Why this session matters to us
Of everything on Day 1, this was the session built for measurement. The programme asked seven panellists to identify the priority scientific gaps and system-level challenges facing the Science Decade: where the world is still failing to mobilise science at the scale the SDGs require, and how science should be organised, funded and governed instead. Then it asked them to close the day by setting an analytical frame for Day 2.
A diagnosis is a measurement claim. It asserts that something is short by some amount. So the test for a session with this title is narrow and fair: across ninety minutes, did anyone put a number on the gap? Mostly they did not. What the panel produced instead was a set of institutional achievement figures, several sharp qualitative namings of exactly the right problems, and one structural number whose value was zero.
The numbers in the room were achievements, not gaps
Inventory the figures actually spoken and nearly every one describes what an institution has done rather than what the system is missing.
- ICTP has hosted “two hundred thousand scientists from all over the world in the past sixty years,” its director said, of whom “about two thirds of them are from the developing countries who are fully funded by ICTP, and about one third of them are women.”
- The European Physical Society is “a society comprised of forty two national physical societies and we represent more than one hundred thousand physicists in Europe.”
- SESAME in Jordan has run continuously since 2018, publishes at an average impact factor “around five,” and is “the only one in the world powered by solar energy.”
- The Youth as Researchers model, adopted by UNESCO since 2020, “has resulted in 60 youth-led knowledge generation projects throughout the globe.”
- The SDG synthesis coalition, formed in 2022, “was made up of some forty UN organisations.”
These are real numbers and mostly good ones. None of them measures a gap. Nobody on the panel gave R&D spending as a share of GDP, or researchers per million people, or the North–South split in publications, authorship or research funding, or any figure at all for who is failing to mobilise science at the scale the goals require. Magdalena Skipper, the Editor-in-Chief of Nature, came closest by naming the category outright, “persistent data and monitoring gaps,” particularly in local, disaggregated and real-time data, and in community-level data that would show who actually benefits from an intervention. She named the gap. She did not size it. That is a fair summary of the afternoon.
The exception came from Rolf-Dieter Heuer, and it is the sharpest thing in the transcript precisely because it is a count. Having described SESAME as the only synchrotron light source in the Middle East, built into a membership of Cyprus, Egypt, Iran, Israel, Jordan, Palestine, Pakistan and Turkey, he turned to the continent that has none.
He then did what almost nobody else did: he made it a target with a deadline attached to the Decade itself. “A decade is ten years,” he said, and his ask was to see the first stone laid for an international facility in Africa, with a first step of “a beam line and experimental station at Sesame.” It is a small proposal and a checkable one, which is more than most of the afternoon produced.
The panel’s other system-level number was a comparison, and it was loose. Cetto put annual world military and defence spending near one trillion dollars and set it against funding limits that she said are compromising public health and public education in many countries. Dabholkar reached for the same contrast in his closing recommendation.
The argument holds. The arithmetic was approximate. SIPRI, which produces the standard series, put world military expenditure at USD 2,887 billion in 2025, about 2.5% of world GDP and the highest total it has ever recorded. Whatever total the panel had in mind, the one comparative figure the session offered against the science and climate financing gap arrived understating its own case by roughly a factor of three.
The publishing system, described from inside it
Skipper had the rarest seat on the panel, and she used it to describe the incentive structure she administers. Her second point was that research translates poorly into policy and practice, and that this is a design feature rather than an accident.
That is the publication bias question named by the person best placed to name it, and it is worth registering that she named it without a number: no figure for what share of the literature is replication, or synthesis, or negative results, or for who gets published where. She argued for a shift toward implementation science, away from asking only what works and toward “what works at scale,” “what works in low-resource setting[s],” what the unintended consequences are, and in every case “who benefits and who loses.” She reported that Nature has widened its own scope in that direction, and pointed to a specific unopened stock of evidence: the thousands of evaluation reports “hidden in institutional vaults [that] need to be opened up and synthesised.” She also called out the university itself, noting that most institutions “continue to be organised in very siloed traditional discipline ways” while the problems are not.
