A new analysis of international computing conferences shows that the field's most prestigious gatherings have been slow to adapt to Asia's rising research prominence, creating persistent imbalances in who gets to present work and who gets to decide what gets published. Researchers examined thirteen major systems research conferences over thirteen years, finding that while most conferences gradually increased Asian participation throughout the past decade, a group of top-tier conferences only began reflecting this shift in the last four years. This delay highlights how established power structures in academic computing may be resisting the broader geopolitical and scientific transformations reshaping global research.
The study reveals that conferences like NSDI, ASPLOS, OSDI, and SIGCOMM historically maintained American participation at around 80% for much of the past decade, only dropping to about 50% in recent years as Asian contributions increased. In contrast, other conferences such as ICDCS, ATC, EuroSys, IEEE Cloud, and CCGRID showed more gradual adaptation to Asian growth throughout the period. The most dramatic shift occurred at USENIX ATC, where Asian papers rose from nearly 5% in 2012 to 63% in 2024, while American papers fell from 78% to 27% over the same period. These changes indicate that the computing research landscape is evolving from a two-block system dominated by America and Europe toward a more multipolar world with three major hubs.
To conduct their analysis, the researchers created a crawler that gathered data from conference websites and three major research resources: DBLP, Semantic Scholar, and OpenAlex. They collected information on 9,712 accepted papers and 14,996 program committee members across thirteen conferences from 2012 to 2024. The team mapped papers and committee members to continents based on author affiliations, using a heuristic that assigned entities to the continent where most authors were located. When data was incomplete, they manually retrieved information from official conference websites to ensure accuracy, though approximately 5% of papers had ambiguous assignments that didn't significantly affect overall trends.
, Visualized in Figures 1 through 4, show clear patterns in geographical representation. Figure 1 illustrates the variability in Asian paper proportions over time, with most conferences showing increasing Asian participation during the past decade. Figure 2 compares continent diversity in accepted papers between 2012-2019 and 2020-2024, revealing substantial changes in a relatively short period. The researchers used the Gini-Simpson index to quantify diversity, finding that conferences like CCGRID, IEEE Cloud, and ICDCS had greater diversity in accepted papers (with indices of 0.658, 0.654, and 0.621 respectively) compared to NSDI, ASPLOS, and SIGCOMM (0.405, 0.467, and 0.484). However, by 2024, these indices had become more closely aligned, suggesting recent convergence.
Perhaps more concerning is the persistent gap between who publishes papers and who serves on program committees. Figure 3 shows that program committee diversity hasn't changed as substantially as paper diversity, with Asian researchers consistently underrepresented. Figure 4 quantifies this gap for 2020-2024, revealing that North America remains over-represented in program committees across nearly all venues, with the strongest positive gaps in ATC (+31.9 percentage points), OSDI (+21.8 pp), and ASPLOS (+14.6 pp). Meanwhile, Asia is systematically under-represented, with the most pronounced gaps in ATC (-38.0 pp), ICDCS (-21.2 pp), and Euro-Par (-18.8 pp). This imbalance means decision-making and peer-review power remain concentrated in North American institutions despite Asia's growing publication output.
The study also found correlations between geographical diversity and corporate involvement. As shown in Figure 5, conferences with higher percentages of U.S.-based authors tend to have greater shares of publications from major technology companies like IBM, Microsoft, Google, and Amazon. These include NSDI, SIGCOMM, ATC, and OSDI, where around 40% of papers come from Big Tech companies. In contrast, conferences with more international character, such as ICDCS, IEEE Cloud, and CCGRID, publish papers primarily from academic institutions. The researchers hypothesize that the intense competition created by industry participation in top-tier conferences may make it harder for academic researchers worldwide to participate on equal footing, though they note this correlation doesn't confirm causation.
Several limitations affect the study's conclusions. The assignment of papers and committee members to continents relied on a heuristic that could misclassify some entities, particularly those with authors from multiple continents or missing affiliation data. While the researchers manually vetted data consistency using randomized samples, approximately 5% of papers were assigned despite most authors lacking affiliation data, and no conference had more than 5% of papers associated with multiple continents. Additionally, the analysis doesn't account for individual researcher mobility or collaborations that might blur geographical boundaries, and it focuses only on systems research conferences, leaving open whether similar patterns exist in other computing subfields.
Of these extend beyond academic fairness to the very quality and direction of computing research. As technological innovation in AI and cloud computing becomes increasingly geopolitically significant, maintaining diverse perspectives in research venues is crucial for fostering innovation that serves global needs. The study's public website and open datasets allow the community to explore these trends further, potentially encouraging conference organizers to reflect on their practices. With USENIX ATC announcing its conclusion in 2025 after dramatic shifts in participation, the computing research community faces important questions about how to balance quality standards with equitable global participation in an increasingly multipolar world.
Original Source
Read the complete research paper
About the Author
Guilherme A.
Former dentist (MD) from Brazil, 41 years old, husband, and AI enthusiast. In 2020, he transitioned from a decade-long career in dentistry to pursue his passion for technology, entrepreneurship, and helping others grow.
Connect on LinkedIn