download references

Acknowledgements

The present study is an extended version of a paper presented at the 19th International Conference on Scientometrics and Informetrics 2023 (ISSI 2023), Bloomington, Indiana (USA), 2-5 July 2023 (Cai et al., 2023). This study is partially supported by the Major Projects of National Social Science Foundation of China (22&ZD194), the National Natural Science Foundation of China (71974030), and the LiaoNing Revitalization Talents Program (XLYC2007149). Wencan Tian is financially supported by the China Scholarship Council (202106060134). The authors are grateful to the anonymous reviewers for their helpful comments and suggestions.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Zhigang Hu.

Ethics declarations

Competing interests

The authors declare that they have no conflicts of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix

Appendix

GERGM diagnosis

To evaluate the fit goodness of the estimated model, nine typical network structure attributes were chosen from the simulated network and compared with the observed network (see Fig. 3). To facilitate comparative analysis, the statistics displayed in Fig. 3 have been normalized (i.e., by subtracting the expected value of each statistic in the random network model and then dividing by the standard deviation). The findings indicate that the simulated network closely corresponds to the observed network, and the model exhibits a satisfactory fit.

Fig. 3
figure 3

Goodness of fit plots for the model fitted to the research leadership in international collaboration networks

To ascertain the non-degeneracy of GERGM, we conducted a hysteresis analysis as proposed by Snijders (2006). The result (see Fig. 4) reveals that the average density of the simulated network during the periods of 2001–2010 and 2011–2020 exhibits a smooth, upward trend of connectivity, thereby indicating that the GERGM estimates are not compromised.

Fig. 4
figure 4

Hysteresis plots for the simulated network

Robustness test

We conducted a robustness test to validate the reliability and consistency of our primary findings obtained through the GERGM. The necessity for this robustness test arises from the inherent complexity and potential variability in modeling international scientific research collaboration networks. To ensure that our results were not artifacts of the specific modeling choices inherent to GERGM, we employed the ERGM as an alternative approach. ERGM was chosen due to its ability to model network data with fewer constraints compared to GERGM, particularly in handling unweighted network ties, thus providing a complementary perspective. Notably, ERGM identifies social networks with binary (0–1) variables only; therefore, the directed and weighted research leadership network of international collaboration was converted into a directed network with binary values only. Moreover, to replace the R&D expenditures, some explanatory variables were substituted with the node’s in-degree and out-degree. Additionally, we also used MCMC MLE to parametrically test ERGM (the result can be seen in Table 4).

Table 4 Results of ERGM estimates

The mutuality coefficients were estimated to be 1.7 and 1.58 for 2001–2010 and 2011–2020, respectively, and both were statistically significant at the 0.1% level. The estimated coefficients of the ln_paper_sender for 2001–2010 and 2011–2020 were 0.01 (not statistically significant) and 0.18 (statistically significant at the 5% level), respectively. Furthermore, the estimated coefficients of ln_paper_receiver for 2001–2010 and 2011–2020 were 0.19 (statistically significant at the 5% level) and 0.46 (statistically significant at the 0.1% level), respectively. These results indicate that the estimates of mutual effect and level of research productivity from ERGM align with those of GERGM.

The estimates for 2001–2010 and 2011–2020 indicate that the coefficients of degree_sender are 0.92 and 0.69, respectively, both of which are significant at the 0.1% level. Similarly, the coefficients of degree_receiver are 0.68 and 0.27 for 2001–2010 and 2011–2020, respectively, with significant levels of 0.1% and 1%. These results demonstrate that a higher number of countries engaged in research partnerships increases the likelihood of establishing and leading research collaborations. Additionally, the analysis confirms the prevalence of the “Matthew effect”, which means that countries with a more extensive history of international research collaborations are more likely to lead and be led in research partnerships. Conversely, the estimated coefficients for edgecov_lan are insignificant for both time periods. Furthermore, the coefficients of geographical distance are not statistically significant and are thus removed from the ERGM model to improve estimation accuracy. This indicates that it no longer significantly influences the explanatory variable of research leadership in international collaboration.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Cai, R., Tian, W., Luo, R. et al. The generation mechanism of research leadership in international collaboration based on GERGM: a case from the field of artificial intelligence. Scientometrics 129, 5821–5839 (2024). https://doi.org/10.1007/s11192-024-04974-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11192-024-04974-9

Keywords

Navigation