Source-linked AI summary
International Research Collaboration: Novelty, Conventionality, and Atypicality in Knowledge Recombination
Caroline S. Wagner, Travis A. Whetsell, Satyam Mukherjee
TL;DR
The paper examines novelty and conventionality in international collaborative publications through an inductive measurement approach. Its analysis does not uphold the expectation that international collaboration produces more novel knowledge; instead, it indicates high conventionality and low novelty, with possible creativity losses associated with collaboration.
Problem
The paper addresses how international collaborative publications combine knowledge, focusing on measurement of novelty and conventionality rather than theoretical explanations.
Method
The paper applies measures of novelty and conventionality to international collaborative publications, including measures of conventional combinations defined by comparison with chance.
Results
International collaboration, in aggregate, indicates high conventional measures and low novelty measures for reference knowledge, while the negative finding concerns highly atypical referencing behavior.
Takeaways & Limitations
The findings suggest that international collaboration may involve losses of creativity and novelty, contrary to the expectation that collaboration produces more novel research.
Takeaways & Limitations
Distance collaborations may not allow implicit communication, and reliance on information-communications technologies may limit collaboration and lessen opportunities for novel outcomes.
Abstract
from arXiv · showhide
Research articles produced through international collaboration are more highly cited than other work, but are they also more novel? Using measures developed by Uzzi et al. (2013), and replicated by Boyack and Klavans (2014), this article tests for novelty and conventionality in international research collaboration. Scholars have found that coauthored articles are more novel and have suggested that diverse groups have a greater chance of producing creative work. As such, we expected to find that international collaboration tends to produce more novel research. Using data from Web of Science and Scopus in 2005, we failed to show that international collaboration tends to produce more novel articles. In fact, international collaboration appears to produce less novel and more conventional knowledge combinations. Transaction costs and communication barriers to international collaboration may suppress novelty. Higher citations to international work may be explained by an audience effect, where more authors from more countries results in greater access to a larger citing community. The findings are consistent with explanations of growth in international collaboration that posit a social dynamic of preferential attachment based upon reputation.
1 - Introduction
International collaboration is highly cited and expected to support novel knowledge combinations, but whether it actually produces more novel research remains unresolved. This study applies novelty and conventionality measures to international coauthored work and finds that the expectation is not supported.
- Internationally coauthored articles are more highly cited, more interdisciplinary, and associated with elite scholars than domestic or sole-authored work.
- The literature does not establish whether international research is more novel than domestic or sole-authored research.
- The study applies measures of novelty and conventionality to international collaborative publications, extending the Uzzi et al. approach with WoS and Scopus data.
- The authors expected international collaboration to produce atypical combinations because collaboration may combine diverse expertise and information.
- The analysis focuses on antecedents of novelty, conventionality, and atypicality, examining differences across broad fields and six disciplines.
2 – Literature Review
International collaboration has expanded and is associated with citation advantages, broader expertise, and potentially atypical knowledge combinations. Scholarship also treats novelty as a risky balance between disciplinary mastery and unusual recombination, whose impact may take time to recognize.
- 2.1 Growth of International Collaboration in Science: International collaboration has grown substantially, with participation expanding across nations and international coauthorship increasing over time.
- 2.1 Growth of International Collaboration in Science: Internationally coauthored articles generally receive more citations, and articles involving more countries are linked to higher citation rates.
- 2.1 Growth of International Collaboration in Science: International collaboration is expected to broaden information and expertise, giving heterogeneous teams greater opportunity to support knowledge creation.
- 2.2 Atypical combinations reflecting novel contributions: Originality requires both mastery of a field and new ideas, while novel cross-disciplinary combinations can be risky and slower to assimilate.
- 2.2 Atypical combinations reflecting novel contributions: Citation counts may reflect social recognition and communication as well as originality, and highly novel findings may require time for scholarly validation and dissemination.
3 - Data, Variables, and Methods
The study measures novelty and conventionality through citation-network reference combinations, then tests their relationship with international collaboration using WoS and Scopus data from 2005. It uses continuous, binary, and four-category regression models with author, reference, and field controls.
- The primary data combine roughly 850,000 articles from the Uzzi et al. study with WoS metadata and a roughly one-million-observation Scopus replication.
- The method compares observed journal-pair frequencies with randomized citation-network expectations to derive z-scores for each reference combination.
- The 10th percentile z-score measures article novelty, while the median z-score captures the central tendency of its journal combinations.
- A journal pair with a 10th percentile z-score below 0 is classified as a novel combination, while a median z-score above 0 indicates conventionality.
- The study binarizes novelty and conventionality, combining them into four categories: HN/HC, HN/LC, LN/HC, and LN/LC.
- OLS, logistic, and multinomial logistic regressions include controls for authors, references, and field fixed effects.
4 - Results
Across descriptive, regression, categorical, disciplinary, and robustness analyses, more countries per article are associated with lower novelty and higher conventionality, contrary to expectations. More authors show the opposite pattern, while discipline-specific results are mixed but generally consistent with the main findings.
- Descriptive results: All correlations are significant at p<0.0001, except the correlation between Novelty and Authors, which has p=0.1087.
- Descriptive results: The average novelty score decreases slightly as the number of countries grows, while the average conventionality score increases.Figures 5 and 6 illustrate these relationships using the mean tenth-percentile and median z-scores.
