arxiv: 2604.17602 · v1 · submitted 2026-04-19 · ⚛️ physics.soc-ph · cs.AI· physics.app-ph

Recognition: unknown

Polarization and Integration in Global AI Research

Luca Gallo , Riccardo Di Clemente , Bal\'azs Lengyel

Authors on Pith no claims yet

Pith reviewed 2026-05-10 05:02 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cs.AIphysics.app-ph

keywords AI researchinternational collaborationpolarizationcitation networksUS-Chinascientific integrationglobal innovationresearch networks

0 comments

The pith

US and China have diverged in AI collaboration and citations to form two distinct poles that organize global research.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper analyzes three decades of AI publication data to track how countries collaborate and cite one another. It shows the United States and China diverged early and now act as separate poles, with the United Kingdom and Germany linking only to the US, many European nations linking to both, and developing countries linking mainly to China. A reader would care because AI drives security concerns and regulatory efforts, so these network patterns determine how knowledge and influence actually spread across borders. The findings directly address whether strategies like US mobility restrictions or Chinese global governance initiatives can succeed given the observed alignments.

Core claim

By comparing observed cross-country collaboration and citation links in AI publications to their random realizations, the authors find that the US and China diverged long ago in both dimensions and now form two poles. The United Kingdom and Germany integrate exclusively with the US pole, many European countries converge with both poles, and developing plus further developed countries integrate only with China, which signals its expanding influence over the international AI research landscape.

What carries the argument

Comparison of real cross-country collaboration and citation networks against randomized null models to quantify polarization and integration.

If this is right

The global AI research system revolves around two poles rather than a single integrated network.
UK and Germany alignments reinforce separation from China, while many European countries maintain dual ties.
Developing countries' exclusive integration with China indicates a shift in its international reach.
National policies on knowledge flows and global regulatory efforts must account for these specific alignment patterns.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Persistent two-pole structure could reduce cross-pollination between US-led and China-led AI advances.
Countries with dual ties may act as bridges for standards or safety discussions that single-pole nations cannot.
The patterns imply that US restrictions on mobility may accelerate rather than slow the separation of research communities.

Load-bearing premise

That comparing observed collaboration and citation links to random realizations accurately measures polarization and integration, and that large-scale publication data fully and unbiasedly represents global AI research activity.

What would settle it

New publication or patent data from recent years showing that developing countries now form substantial collaboration or citation links with the US at levels comparable to China, or that US-China divergence has narrowed rather than persisted.

read the original abstract

The AI race amplifies security risks and international tensions. While the US restricts mobility and knowledge flows, challenges regulatory efforts to protect its advantage, China leads initiatives of global governance. Both strategies depend on cross-country relationships in AI innovation; yet, how this system evolves is unclear. Here, we measure the processes of polarization and integration in the global AI research over three decades by using large-scale data of scientific publications. Comparing cross-country collaboration and citation links to their random realizations, we find that the US and China have long diverged in both dimensions, forming two poles around which global AI research increasingly revolves. While the United Kingdom and Germany have integrated exclusively with the US, many European countries have converged with both poles. Developing and further developed countries, however, only integrate with China, signaling its expanding influence over the international AI research landscape. Our results inform national science policies and efforts toward global AI regulations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The abstract maps US-China polarization in AI research networks but the random null model is underspecified and may just capture publication volume differences.

read the letter

The paper's main claim is that over three decades US and China have become distinct poles in AI collaboration and citation networks, with the UK and Germany linking only to the US, other European countries to both, and developing nations mainly to China. That pattern, if it survives proper controls, is the useful descriptive result here. It applies standard network comparison techniques to a long publication time series in a domain that matters for current policy debates on research flows and security. The time span and the specific country groupings for AI are not already in the cited literature, so the measurements add something concrete even if the underlying idea of comparing observed links to random ones is not novel. The work stays empirical and avoids self-referential modeling, which keeps it straightforward. The abstract is clear on the headline patterns and their policy angle. The soft spot is exactly the one flagged in the stress test. The abstract only says links are compared to random realizations without saying whether those realizations preserve each country's total AI output or total degree. If they do not, then the US and China will automatically show more cross-links simply because they produce far more papers; that mechanical effect gets labeled as long-term divergence. Without a configuration-model null or equivalent, the reported splits for Europe and developing countries could be artifacts of size heterogeneity rather than genuine integration or polarization processes. The abstract also gives no details on the exact data source, how divergence is quantified, or any robustness checks, so verification is impossible from what is shown. If the full paper fixes the null model and reports the checks, the claims become more credible; otherwise the central evidence is weak. This is for readers who track science policy, AI geopolitics, or applied network studies. Someone working on national research investment or international governance would find the descriptive groupings worth seeing, even if they treat the statistical support as provisional. It is important enough and grounded enough in large-scale data to deserve a serious referee rather than a desk reject, provided the methods section addresses the null-model issue directly.