A panellist diagnosed the panel
The most direct diagnosis of the session came from Ana Maria Cetto, and its subject was the session. Physics dominated the room: ICTP, SESAME and CERN, the European Physical Society, and Cetto herself, a physicist working on the foundations of quantum mechanics. She said so out loud.
Her diagnosis proper was the bluntest of the day and none of it was quantified: that “we live in a world that is deeply divided between extreme wealth and systemic marginalisation”; that much of the science and technology under discussion “is geared toward making the world less sustainable”; that privately funded, business-oriented institutions are displacing public higher education and research; and that “scientific careers are becoming less appealing to the younger generation in many countries as there are no new positions available.”
One of her observations cuts directly against the geography the Decade’s own reporting emphasises. The erosion of scientific careers, she said, “is not a global north versus global south divide, it’s happening in countries in the north, in the countries in the south, everywhere.” That is a claim the Decade does not currently measure. Her prescription was for more basic science rather than more technology, and for the natural sciences to work with neurology, psychology, philosophy, history, economics and sociology on why humans behave as they do. Her closing line was the honest one: “But if there is no political will, we will come nowhere.”
AI took the frontier slot
The programme reserved space for AI and frontier technologies, and it went to Jian Wang, Director of Zhejiang Lab and a member of the UN Independent International Scientific Panel on AI, joining online. He argued that AI belongs in the same class of object as paper: a general-purpose instrument that extends what a brain can do, and now “as fundamental as mathematics.” He cited the panel’s own report for the proposition that an open source model is “a critical pillar of the modern technology landscape and can catalyze distributed global innovation,” then pushed past it: an open AI model is not like open source software, because behind any model sit electricity, computing power and data. He would rather call it an open resource model, which makes openness a question of sharing physical resources and therefore a question of governance.
His concrete recommendation was a measurement one. Almost every model today is built on text, he said, and scientific data itself is not a first-class input: “The scientific data itself are not the natural citizen for the AI models.” Most scientific data, on his account, is read closely by a handful of specialists and otherwise untouched. His ask was that scientists worldwide “build what I call the science foundation model” trained on scientific data rather than papers about it. He grounded it in GeoGPT, the geoscience model his lab has spent five years building, which he said is an endorsed initiative of the Decade and which he described lowering the barrier for young researchers in Africa to do work they could not otherwise do. It runs under a global governance committee.
This is the cross-thread worth marking. The AI panel’s preliminary report assessed that AI capabilities are advancing faster than the ability to measure or govern them, and that the AI divide is about the capacity to shape AI rather than merely to access it. Wang’s intervention is the same argument arriving from the supply side, and it is the one place in this session where a panellist proposed building the missing measurement infrastructure rather than lamenting it. Dabholkar made the institutional version of the same point, warning that “if these resources remain locked up into just a few organizations or few companies in the West coast, it’s not going to be good for the world,” and describing an ICTP consortium with CERN to put computing into the University of Benin in Nigeria. De Teresa supplied the counterweight, arguing AI is reshaping the job market and putting human control at risk, that it “clearly has to be regulated,” and that “we can’t delegate education on AI”: what is needed is “an ethical education for a changing world.”
Organized, funded and governed: what was actually proposed
Running long, the moderator cut the discussion and asked each panellist for one concrete recommendation to strengthen the Decade. The answers are the honest measure of the session, and they split cleanly.
Some were operational. Heuer’s first stone in Africa. Wang’s science foundation model. Dabholkar’s argument for building scientific communities in place rather than exporting outputs to them, made with the sharpest example of the day: if Ghana depends on hydroelectricity it needs climate predictions, and “if you don’t have a community of climate scientists working in Ghana, having some climate model far away is not going to be useful.” Only then can countries in the global south “be sitting on equal terms at the negotiating tables.” John Canavan asked for the thing closest to accountability: a high-level UNESCO statement of policy guidance on public engagement in science, secured with member state support, “but to implement and to monitor same.”
Others were a call for more cooperation. Skipper’s recommendation was unity across scientists, governments, publishers and society, aiming to “turn every single inhabitant of this planet into an advocate for science and innovation.” The moderator’s reply is the most quietly useful line in the transcript.