- Regression results: Against expectation, Countries is negatively associated with Novelty and positively associated with Conventionality, whereas Authors shows the opposite associations.These models use continuous z-scores and include intercept, Countries, Authors, References, and field fixed effects.
- Regression results: The binary logistic models replicate the main pattern: Countries is negatively associated with Novelty and positively associated with Conventionality, while Authors shows the opposite pattern.All estimates are significant at p<.0001, except Authors in model 1, where p=0.0063.
- Multinomial results: In the All Fields model, Countries is most strongly associated with Category 3 (LN/HC), while Authors is most strongly associated with Category 2 (HN/LC).The Sciences model shows a similar pattern; the Social Sciences and Arts and Humanities models differ in some coefficient signs and significance.
- Limitations: The Arts and Humanities analysis omitted field fixed effects because the model could not converge, and many articles lacked reference lists needed for the analysis.The authors note that results could vary by specific subfield.
- Disciplinary analyses: Across six scientific specialties, international collaboration is either positively associated with conventionality, negatively associated with novelty, or non-significant.Results were non-significant or supported the main Table 4 findings, with field-specific variation across specialties.
- Robustness checks: Robustness checks using additional years, alternative confounders, logarithmic transformations, and field-based scores produced results consistent with the main findings.Countries retained a positive sign for Conventionality and a negative sign for Novelty, with the opposite pattern for Authors.
5 - Discussion
Contrary to expectations, international collaboration is associated with more conventional and less novel reference combinations. The discussion considers communication and coordination barriers as possible constraints on novelty, while audience reach and reputation-based network dynamics may explain higher citation strength.
- International collaboration indicates high conventionality and low novelty in reference combinations.
- The analysis tests these patterns using Web of Science and Scopus source publications from 2005, plus additional Web of Science years.
- Higher transaction costs and coordination complexity may slow idea integration and suppress atypical knowledge pairings.Potential obstacles include time zones, travel, communication routines, and clashes between management systems.
- Distance, ICT-mediated communication, and language differences may limit tacit-knowledge exchange and creative discussion.ICTs favor codified knowledge, while tacit knowledge is described as important to scientific knowledge and innovation.
- Larger author teams are associated with higher novelty and lower conventionality, but this pattern does not extend straightforwardly to international teams.The discussion suggests international collaborations may use more hierarchical governance, with differences of opinion deferred to fewer leaders.
- An audience effect and preferential attachment may explain why international work receives more citations despite limited novelty.More international ties expand reach, while reputation rather than novelty may drive citation strength and collaboration growth.
6 - Policy Implications & Conclusions
The findings suggest that international collaboration may involve losses of creativity and novelty, despite its usefulness and varied research purposes. Policy responses include improving communication, limiting unsuitable large teams, and nurturing exploratory projects.
- Conclusions: International projects may involve losses of creativity and novelty.The conclusion qualifies this as an association rather than a universal outcome.
- Policy Implications: Some international collaborations require shared infrastructure, field access, data, or samples that no single country can provide.Examples include large-scale equipment, outbreak research, and projects requiring shared data or samples.
- Policy Implications: Collaborations pursued for audience effects or reputation enhancement may lack characteristics associated with highly creative work.These collaborations are contrasted with projects whose research requirements make collaboration necessary.
- Policy Implications: Frequent face-to-face meetings may facilitate richer communication in international projects.The paper suggests providing enough budget to support such meetings.
- Policy Implications: Very large teams, especially those not centered on large equipment, may be antithetical to the novelty/conventionality sweet spot.The paper links this concern to the communication conditions identified by Uzzi et al. (2013).
- Policy Implications: Exploratory research requiring substantial cross-disciplinary discussion may warrant alternatives to international collaboration.The paper also suggests explicitly nurturing the novelty/conventionality sweet spot in internationally collaborative projects.
Appendix - Robustness Checks & Tests
Robustness checks across regression specifications, years, variables, and field-adjusted measures generally reproduce the primary relationships between countries, authors, novelty, conventionality, and citations.
- Robustness Checks: The robustness analysis used logistic, multinomial logistic, OLS, fixed-effects, alternative-variable, log-transformed, and citation models.Checks covered Scopus data, multiple years, continents, languages, field and year controls, and nonlinear specifications.
- Robustness Checks: The primary results remained consistent across the robustness checks.The checks included alternative model forms, additional years, and different potentially confounding variables.
- Regression Results: The number of affiliated countries was negatively and significantly related to novelty and positively and significantly related to conventionality.This result was obtained in the Scopus logistic regression analysis.
- Regression Results: In the multinomial model, the number of countries was most strongly associated with Category 3 and negatively associated with Category 2.Category 4 served as the reference category.
- Across-Time Tests: Across years, countries were positively associated with conventionality and negatively associated with novelty, while authors showed the opposite pattern.The appendix reports these relationships as consistent with the primary results.
- Alternative Variables: Continents and languages were modeled separately because of high variance inflation between those variables.The language measure distinguished articles with no common languages from those including at least one common language.
- Alternative Specifications: Log-transformed country counts remained positively related to conventionality and negatively related to novelty, whereas log-transformed author counts showed the opposite.These results were reported as consistent with the primary findings.
- Field-Adjusted Analysis: Field-based median z-scores produced similar multinomial results, with Category 2 remaining negative and significant and Category 3 positive and significant.The Category 1 country effect became positive but nonsignificant under this specification.