Referee Report

2 major / 2 minor

Summary. The paper claims that analysis of large-scale AI publication data over three decades shows the US and China have diverged in collaboration and citation networks, forming two poles; the UK and Germany integrate exclusively with the US, many European countries converge with both, while developing countries integrate only with China, as measured by comparing observed cross-country links to random realizations.

Significance. If the empirical patterns hold after proper null-model controls, the work provides quantitative evidence on geopolitical polarization in AI research, with direct relevance to national science policies and global regulatory efforts. The network-based approach to tracking integration over time is a strength if the comparisons are robust.

major comments (2)

[Methods] Methods section (null model description): the random realizations for collaboration and citation links are not specified as preserving country-level publication volumes or total degree sequences. Without a configuration-model or similar null that fixes marginals, the reported long-term divergence of the US and China (and the differential integration patterns for Europe vs. developing countries) could be driven by size heterogeneity rather than genuine polarization processes.
[Results] Results (polarization claims): the central finding that the US and China 'have long diverged in both dimensions' and that developing countries 'only integrate with China' depends on the statistical comparison to random realizations; absent details on the exact test statistic, number of realizations, or p-value thresholds, it is impossible to assess whether the observed patterns exceed what volume differences alone would produce.

minor comments (2)

[Abstract] Abstract and introduction lack any mention of the data source (e.g., specific database, time span, AI keyword filters), the precise definitions of collaboration vs. citation links, or robustness checks such as alternative null models.
[Figures/Tables] Figure captions and table legends should explicitly state the number of random realizations used and whether country sizes are preserved.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which highlight important aspects of methodological clarity. We address each major comment below and will incorporate revisions to strengthen the paper.

read point-by-point responses

Referee: [Methods] Methods section (null model description): the random realizations for collaboration and citation links are not specified as preserving country-level publication volumes or total degree sequences. Without a configuration-model or similar null that fixes marginals, the reported long-term divergence of the US and China (and the differential integration patterns for Europe vs. developing countries) could be driven by size heterogeneity rather than genuine polarization processes.

Authors: We agree that the null model must preserve country-level publication volumes and degree sequences to isolate genuine polarization from size effects. The current manuscript description of the random realizations is insufficiently detailed on this point. In the revised version, we will update the Methods section to explicitly state that we use a configuration-model null that fixes the marginal distributions (publication counts per country and total degrees in the collaboration and citation networks). We will also describe the generation procedure and confirm that all reported comparisons use this controlled null. revision: yes
Referee: [Results] Results (polarization claims): the central finding that the US and China 'have long diverged in both dimensions' and that developing countries 'only integrate with China' depends on the statistical comparison to random realizations; absent details on the exact test statistic, number of realizations, or p-value thresholds, it is impossible to assess whether the observed patterns exceed what volume differences alone would produce.

