Science Diplomacy

Established Science diplomacy is three claims in a trenchcoat, and only one of them has evidence. The standard taxonomy, set out in a 2010 joint report by a national academy and a scientific association, separates science in diplomacy (technical advice inside negotiation), diplomacy for science (treaties and agreements that make facilities and access possible) and science for diplomacy (scientific cooperation used to improve political relations). The middle claim is documented to the point of banality: shared laboratories exist because governments signed things. The third claim — that cooperation between scientists produces cooperation between states — is the one the field is named after and the one with no measured support anywhere in the literature.

Frontier This brief owns the joint question its neighbours each touch and none answers: what keeps scientific cooperation alive when the politics turns hostile? Global Cooperation Models carries the compliance record of treaty regimes, Scientific Institutions Through History carries the institutional forms, and Future Trade Systems carries the measured flow data on fragmentation — including the finding that outside Russia and the two largest economies, East–West trade outperformed world trade. None of them asks what the shared-facility record shows about survival under rivalry.

Frontier The answer the record supports is unflattering to the field’s self-description: cooperation persists where exit is expensive, and stops where exit is cheap. After February 2022 the same shock hit three flagship arrangements and produced three different outcomes. The particle-physics laboratory let its cooperation agreements with Russia and Belarus lapse, and nothing physical broke. The fusion project kept its Russian member delivering in-kind components, because the components are on the critical path and cannot be re-sourced inside a decade. The space station kept flying, because one segment supplies propulsion and reboost and the other supplies power and attitude control, and neither works alone. The variable that predicts continuity is not shared scientific values; it is the cost of separation.

Speculative If that is right, the policy implication inverts the usual advice. Building scientific ties to improve relations is the weakly evidenced intervention. Building physically interdependent infrastructure, and financing it through multi-year in-kind commitments that are painful to withdraw, is the intervention with a track record — and it is the one currently being dismantled, since the station that best exemplifies it is scheduled for destruction at the end of this decade.

Established A note on sourcing. This brief was commissioned in September 2026 from the Institute’s research base. Reading-list entries without links are cited from the bibliographic record rather than re-fetched, and claims are dated no later than early 2026 unless carried by a linked source.

Established The founding instruments are explicit that openness is the price of entry, and they are old. The 1953 convention establishing the European particle-physics laboratory states that the organisation shall have no concern with work for military requirements and that the results of its experimental and theoretical work shall be published or otherwise made generally available. The 1959 Antarctic Treaty, drafted directly out of the 1957–58 International Geophysical Year, reserves the continent for peaceful purposes, provides for freedom of scientific investigation, and requires that observations and results be exchanged and made freely available. Both were signed at the height of a nuclear standoff by the states conducting it. That is the strongest fact in the subject, and it is a fact about what governments will sign, not about what scientists achieved.

Frontier The cost structure of shared facilities is deliberately unknowable, and the flagship project says so out loud. The international fusion project distributes roughly nine-tenths of its construction as in-kind components manufactured by members and delivered to the site. At the July 2024 rebaseline the organisation published a revised schedule — start of research operations 2034, deuterium–deuterium 2035, full magnetic energy 2036, deuterium–tritium 2039 — attributing the slip to supply shocks, first-of-a-kind manufacturing and planning that proved too optimistic. Independent reporting on the same rebaseline records something sharper: the Director-General declined to state a total project cost at all, on the ground that in-kind contributions make one meaningless, against reference points of roughly five billion euros in 2001 and fifteen billion in 2010. A cooperation instrument whose cost cannot be stated cannot be evaluated against alternatives, and that is a design feature rather than an accounting failure: in-kind contribution is what makes members deliver.

