Missions – possible?

Everybody loves a good mission. The Manhattan Project compressed a decade of theoretical physics into three years of industrial-scale engineering and produced a device that ended a world war. Apollo took a president’s dare and, within the decade, landed two men on a body a quarter of a million miles away.

But the reason these examples are so oft-cited, more than half a century on, is that they’re rare. Most scientific progress doesn’t work this way, and we oughtn’t pretend it does. National missions are grand undertakings – they run on both ambition & architecture; the careful selection of goals matched with specific structures to deliver. They require the singular and deliberate concentration of effort, attention, talent, and a huge amount of resources on a well-defined objective; if everything’s a ‘mission’, nothing is.

The language of missions has been loose in British policy for some time now. Mariana Mazzucato popularised the concept amongst the wonks of Whitehall; the Conservative government’s Industrial Strategy adopted it; Labour now governs under five grand “missions” that straddle huge areas of domestic policy. The word has become somewhat promiscuous – it was only a matter of time before it reached DSIT.

The government’s AI for Science Strategy (published last November) does something refreshing in this space.1 Instead of paying lip service to mission-ology, it makes a concrete commitment to one flagship mission: a specific, time-bound, falsifiable target to develop trial-ready drugs within 100 days by 2030. It also commits to selecting a further round of missions in 2026.

We were each co-authors of the Tony Blair Institute’s AI for Science paper that fed into the strategy. That paper did not recommend missions. But with one mission announced and more on the way, the stakes are high enough to warrant a more thorough examination. History shows that well-executed missions can produce spectacular results; recent experience shows there are plenty of ways for government to fall short. We want to try to help narrow that gap as much as possible, and make the AI for Science missions a success.

First, principles

AI for science matters to Britain. We are fundamentally a scientific nation; a considerable share of our prosperity and national character is derived from this fact. Britain’s scientific inheritance is one of the few unambiguous advantages we carry into this century, and AI will reshape what that inheritance is worth.

Science is about to transform: radically and within a generation. Getting this transformation right is among the more consequential things the government can do. Few countries have more to gain – or more to lose – than we do.

Our greatest challenge is embedding AI across the whole scientific landscape: building enabling infrastructure, generating and sharing high-quality data, incentivising upgrades to the tools and workflows on which scientists depend. But alongside that broader effort, there’s space for something more concentrated: specific programmes that marshal resources and attention toward defined scientific goals.

In this spirit, the AI4S strategy commits to a handful of missions – but what does that mean?

There are many species of ‘mission’: scientific moonshots versus the creation of broad societal goods; government as principal versus government as conductor. Do they mean a demanding, resource-intensive push toward a technical objective, like sequencing the human genome? Or do they mean a broad, systemic goal, like using AI to tackle the reproducibility crisis? These are each fundamentally different propositions, with different selection criteria, different delivery models, and different measures of success.

As a baseline, we should outline what we think DSIT intends. Per the strategy, the AI for science missions are aiming for technical and scientific breakthroughs, and sit closer to a conductor model – government orchestrating activity across academia, industry, and public institutions, concentrating resources, and using its convening power and national assets.

This is helpful framing as far as it goes, but it doesn’t go very far. We have a rough idea of the kind of missions they’re aiming for, but not yet any detail about how they’ll be run. And it is in these details that missions succeed or fail.

The credibility and success of the missions programme depends entirely on two things: the quality of what gets selected, and the delivery architecture that surrounds it. This post covers the first of these.

The second half of the problem – implementation – matters just as much as selection. We shouldn’t take it for granted that delivery will be up to par. Anyone who’s worked inside the system knows that government’s default mode is to treat the announcement as the product. Keeping a mission going depends on sustaining a punishing standard of rigour, and once political attention moves on, it can be difficult for senior civil servants to justify spending significant ongoing resources on yesterday’s news. Indeed, delivering missions well may require entirely new institutional architecture – both inside and outside government. We will return to this at a later date.

The good, the bad, and the ugly

Now that missions have been announced, proposals are sure to come thick and fast. Researchers, industry lobbyists, and think-tank operatives will rush to the new seam to pan for gold. Some ideas will be strong. Some less so: there is little cost to throwing ideas out there, and the policy ecosystem is not always known for its quality control. The reputational upside of having your domain selected is considerable; even unsuccessful bids tend to gain disproportionate airtime, while proponents seldom bear the delivery risk.