Authors: We acknowledge that the Results section lacks the necessary statistical details for readers to evaluate the comparisons. In the revision, we will add the exact test statistic (normalized deviation of observed from expected links), the number of random realizations generated (1,000 per network and time window), and the significance thresholds applied (including any multiple-testing correction). These additions will allow direct assessment of whether the US-China divergence and country-specific integration patterns are statistically distinguishable from volume-driven expectations under the revised null model. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical null-model comparison is self-contained

full rationale

The paper measures polarization and integration by directly comparing observed cross-country collaboration and citation links against random realizations drawn from the same publication data. This is a standard empirical procedure with no equations that define a quantity in terms of itself, no fitted parameters relabeled as predictions, and no load-bearing self-citations that substitute for independent justification. The central claims follow from the data-to-null contrast rather than reducing to the inputs by construction. Any concerns about whether the null model preserves country-level publication volumes are methodological (correctness) issues, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review based solely on abstract; no free parameters, axioms, or invented entities are described or invoked in the provided text.

pith-pipeline@v0.9.0 · 5454 in / 1051 out tokens · 51812 ms · 2026-05-10T05:02:01.147361+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

46 extracted references · 1 canonical work pages · 1 internal anchor

[1]

Physical Review E—Statistical, Nonlinear, and Soft Matter Physics , volume=

Constrained randomization of weighted networks , author=. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics , volume=. 2011 , publisher=

2011
[2]

Physica a: Statistical mechanics and its applications , volume=

Characterization and modeling of weighted networks , author=. Physica a: Statistical mechanics and its applications , volume=. 2005 , publisher=

2005
[3]

Journal of Complex Networks , volume=

Switching edges to randomize networks: what goes wrong and how to fix it , author=. Journal of Complex Networks , volume=. 2017 , publisher=

2017
[4]

Nature , year=

Massive open index of scholarly papers launches , author=. Nature , year=
[5]

Nature Reviews Physics , volume=

The statistical physics of real-world networks , author=. Nature Reviews Physics , volume=. 2019 , publisher=

2019
[6]

Proceedings of the National Academy of Sciences , volume=

Evolution and convergence of the patterns of international scientific collaboration , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

2016
[7]

Physical review letters , volume=

Generalized bose-fermi statistics and structural correlations in weighted networks , author=. Physical review letters , volume=. 2009 , publisher=

2009
[8]

Physical Review E—Statistical, Nonlinear, and Soft Matter Physics , volume=

Maximum likelihood: Extracting unbiased information from complex networks , author=. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics , volume=. 2008 , publisher=

2008
[9]

Geoforum , volume=

Globalization of science and international scientific collaboration: A network perspective , author=. Geoforum , volume=. 2019 , publisher=

2019
[10]

IEEE transactions on information theory , volume=

Least squares quantization in PCM , author=. IEEE transactions on information theory , volume=. 1982 , publisher=

1982
[11]

Science, technology, & human values , volume=

Understanding patterns of international scientific collaboration , author=. Science, technology, & human values , volume=. 1992 , publisher=

1992
[12]

Science , volume=

Specificity and stability in topology of protein networks , author=. Science , volume=. 2002 , publisher=

2002
[13]

New Journal of Physics , volume=

Enhanced reconstruction of weighted networks from strengths and degrees , author=. New Journal of Physics , volume=. 2014 , publisher=

2014
[14]

OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts

OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts , author=. arXiv preprint arXiv:2205.01833 , year=

work page internal anchor Pith review arXiv
[15]

Research Policy , volume=

Transformation of Global Science core--periphery structure towards a multi-polar horizon: The Rise of China and the Global South from 1980--2020 , author=. Research Policy , volume=. 2026 , publisher=

1980
[16]

New Journal of Physics , volume=

Analytical maximum-likelihood method to detect patterns in real networks , author=. New Journal of Physics , volume=. 2011 , publisher=

2011
[17]

, author=

Smoothing and differentiation of data by simplified least squares procedures. , author=. Analytical chemistry , volume=. 1964 , publisher=

1964
[18]

IEEE transactions on acoustics, speech, and signal processing , volume=

Dynamic programming algorithm optimization for spoken word recognition , author=. IEEE transactions on acoustics, speech, and signal processing , volume=. 2003 , publisher=

2003
[19]

New Journal of Physics , volume=

Unbiased sampling of network ensembles , author=. New Journal of Physics , volume=. 2015 , publisher=

2015
[20]

Research policy , volume=

Network structure, self-organization, and the growth of international collaboration in science , author=. Research policy , volume=. 2005 , publisher=

2005
[21]

Scientometrics , volume=

Growth of international collaboration in science: revisiting six specialties , author=. Scientometrics , volume=. 2017 , publisher=