Established The space station is the clearest case of interdependence as the binding agent, and it has a termination date. The station masses about 420 tonnes and took 42 assembly flights. A 2026 government audit records deorbit at the end of 2030 or early 2031, an $843 million contract for the deorbit vehicle, operations at about $3 billion a year, and $1–1.5 billion projected for commercial replacements across 2026–31 — which the agency clarified in March 2026 is sufficient to support only one commercial station, against six in development. Frontier The partnership survived the 2014 Crimea crisis and the 2022 invasion intact, with Russian participation extended to 2028. The mechanism usually credited is goodwill; the mechanism visible in the engineering is that the Russian segment supplies propulsion and orbital reboost and the American segment supplies power and attitude control.

Frontier The sanctions response was not uniform across facilities, and the variation is the data. The particle-physics laboratory suspended Russia’s observer status in March 2022 and allowed its international cooperation agreements with Russia and Belarus to expire during 2024, affecting on the order of several hundred users affiliated with Russian institutes. The fusion project retained its Russian member and its component deliveries. The space station continued to fly crews of both nationalities. Three institutions, one shock, three outcomes — and the ordering matches the cost of separation rather than any ranking of political outrage.

Established Where a cooperation regime has an independently instrumented outcome series, it can be scored, and almost none does. The ozone regime is the exception: the 2022 scientific assessment reports stratospheric chlorine down about 420 parts per trillion from its 1993 peak and bromine down about 3.2 from 1999, with return to 1980 values projected around 2035 for mid-latitudes and 2066 for the Antarctic, and it also carries its own corrective on the CFC-11 emissions anomaly. That is what a scorable instrument looks like. No shared research facility publishes anything comparable about its own effect on anything.

Frontier The bilateral agreement that everyone cites as the barometer is now a narrower document. The United States–China science and technology agreement of 1979 was renewed on five-year cycles for four decades. It lapsed into short extensions from August 2023 and was signed again in amended form in December 2024, with scope narrowed toward basic research and new provisions on safeguards, data and dispute consultation. Read carefully, the renegotiation is evidence for the claim in section 1: the agreement survived in the form where the cost of ending it was highest — the framework functions that underpin existing joint programmes — and shed the open-ended parts.

Frontier The access dispute now runs through pathogens rather than through instruments. The World Health Organization’s pandemic agreement opens for signature only once its pathogen access and benefit-sharing annex is adopted. Member states extended negotiations in May 2026; the seventh meeting closed on 17 July 2026 with contractual arrangements, laboratory networks and the definition of benefits arising from pathogen sharing all still open, and an eighth meeting scheduled for 14–18 September 2026. A secretariat reporting that its own negotiation has not closed on three substantive questions is reporting against interest, which is why the deadlock is treated here as established and the outcome as open.

Established The scale claim behind all of this is real and is usually stated too strongly. The largest collaborations now publish papers with author lists longer than their text: the record is a 2015 joint detector paper with 5,154 authors, 24 pages of names and 9 pages of science. Big collaboration is a fact. That it produces diplomacy is a separate claim requiring separate evidence.

Frontier Does scientific cooperation cause political cooperation, in any measured instance? The field’s canonical examples — the Antarctic Treaty out of a geophysical year, arms-control channels run by physicists from 1957, a synchrotron in Jordan whose members include states without diplomatic relations — are all consistent with scientists having been used as an available channel by governments that had already decided to talk. Distinguishing channel from cause requires a counterfactual nobody has constructed.

Frontier Is the decline in cross-bloc collaboration a fragmentation of science or a normalisation? The trade data in the sibling brief shows aggregate flows holding up better than the fragmentation narrative predicts once two specific pairs are excluded. Whether research flows behave the same way is unmeasured at the level of design, and the answer decides whether current policy is treating a crisis or a trend.

Speculative Can a facility be designed so that withdrawal is expensive on purpose? In-kind contribution does this accidentally: a member that leaves strands components it has already built and loses the industrial capability it acquired. Whether that property can be engineered deliberately — and whether doing so is legitimate — has not been studied, and it is the most policy-relevant open question in the subject.

Frontier What happens to cooperation instruments when the anchor facility ends? The space station is the last large piece of infrastructure whose operation requires two rival states daily. After 2031 there is no equivalent, and the fusion project’s components will largely have been delivered. The prediction from this brief’s framing is that cooperation weakens sharply when nothing physical is at stake, and the 2030s will test it without anyone having designed the test.