If you squint, just about anything can look like a “mission”, and we should be wary of uncritically memeing misguided ideas into the mainstream. More important than rushing to stake claims on specific domains is to establish what kinds of proposals are genuinely well-suited to be national missions; if the government is to select well, it needs to know what it’s selecting for.

The strategy selects five broad areas of science and technology where AI could unlock progress: engineering biology, fusion energy, materials science, medical research, and quantum technologies. Most of the strategy’s efforts will be directed at these domains – including missions.

Each area was selected on the grounds that it shows existing UK strength, alignment with wider UK strategy, and opportunities for AI-driven progress. It would be hard to disagree that any of the five clears that bar. But each is a hefty field. A huge variety of work within these areas will be important and deserving of funding and support; few problems are specifically national mission-shaped. The bar for domain selection was deliberately low: the bar for mission selection must be considerably higher.

To raise that bar, we need to know what we’re aiming for – to reverse-engineer what’s mission-shaped from an understanding of what national missions are, and how they work.

There’s good prior thinking on this within government that deserves to be dusted off. In 2018, the Council for Science and Technology – then chaired by Patrick Vallance, now Science Minister – published advice on how to run a successful mission programme. In their view:

“Missions should be ambitious, transformative programmes that address grand challenges. They must have a clear vision, be problem-led without pre-defined solutions, and have a quantifiable outcome. They should not consist of existing Government policy re-branded.”

They go on to say that successful missions require (and are characterised by) a fundamentally different approach to management and delivery: a single accountable leader, a dedicated cross-government team of sufficient critical mass, ring-fenced resources, simplified governance, and access to external expertise.

These elements are a good jumping off point for thinking about what makes a strong mission candidate, but they are generalised to any mission programme. What follows is our attempt to build on them: a set of criteria, some analytical, others practical, that we think should govern the selection of AI for science missions. It is not exhaustive, but it covers the ground where we think the risks of getting it wrong are highest.

1) Definitional clarity

What is AI for science?

“AI for science” is almost as promiscuous a term as mission. Does it mean science that makes use of machine learning? AI applied to the research ecosystem? Does it include research into artificial intelligence? And does it matter?

The government says the missions should aim for goals “that can only be reached through breakthrough scientific progress enabled by AI” and that missions may occur in priority areas where AI is “poised to play a significant role in accelerating progress.”

This is sentimentally a good starting point – we like breakthroughs. But as the programme gets underway and selection approaches, government should articulate a considerably more specific definition. Almost every field and type of science now uses AI in some form – as a workhorse for data analysis, as an aid to simulation, or as an accelerant to processes that were previously manual and slow. Most serious new research programmes launched today will have AI woven in as a matter of course. If the test is simply whether a mission involves science and is significantly enabled by AI somewhere along the way, then virtually anything qualifies.

We think the threshold must be more demanding: an AI for science mission should be one where AI is not merely present but constitutive – where the goal is unachievable without specific novel AI capabilities, or where the mission itself must build the foundational assets on which those capabilities depend. This is a harder test than “science enabled by AI,” but it is the only one that gives the label real discriminating power.

The mission programme may buckle if swathes of proposals can claim the “AI4S” label with equal plausibility. It must be able to distinguish genuinely exciting opportunities for value-creation from rebranded scientific programmes that would have been proposed regardless.

2) Quality over quantity

Can we afford another?

Good national missions are expensive. The Vaccine Taskforce agreed a £5.2 billion programme business case with the Treasury. The London 2012 Olympics cost north of £9 billion. AI for science missions will not build stadiums or procure vaccines at breakneck speed during a pandemic; but the resources required to stand up anything deserving of the name are nevertheless substantial.

The government has committed up to £137 million to the entire AI for Science Strategy. That figure covers a lot: data infrastructure, fellowships, doctoral training, interdisciplinary teams, and missions. Not all of it – probably not even most of it – will go to national missions. The strategy also doesn’t say how many missions it intends to select: it commits to “a handful”, so probably more than two, fewer than ten. In the most restrained possible case, we are talking about an absolute maximum of low tens of millions each.