Canavan had already supplied the reason that answer is not enough. Public engagement, he said, is often not recognised, valued or resourced by institutions, and the deeper barrier is “the unwillingness of scientists to give up the levels of power and control necessary to achieve meaningful engagement.” He was also the only speaker to name the failure mode of the very thing everyone was recommending: academics and policymakers “can be involved in gaming, ticking the box of public engagement with no genuine value-based commitment to it.” Heuer added the discipline that ought to travel with any of this, warning that advice must carry its uncertainty: “We can only advise, the politicians have to take the decisions”, and when scientists do advise, “we have to be very very clear what uncertainty is on this advice, because all results have an uncertainty.” Dabholkar put the same modesty in one clause: scientists “should also be modest about our impact.”
Cetto’s proposal was to revive a recommendation from UNESCO’s 2024 General Conference for a high-level forum of intellectual leaders, an idea traced back to the exchange in which Einstein asked Freud why war. The moderator offered her own: a “capturing the stories” initiative in which students document what each of the Decade’s funded initiatives is actually doing.
The session was billed as concluding Day 1 with an analytical frame for Day 2. It did not produce one, and the moderator said so: “I’m not going to try and summarise this discussion. First of all, there were out of time.” The recommendations were to be written up and passed to the Decade’s leadership afterwards. Then a band played and the day ended.
Our read
The panel found the right problems and left them unmeasured. Read the transcript against the Global Report launched that same morning and the two are describing the same continent-shaped hole from opposite directions. The report says Africa holds under 10% of the Decade’s initiatives against 17.5% of the world’s population. The panel said there is no synchrotron in Africa, and that a climate model built far from Ghana is not much use in Ghana. Same finding. One of them counted it.
That is not a small difference, because a diagnosis you cannot count is a diagnosis you cannot check next year. The Decade now has a baseline for its own initiatives. It does not have one for the frontiers this session was convened to prioritise, and the session did not propose building one. Heuer’s zero is the exception that shows what was missing: it is falsifiable, it has a target, and in ten years someone can go and look. Cetto’s complication deserves to survive alongside it. If the collapse of scientific careers really is happening “in countries in the north, in the countries in the south, everywhere,” then a geography-only reading of the gap will miss it, and geography is the axis the Decade currently counts on.
Why it matters for the SDGs
Every panellist who was asked to name the most important goal named the same one, and it was SDG 17. The moderator opened by calling partnership the most important SDG and, in the same breath, “the one that has the least scholarly study.” Skipper said she reached the same conclusion when she first read the goals. Heuer built his entire intervention on it. That consensus is the session in miniature: the goal everyone agrees matters most is also the one nobody has the numbers for, and this panel did not add any. The concrete SDG business here sits in target 17.18 on disaggregated data, which Skipper named as a persistent gap, and target 17.6 on science and technology cooperation, which is what SESAME, ICTP and the proposed African facility actually are. Underneath both is SDG 9.5, the target that commits countries to enhance scientific research and increase R&D spending and researchers per million people. It is the only place in the framework where the world already agreed to count exactly what this session was convened to diagnose, and it went unmentioned.
Watch & read
- Full IDSSD Global Conference 2026 coverage, including the opening day and the first IDSSD Global Report, the Decade’s own count of its 397 endorsed initiatives across 79 countries.
- The cross-thread on AI: the Independent International Scientific Panel on AI, on which Jian Wang sits, and its preliminary report, the document he cited from the panel.
- Trends in World Military Expenditure, 2025, SIPRI, the standard series behind the figure the panel reached for.
- The institutions the panellists described: SESAME · ICTP · European Physical Society.
- The 2021 UNESCO Recommendation on Open Science, the instrument three panellists argued from, and the Decade’s endorsed initiatives.
- Conference announcement and programme, UNESCO.
This summary is drawn from an automated transcript of the UNESCO webcast; the conference publishes no official transcript, and quotations are lightly edited and should be read as close paraphrase. Speaker labels are not reliable across the recording, so speakers are named here only where the content makes the attribution unambiguous, and are otherwise cited by role. Figures are as speakers stated them and were not independently verified, except where this page names an outside source. Non-English passages are summarised, not quoted.