Established In-kind contribution is the central technical bottleneck of shared facilities and is treated as an accounting curiosity. Members deliver components rather than money, so schedule risk is distributed across national industrial bases with different qualification regimes, different currencies and different political calendars, and no single authority can accelerate a late item by paying more. This is why the fusion rebaseline attributes slip to first-of-a-kind manufacturing and supply shocks rather than to a funding shortfall.

Frontier Export control is now inside the laboratory rather than at the border. Detector electronics, cryogenic systems, superconducting materials and the software that drives them are dual-use in the regulatory sense, so a multinational collaboration must run licence determinations for equipment it jointly designed. The administrative load falls on the collaborations least able to carry it, and there is no published estimate of its size at any facility.

Established Data access is a harder constraint than instrument access, and the gradient runs the wrong way. Detector and telescope data are increasingly open; the compute, storage and expertise needed to use them are not. A country can be a formal member of a facility and remain unable to analyse its output, which converts access agreements into a weaker instrument than their signatories believe.

Frontier Publication is the one channel sanctions cannot cleanly cut, and it is where the collateral damage lands. Restricting institutional cooperation does not stop a paper being co-authored; it stops visas, shipments and shared budgets. The result is that sanctions bite hardest on the physical and financial layers and least on the layer everyone talks about, which is why measures aimed at signalling are often paid for in graduate students.

Established The first dependency is a cost accounting for in-kind contributions. Until one member’s delivered components can be priced on a common basis, no shared facility can be compared with any alternative use of the same money, and the field’s central claim — that these arrangements are good value for the cooperation they buy — cannot be tested at all.

Frontier The second is a counterfactual method for political effects. Synthetic-control and matched-pair designs are standard in policy evaluation and have never been applied to scientific cooperation events: facility accessions, agreement lapses, sanctions suspensions. The data are public and the designs are off the shelf.

Established The third is a published continuity instrument. Every large facility has an implicit answer to what happens when a member is sanctioned, and none has a written one that survives contact with a real case. The 2022 responses were improvised, differed across institutions, and set precedents nobody negotiated.

Frontier The fourth is an access metric that is not membership. Formal membership is counted; beam time used, data downloaded and papers led by researchers from lower-income members are not published in comparable form by any facility, so the equity claims made on all sides are unfalsifiable.

Frontier The decisive test is the natural experiment that ran in 2022 and has never been written up: one shock, three institutional responses, three measurable outcome series. The particle-physics laboratory ended cooperation agreements with two member-adjacent states; the fusion project retained its sanctioned member and its in-kind deliveries; the space station continued joint operations. The prediction from this brief is that continuity tracks the cost of separation and not the strength of scientific ties, and it is testable against data that already exist: component delivery schedules before and after, co-authorship by affiliation before and after, user counts and beam-time allocations, crew rotation and reboost manoeuvres. Nobody has assembled it, and every series is held by an institution that publishes annual reports.

Frontier The second is a synthetic-control study of accession events. States join facilities on dated, staggered schedules. Build a synthetic control from non-joining comparators and measure bilateral political outcomes — treaty co-signature, votes in multilateral bodies, trade in scientific instruments — before and after. If the science-for-diplomacy claim has any measurable content, this is where it shows up; if it does not, the field should say so and defend its facilities on scientific grounds.

Established The third has already run, in the least flattering circumstances, and it worked. When Soviet state science funding collapsed, a privately funded international grant programme supporting more than 28,000 scientists produced a clean eligibility discontinuity: grants more than doubled publication output, significantly raised the probability of staying in research, and had lasting effects. That is the single best-identified result in the whole subject, and what it demonstrates is diplomacy for science — money and access sustaining a research base — rather than science for diplomacy.

Frontier The fourth is already scheduled and nobody calls it an experiment. The station deorbits at the end of 2030 or early 2031, removing the last operational dependency between rival space programmes. Measuring joint activity for a decade either side is a before-and-after design with a fixed date, and the instruments are already reporting.