In scientific programmes, this does not get you very far. Crowding in private or philanthropic funding then – as the government has said it wants to do – becomes a condition of viability. But it is also in a sense definitional: missions require others to have skin in the game, so the selection process should favour candidates where such co-investment is not just possible but natural.

Each new mission is also competing for the same inputs: leadership talent with both scientific credibility and delivery capability, compute allocation (the strategy’s own “system takeover” model ties up 80% of national capacity for two weeks per mission), and sustained political attention. These are extraordinarily difficult to parallelise.

The government has said that part of the reason it is pursuing missions is to ensure resources “are targeted towards shared goals instead of being spread thinly across diffuse opportunities.” That’s the right instinct, but the pressure to add missions will be considerable. Announcing the next one is virtually costless and would confer considerable political upside in the short-term (and the goodwill of the research base). Many ideas will be advocated for with great force. Resisting that pressure will be the government’s first and most important act of discipline.

3) Meaningful, falsifiable targets

Can describe success precisely enough to know if we got there?

The most basic question a mission candidate must answer is also the one most easily fudged: what, precisely, would it mean to succeed? A mission needs a target that’s concrete, specific, and tied to a deadline: we will cross this threshold by this date. “Trial-ready drugs within 100 days by 2030” is a target that government can succeed or fail at, which makes it stick out from the usual mush of “strengthen the UK’s position” or “unleash innovation.” A target that cannot be clearly met or missed is moot.

The target must also be set at the right altitude. Too modest, and it fails to justify the machinery of a national mission. Too unrealistic and no one will take it seriously. The target should sit in the narrow band where achievement is genuinely uncertain but plausibly within reach – stretching enough to demand extraordinary effort, credible enough that failure reflects on the programme rather than our level of ambition.

A target must also be compatible with the delivery model it depends on. If meeting the goal would require fundamentally re-orienting ongoing activities across research communities, industry, or other areas of government, then either the mission team needs the mandate and resources to drive that reorientation, or the goal needs to bend to accommodate reality. A target that assumes a level of systemic coordination the delivery architecture cannot provide is a fiction.

As any scientist will tell you, perhaps the most critical consideration when setting a goal or readout metric is correct causal attribution. You need to be confident that your mission was the cause of (or at least a significant contributing factor to) the outcomes you hope to observe. This rules out a lot. A target pegged to macro-level indicators like total factor productivity, R&D intensity, or private investment in research would fail immediately: these indicators are driven by such a wide range of forces that no single programme could honestly claim ownership of the result.

Equally, a target pegged to broad research assessment processes would fail. The REF, for instance, is shaped by so many variables that no intervention could credibly take credit – and its cycles are far too slow and too infrequent to function as a meaningful feedback mechanism in the first place. Good mission targets need verifiable causal chains between the intervention and the result. This also means that, wherever possible, progress should be independently verifiable – not self-assessed by the mission team or the department that sponsors it.

And finally, a really serious mission must be capable of embarrassing its sponsors. To know when you have succeeded, you have to know when you’ve failed – and that failure must be visible enough to sting. The London 2012 Olympics could not be redefined as a partial success if the stadium was unfinished on opening night: the date was fixed and the world was watching. Pressure is the mechanism by which missions sustain political weight; visibility keeps everyone honest.

4) Multi-faceted challenges…

Is the problem the right shape for a mission?

Scientific research can come in all shapes and sizes. Some efforts are hell-bent on chasing a single quarry; some are sequential, linear pipelines; some traverse vast and directionless oceans of understanding. AI can likely play a role in all of these. What matters for mission suitability is the shape of the problem.

At one end of the spectrum, AI is already accelerating well-contained processes across dozens of fields – improving image classification in pathology, say, or optimising the geometry of a turbine blade. These are discrete technical challenges where AI is applied directly to a known bottleneck, without requiring the government-anchored orchestration of multiple actors. Some of these efforts are enormous in their ambition and impact, but the bottleneck is capability, not coordination, and they don’t need the machinery of a national mission.

