The Securitization of Science: A Systems Perspective on Policy and Measurement

Technical Summary: The Securitization of Science: A Systems Perspective on Policy and Measurement

Problem Statement
The paper addresses the growing tension between international scientific collaboration and the rise of "research security" policies in the United States and Europe. These policies, driven by geopolitical competition with China, aim to restrict knowledge transfer in sensitive, dual-use fields by raising the "friction costs" of collaboration (e.g., visa restrictions, funding conditions, disclosure requirements). The prevailing assumption is that such policies will induce a selective withdrawal from high-risk fields, effectively severing ties in areas like computer science and engineering. However, the author argues that this view misunderstands the structural dynamics of science, which operates as a self-organizing "invisible college" driven by researcher-level logic rather than state-level directives. The central problem is determining where the boundary of legitimate connection, redrawn by political actors, actually registers in the observable record of scientific output.

Methodology
The study employs a bibliometric analysis of China's scientific output from 2018 to 2025, utilizing Scopus data across 27 subject categories. The methodology focuses on relative measures rather than absolute counts to isolate dynamic shifts:

• Collaboration Share: Tracking the proportion of internationally co-authored publications versus domestic-only publications.
• Field-Weighted Citation Impact (FWCI): Analyzing the quality of domestic versus international outputs to determine if domestic capacity is maturing independently.
• Dual-Use Classification: Subject categories were categorized into high, medium, and low "dual-use proximity" tiers based on explicit inclusion in US research security targeting (e.g., Export Administration Regulations) and documented pathways to security-relevant technologies.
• Linguistic Analysis: A comparative text analysis of titles and abstracts in Computer Science (CS) papers. The study compares China-affiliated CS papers against a global baseline (2014–2017 vs. 2019–2023) to detect shifts in vocabulary, specifically looking for the decline of security-sensitive terms and the rise of civilian-application language.

Key Results
The empirical findings challenge the "deterrence" hypothesis that security policies will cause a targeted collapse in sensitive fields:

1. Universal Decline Without Withdrawal: China's share of internationally co-authored publications declined universally across all 20 analyzable fields between 2018 and 2022. However, this decline correlates with a massive expansion in domestic-only output (which grew from 77% to 82% of total output). The data suggests the decline is an arithmetic consequence of domestic capacity maturation rather than a policy-induced withdrawal from international partners.
2. The Dual-Use-Proximity Pattern: Contrary to the expectation that sensitive fields would contract the most, the fields with the highest dual-use proximity (Computer Science, Materials Science, Physics, Engineering) showed the least contraction (3–6 percentage point decline). Conversely, fields with minimal dual-use salience (Psychology, Nursing, Dentistry, Health Professions) experienced the steepest declines (8–24 percentage points).
3. Linguistic Adaptation: In Computer Science, while the global baseline saw a massive increase in dual-use terminology (e.g., "drone," "autonomous," "LiDAR" increased by 79–192%), China-affiliated papers exhibited a distinct divergence. While some dual-use terms increased, specific security-sensitive terms (e.g., "facial recognition," "jamming," "malware," "cryptography") declined sharply or disappeared in China-affiliated papers, even as they remained stable or grew globally. Simultaneously, China-affiliated papers saw a surge in civilian-application vocabulary (e.g., "agriculture," "education," "sustainability") far exceeding global growth rates.

Key Contributions

• Theoretical Reframing: The paper proposes a "systems perspective" that views research security not as an external shock to a "stateless commons," but as an interaction between two coupled systems: the political support system (which sets agenda and boundaries) and the self-organizing research community (which selects partners based on scientific value).
• Diagnostic of Policy Limits: The study provides empirical evidence that agenda-setting power can raise friction costs but cannot redirect epistemic choice at the high-value core of the network. The "invisible college" logic persists: researchers maintain ties where the scientific value justifies the friction, while peripheral, lower-value ties dissolve.
• Evidence of Adaptive Participation: The paper documents "adaptive participation" at the researcher level, where scientists maintain frontier collaborations but adjust the form and language of their engagement to navigate political constraints (a form of "desecuritization" at the paper level).

Significance and Claims
The paper claims that the observable record of research security policy is one of "boundary-redrawing" that registers primarily at the periphery of the scientific network rather than the core. The findings suggest that governments have successfully raised the transaction costs of collaboration, leading to a general decline in international share, but they have failed to suppress collaboration in the specific dual-use fields they targeted. Instead, the network has reorganized: high-value ties in sensitive fields persist through linguistic adaptation and strategic framing, while low-value ties in non-sensitive fields have dissolved due to domestic capacity growth.

The author explicitly notes the limitations of these claims: the data cannot capture classified research, collaborations that were never formed (the "chilling effect"), or isolate specific policy instruments from other geopolitical factors. Therefore, the paper does not offer a causal estimate of policy effectiveness but rather a diagnostic of where the political boundary intersects with the self-organizing dynamics of science. The significance lies in demonstrating that the "invisible college" remains robust at its core, operating on its own selection criteria despite the imposition of new political constraints.