Speculative The fifth would settle the access dispute and requires only publication. Ask every major facility to report beam time, data volume and lead authorship by member country and income band, annually, on a common template. The equity argument in science diplomacy is currently conducted entirely without these numbers.

Established Physical interdependence is an engineering choice with diplomatic consequences, and it is made unconsciously. The station’s split between propulsion on one segment and power and attitude control on the other was a design compromise, not a treaty strategy, and it has done more to hold the partnership together than any agreement text. Future facilities will make the same class of choice, and nobody analyses it in these terms.

Frontier Schedule coupling is the practical failure mode of in-kind delivery. One late component stops a critical path that no other member can accelerate, so the project inherits the worst national calendar in the partnership. The fusion rebaseline, moving deuterium–tritium operation to 2039, is the quantified version of that inheritance.

Established Successor facilities are designed inside host laboratories long before they are funded, which is why catalogue records for machines like the muon collider study exist at the host laboratory while the political decision remains open. The engineering work is the slow part and the funding decision is the abrupt part, and the mismatch is the reason expertise decays between projects.

Speculative Distributed facilities weaken the binding mechanism this brief identifies. Remote instruments, federated data and cloud analysis reduce the need for people and hardware to occupy one site — excellent for access and bad for the cost of exit. If the interdependence hypothesis is right, distributed science is more fragile under rivalry than the monumental kind, which is the opposite of the usual assumption.

Established Global Cooperation Models owns the compliance record and this brief defers to it entirely. That brief carries the ozone regime as the only independently instrumented success, the trade-institution effect ranging from zero to +171 per cent depending on specification, and the consent-rule machinery of the health regulations. What it does not do is ask why laboratories survive rivalries that treaties do not, which is this brief’s question.

Established Scientific Institutions Through History supplies the base rates. Its throughput finding — a rejection rate holding between 8.5 and 13.5 per cent for a century while submissions quadrupled — is the corrective to any claim that today’s international structures are historically unprecedented in kind rather than in scale.

Frontier Future Trade Systems carries the flow evidence that constrains the fragmentation story, including the finding that excluding Russia and the largest bilateral pair, East–West trade outperformed world trade by 6–10 per cent. Research flows have not been analysed with the same care, and until they are, claims about scientific decoupling should be stated with that asymmetry attached.

Frontier The facility briefs are where the engineering lives: Commercial Fusion on the fusion project’s schedule and cost opacity, Post-LHC Colliders on what a successor machine requires, Space Habitats on what replaces the station, and Existential Risk Governance on the verification regimes that depend on cross-border scientific access.

Established The institutional forms in use are few and old. An intergovernmental laboratory with a founding convention and member subscriptions; a treaty-based project funded by in-kind contribution shares; an operating partnership built on memoranda of understanding between space agencies; a treaty reserving a territory for science; and a standing bilateral framework agreement renewed on a cycle. Every arrangement in the field is a variation on one of these five, and each has a different failure mode under sanction.

Frontier None of the five has a written continuity rule that has survived a real test. The 2022 responses were improvised by each institution separately: suspension of observer status in one, retention of a sanctioned member in another, business as usual in a third. Those improvisations are now precedent, which is the worst way to acquire a rule.

Established The budget instrument is annual national appropriation, and that is the real constraint. A facility with a thirty-year construction horizon is financed by parliaments that vote yearly. In-kind contribution is partly a workaround: a delivered magnet cannot be rescinded in a budget round the way a cash transfer can. The workaround is why costs are unknowable, so the field pays for its funding stability in transparency.

Frontier Access institutions for lower-income members are the weakest part of the architecture. The pathogen access negotiation is the live case: the instrument cannot open for signature until the annex is adopted, and the annex is stuck precisely on who gets what in return for sharing biological material. It is the same bargain as beam time and data access, with higher stakes and better documentation.