At the other end sit the great diffuse frontiers: broad research spaces that would benefit greatly from widespread AI adoption, but without a defined north star – no singular goal that requires a grand coordination of effort. Open-ended exploration across multiple fronts lacks the bounded structure a mission demands, with no clear map of pathways that can be shortened. ‘Understanding the mechanisms of neurodegeneration’, for instance, is a functionally edgeless space containing countless local bottlenecks that AI might help relieve. But the science advances on a broad front; you can’t set a deadline for a problem with no vision of what a ‘solution’ looks like. A mission is not a general-purpose strategy; it’s a siege engine – and it needs a wall to batter against.2

The sweet spot lies between these poles. The right candidate for a mission should aim for a scientific or technical goal that was previously out of reach, requiring a multi-stage process and coordination between disparate actors who won’t align on their own – a network of distinct challenges. Crucially, there must be a credible account of why an AI-oriented mission can shift the dial.

5) ...that need government intervention

Does government play a necessary role here?

There should also be an understanding of why government involvement is necessary, and what a national mission can offer that can’t be achieved otherwise. The strategy talks about missions “galvanising activity and coordination across UK industry and academia”. This is a natural place for government to be, and aligns with the sweet spot described above. But government can only provide so much funding and bandwidth, so it should be very strict about discriminating on the basis of additionality.

That said, additionality should be assessed against the full range of what government can uniquely provide: access to, or redeployment of, existing national assets – including public facilities or secure datasets; coordination and brokerage power that can anchor the disparate efforts of multiple actors inside and outside government; and help with addressing policy or regulatory blockers.

We may think more about some of these concepts in a future delivery-focussed piece. For now, it’s sufficient to say that if the private sector or public institutions could get there with a push in the right direction (or with barriers removed), it’s probably not a good candidate.

Let’s apply these last two tests to the drug discovery mission. On the shape of the research problem: the preclinical pipeline – target discovery, hit identification, binding affinity prediction, lead optimisation, ADMET modelling, and preclinical testing – is sequential, well-understood, and (critically) slow for reasons that AI seems specifically suited to address. A mission approach is sensible because there are multiple specific stages where specific AI capabilities produce specific acceleration.

But the problem isn’t merely scientific – if it were, then the research base and pharmaceutical industry would get there on their own. Critical enablers require targeted action from government: offering usage of national facilities, establishing new public infrastructure like the Health Data Research Service, and reforming regulation to enable the Medicines and Healthcare products Regulatory Agency to accept AI-generated evidence. These are public goods or pre-commercial activities that the market won’t supply, and collective action problems that no one company or institution has the incentive to address.

6) Consideration of existing national assets

Are we playing to our strengths?

If you want to play missions on easy mode, you don’t begin from a standing start – with a budget this modest you can’t afford to. Far better to leverage what already exists to give the mission every possible advantage. What counts as “enough” will vary by domain, but the question for any candidate is simple: does the UK already possess the assets needed to deliver, or are we planning to build them as we go?

For drug discovery, the existing asset base is comprehensive. The UK has considerable infrastructural strength and depth, across research talent, high-quality data, and world-leading institutions and infrastructure. But there is also an established commercial and regulatory ecosystem beyond the lab ready to translate AI-derived treatments into products and improved patient outcomes. Not every mission will need this kind of end-to-end readiness, but existing assets mean the mission can focus its limited resources on the genuinely hard problems. Whatever the relevant assets are for delivery in a given programme, the mission is much more likely to succeed if they’re already in place.

But ‘existing assets’ requires more granular thinking than simply listing major institutions. An asset could equally be a specific capability – such as a recent breakthrough in experimental technique – that an AI mission can exploit. For example, recently, a breakthrough at the Diamond Light Source – a development in ultra-high throughput crystallography – has enabled researchers to generate structural data at orders of magnitude greater scale than was previously possible. This is the kind of step-change that can make an AI mission newly viable, but it required deep domain awareness (and external expertise) to recognise its potential for the drug discovery mission. Learning from this, DSIT should undertake a detailed capability scan, mapping breakthrough experimental technologies that could generate the foundations on which AI-driven science depends.3 If you don’t know what you have, you can’t play to your strengths.

There is a trade-off here. In some cases it will be acceptable – desirable, even – for a mission to build on a more nascent ecosystem rather than exploiting existing institutional strengths. This might be justified where strategic necessity demands it, or where an emerging field shows extraordinary promise but lacks the critical mass to develop without a deliberate concentration of effort. Missions can be a powerful vehicle for fanning early sparks into something larger. But the less you have in place at the outset, the more honest you’ve got to be about what that means for deliverability and timelines. A mission that must first construct its own foundations before it can pursue its core objective is a fundamentally riskier proposition – and the timescales involved must be compatible with the urgency that justifies a mission approach in the first place.