2017
[22]

Scientific Reports , volume=

China and the US produce more impactful AI research when collaborating together , author=. Scientific Reports , volume=. 2024 , publisher=

2024
[23]

Nature Machine Intelligence , volume=

Towards a new generation of artificial intelligence in China , author=. Nature Machine Intelligence , volume=. 2020 , publisher=

2020
[24]

Nature , volume=

US crackdown harms chinese collaborations , author=. Nature , volume=
[25]

Nature , volume=

China’s AI dreams , author=. Nature , volume=. 2018 , publisher=

2018
[26]

Nature , volume=

The AI race: how the Chinese and US approaches radically differ , author=. Nature , volume=
[27]

Nature Communications , volume=

The global geography of artificial intelligence in life science research , author=. Nature Communications , volume=. 2024 , publisher=

2024
[28]

Nature medicine , volume=

The practical implementation of artificial intelligence technologies in medicine , author=. Nature medicine , volume=. 2019 , publisher=

2019
[29]

Technological Forecasting and Social Change , volume=

Artificial intelligence policy frameworks in China, the European Union and the United States: An analysis based on structure topic model , author=. Technological Forecasting and Social Change , volume=. 2025 , publisher=

2025
[30]

Research policy , volume=

The emergence of China as a leading nation in science , author=. Research policy , volume=. 2006 , publisher=

2006
[31]

Nature human behaviour , volume=

Quantifying the use and potential benefits of artificial intelligence in scientific research , author=. Nature human behaviour , volume=. 2024 , publisher=

2024
[32]

Proceedings of the National Academy of Sciences , volume=

The impact of US--China tensions on US science: Evidence from the NIH investigations , author=. Proceedings of the National Academy of Sciences , volume=. 2024 , publisher=

2024
[33]

Proceedings of the National Academy of Sciences , volume=

The high price of overzealously defending the US research enterprise against theft by China , author=. Proceedings of the National Academy of Sciences , volume=. 2023 , publisher=

2023
[34]

Proceedings of the National Academy of Sciences , volume=

China’s rise as a major contributor to science and technology , author=. Proceedings of the National Academy of Sciences , volume=. 2014 , publisher=

2014
[35]

Proceedings of the national academy of sciences , volume=

China, the United States, and competition for resources that enable emerging technologies , author=. Proceedings of the national academy of sciences , volume=. 2018 , publisher=

2018
[36]

The Next One Will Be Worse

We Warned About the First China Shock. The Next One Will Be Worse. , author=. The New York Times (Digital Edition) , pages=. 2025 , publisher=

2025
[37]

Journal of cleaner production , volume=

Artificial intelligence in sustainable energy industry: Status Quo, challenges and opportunities , author=. Journal of cleaner production , volume=. 2021 , publisher=

2021
[38]

2006 , publisher=

The new argonauts: Regional advantage in a global economy , author=. 2006 , publisher=

2006
[39]

Research Policy , volume=

Academic collaborations and firm innovation performance in China: The role of region-specific institutions , author=. Research Policy , volume=. 2015 , publisher=

2015
[40]

Global Policy , volume=

Fragmentation and the future: Investigating architectures for international AI governance , author=. Global Policy , volume=. 2020 , publisher=

2020
[41]

AI and Ethics , volume=

Mapping global AI governance: a nascent regime in a fragmented landscape , author=. AI and Ethics , volume=. 2022 , publisher=

2022
[42]

European Management Journal , volume=

Chinese foreign direct investment in R&D in Europe: A new model of R&D internationalization? , author=. European Management Journal , volume=. 2012 , publisher=

2012
[43]

Science , volume=

Is Europe evolving toward an integrated research area? , author=. Science , volume=. 2013 , publisher=

2013
[44]

Nature , year=

China wants to lead the world on AI regulation-will the plan work? , author=. Nature , year=
[45]

, author=

Rules to keep AI in check: nations carve different paths for tech regulation. , author=. Nature , volume=
[46]

A European Strategy for Artificial Intelligence in Science: Paving the way for the Resource for AI Science in Europe (RAISE) , type =