Established Sanctions on institutions fall on individuals, and the distribution is documented only anecdotally. When cooperation agreements lapse, the people who lose access are graduate students and postdoctoral researchers with no influence over the conduct being sanctioned. There is a defensible argument that this is the point — that costs must be visible to matter — and there is no published accounting of who actually paid.

Frontier Steelmanning the isolation case: continued cooperation confers legitimacy, and legitimacy is a resource. Flying joint crews and accepting in-kind components from a sanctioned state signals that the ordinary business of great-power science continues, which is exactly what a sanctioned government wants. Anyone arguing for facility continuity has to answer this, and the honest answer is that the trade is real and unquantified.

Frontier Steelmanning the continuity case: the alternative is a world without shared verification. Arms control, pandemic surveillance and climate observation all depend on instruments and data that no single state produces. Cutting scientific ties as a sanction removes capability that is used mainly to reduce the risk of catastrophic misunderstanding.

Established The access bargain is the field’s clearest equity question and it is being litigated in public. Lower-income states supply biological samples, field sites and observational coverage, and receive access on terms negotiated by richer partners. The intergovernmental commentary on the amended health regulations, which entered into force on 19 September 2025, argues that the amendments preserve sovereignty and advance equitable access and production diversification — an interested party, and useful precisely because it states the southern position in its own words.

Frontier The quantified bound on fragmentation is wide enough to be an admission of ignorance. Estimates of geoeconomic fragmentation losses run from about 0.2 per cent to about 7 per cent of world output, with low-income countries facing up to four times the average. The thirty-five-fold spread is the honest finding. The research component of that number has never been separated out, and any confident claim about the civilizational cost of scientific decoupling is drawing on a range that spans a rounding error and a depression.

Established Some observables cannot be nationalised, and that is the durable argument for cooperation. Stratospheric chemistry, seismic monitoring, pathogen genomics and the particle-physics energy frontier require either global coverage or capital beyond any single budget. The ozone assessment exists because measurement was pooled; no national programme could have produced the chlorine and bromine series that made the treaty scorable.

Speculative The long-horizon risk is duplication rather than absence. A fragmented world does not stop doing science; it does the same science twice with incompatible standards, which raises cost and lowers the value of both copies. That is the mechanism by which fragmentation would actually bite, and it is the one nobody measures.

Handwave The claim that science is a universal culture transcending politics is where this subject most often stops being evidence. It is a description of a norm that holds under favourable conditions and dissolves under unfavourable ones, and the 2022 record shows it dissolving at different rates depending on how much equipment was bolted together. Stated as a general truth it is assertion.

These horizons track facility schedules, agreement cycles and negotiation deadlines rather than research capability:

• 10 yr: Established Two dates are fixed. The space station deorbits at the end of 2030 or early 2031, and the fusion project reaches start of research operations in 2034 on the July 2024 baseline. Frontier The station’s end removes the only daily operational dependency between rival space programmes, and this brief predicts a measurable fall in joint activity afterwards. Frontier Expect the pathogen annex to conclude or to be abandoned within this window; it cannot stay open indefinitely, because the parent agreement cannot open for signature until it closes.
• 25 yr: Speculative The fusion project reaches deuterium–tritium operation in 2039 on current plans, and the successor-collider decision will have been taken or deferred past the point where the expertise survives. Speculative If commercial replacements for the station consolidate to one or two operators, the interdependence mechanism does not return in orbit, and the question becomes whether any terrestrial facility inherits it.
• 50 yr: Speculative Either a new class of shared facility exists whose exit cost binds members — a lunar installation, a next-generation collider, a pooled observation system — or the in-kind model will have been replaced by contracted procurement, which removes the binding property along with the opacity. Handwave Which of these happens depends on geopolitics that nothing in this brief constrains.
• 100 / 250+ yr: Handwave Statements at this horizon are statements about whether states still exist in their present form. The only durable observation is structural: cooperation has historically been purchased by making separation expensive, and that mechanism is older than any of the institutions discussed here.

