Scientific software as geopolitics: evidence of national attribution bias from a large-scale replication in bibliometrics

This study examines how seemingly neutral bibliometric software performs geopolitical work by distorting national research metrics. Using a preregistered replication of 44 bibliometric studies, this research reveals how an operational assumption about what counts as a national entity, implemented in just two lines of code in a widely used open-source package, removed Taiwanese scientific contributions in 84% of the replicated studies while simultaneously inflating the academic output assigned to China. Evidence suggests the majority of the software's users were unaware of this reclassification, which is estimated to affect nearly 2,000 studies cited over 15,800 times. The results demonstrate that neither high user skill nor algorithmic transparency is sufficient to prevent bias, highlighting how software infrastructure can become a powerful, yet often invisible, actor in international disputes. The findings underscore the need for greater scrutiny of algorithms, not merely for technical accuracy, but to reveal how they can compel researchers to unknowingly adopt and amplify specific geopolitical stances.