• Depends on The compliance evidence in Global Cooperation Models, the institutional base rates in Scientific Institutions Through History, and the flow data in Future Trade Systems, which between them supply everything this brief does not have to re-derive. No scientific result blocks this topic; what blocks it is that the institutions holding the relevant series do not publish them.
• Requires (not on this map) Five things, none of them a discovery. First, a published cost accounting for in-kind contributions to shared facilities: the fusion project’s Director-General declined to state a total cost because in-kind delivery makes one meaningless, and until members’ deliveries are priced on a common basis no shared facility can be compared with any alternative use of the same money. Second, a counterfactual estimate of scientific cooperation on political outcomes — a synthetic-control or matched-pair design on staggered facility accessions and agreement lapses. This is the one item that is genuinely research, the methods are standard, and the absence of any such estimate is why the field’s defining claim remains unsupported. Third, a written continuity rule for shared facilities under sanctions: the 2022 responses were improvised institution by institution and are now precedent. Fourth, multi-year facility funding outside annual national appropriations, because in-kind contribution exists largely to defeat the annual budget cycle, and it buys stability by destroying cost transparency. Fifth, a pathogen access and benefit-sharing annex in force, since the parent pandemic instrument cannot open for signature until the annex is adopted, and the annex was still open on the three substantive questions in July 2026.
• Enables A measured answer here would let the verification regimes discussed in Existential Risk Governance be designed around a known rather than assumed relationship between scientific access and political continuity. No typed enabling edge is claimed, because the relationship has never been measured in either direction.
• Adjacent The facility engineering in Commercial Fusion, Post-LHC Colliders and Space Habitats; the instrument-cost side of competition in Technological Sovereignty; and the historical record of scientists moved between systems by policy in Soviet Frontier Science.

Handwave Claim: the particle-physics laboratory ended the Cold War, or helped to. The laboratory was founded in 1953 and kept East and West physicists working together throughout, which is true and remarkable. The causal claim attached to it has no counterfactual, no control and no measurement, and the same governments that funded the cooperation were simultaneously building the weapons. The defensible version is narrower: an intergovernmental laboratory with an open-publication convention proved politically survivable for seventy years, which is a fact about institutional design.

Frontier Claim: the space station shows that science transcends politics. The station shows that a partnership survives when one partner supplies propulsion and the other supplies power. The transcendence reading predicts that cooperation would also have survived without the physical coupling, and the 2022 record contains a case where a purely institutional tie was cut within months once nothing physical depended on it.

Established Claim: shared facilities are cheap because contributions are in kind. They may be, and nobody can say. The in-kind structure means the project’s own leadership declines to state a total cost, and reference points of roughly five billion euros in 2001 and fifteen billion in 2010 are the closest thing to a series. An arrangement whose cost cannot be stated cannot be called good value without an argument that does not depend on cost.

Frontier Claim: sanctions do not affect science because publication is open. Publication is the least affected layer. Visas, shipments, shared budgets, beam-time allocations and student placements are the affected layers, and they are the ones that determine who becomes a scientist. Measuring the effect of sanctions by counting co-authored papers systematically understates it.

Established Claim: a treaty with a scientific annex is a scientific instrument. Rarely. The ozone regime is scorable because an independent measurement series exists, not because the treaty says science. Regimes without instrumented outcome series — which is most of them — cannot be graded at all, and the sibling brief on cooperation models makes that point across a much wider sample.

Speculative Claim: open data solves the access problem for lower-income partners. It solves the permission problem and leaves the capability problem, since analysis requires compute, storage and trained people. No facility publishes beam time, data volume or lead authorship by member income band, so the size of the remaining gap is unknown and every party to the argument asserts it.

Handwave Claim: science diplomacy is a profession with a method. It is a label applied after the fact to a heterogeneous set of arrangements: a laboratory convention, a treaty, an operating partnership, a grant programme, a bilateral framework. The taxonomy from 2010 is useful and descriptive; nothing in the literature turns it into a procedure with predictable outputs, and pretending otherwise is the field’s characteristic overclaim.