7) Mission Control

Is it governable at the right level?

Missions succeed or fail on the strength of decisions taken by the right people, at the right level, at the right time. Every Manhattan project needs its General Groves, and the enabling operating system around him that turns individual decisions into coordinated delivery at scale. The science is critical, but what makes a mission a mission is the deliberate concentration of authority, resources, and strategic vision under a defined line of command, oriented towards delivery.

Government departments typically aren’t accustomed to operating this way. The AI for Science strategy contains a long list of activities that appear to support the drug discovery mission, but as yet there isn’t much detail on how this or future missions will be governed.

Everything turns on the answer: it’s the difference between simply defining targets and outlining commitments – words on pages – versus something that can actually be done. Will these “missions” essentially be funds attached to loose collections of already ongoing complementary activities, operating under a shared umbrella heading? Will they be governed through the familiar machinery of “deliverology”, where plans are set at the outset, milestones tracked, and deviations flagged? Or will mission leaders be given genuine authority to prosecute the problem as they see fit – the power to marshal and redeploy resources, and the freedom to pivot?

We think a true mission must conform with the third model, though we find it hard to pinpoint many examples from recent history where the government has done this successfully.4 If this were what DSIT is aiming for in their AI for science missions (and we think it should be), we would propose the following as non-negotiables.

A mission needs legible authority: a responsible owner with a strong mandate and a direct line to ministers. It needs clear delegated decision-making, including on spending; excellent situational awareness of who is doing what, including outside government, and an understanding of where external actors must be drawn in. It needs a control hub – an organising mind capable of seeing the whole board and directing resources accordingly. It needs the right expertise deployed in the right roles. And it needs enough spending power to get the job done.

This isn’t a novel prescription. Vallance’s 2018 mission principles call for essentially this: a single accountable leader, a dedicated cross-Whitehall team of around eight to ten people, ring-fenced funding,5 and simplified governance. If Vallance still wants to follow these principles (and we think he should) then he should appoint mission leads with the requisite technical expertise and empower them accordingly.

These principles also give us a practical test for whether a proposal is genuinely mission-shaped. If you can’t imagine a single person being held accountable for whether the target was met, the mission is probably too diffuse. If the delivery team would need to span so many domains that no group of ten people could cover it and act as the strategic brain, it is probably too broad. The drug discovery mission, whatever its eventual governance arrangements, could plausibly be run this way. That won’t be true of every proposal.

8) Political durability

Can the mission survive political changes?

A mission that’s analytically perfect but politically fragile is a bad mission, because it will be defunded or hollowed out long before it delivers. Missions require sustained political commitment across spending reviews and ministerial reshuffles – a mission that can’t sell itself to a future minister who wasn’t involved in its creation is living on borrowed time.

There are several ways to build this resilience. The first is structural: embedding the mission within the ecosystem – weaving it into the fabric of multiple strategies such that it becomes difficult to unpick without disturbing something else, and is in the interests of many. The drug discovery mission is notable for how many existing government commitments it connects to. The Industrial Strategy, the Life Sciences Sector Plan, the Life Sciences Healthcare Goals, the strategy for replacing animals in research – all converge on the same set of objectives. This reflects the fact that drug discovery sits at a point of unusually high strategic density, where multiple policy goals are served simultaneously. International commitments and private sector co-investment can serve the same anchoring function: the more a mission is entangled with obligations that extend beyond a single department’s discretion, the harder it becomes to kill it.

The second is narrative. A mission that tells a good story and can be easily understood and defended has a structural advantage over one that’s more abstract or bland. Why does this matter to people’s lives? Why does it matter for Britain? A credible link to economic growth, public health, or national security is a load-bearing feature of the mission’s design.

The third is political constituency. Missions that create place-based effects generate local champions. Missions that attract cross-party support and ownership survive changes of government. Missions that connect to long-standing public priorities – health, security, cost of living – are electorally salient. None of this is a substitute for scientific merit, but scientific merit alone has never been sufficient.

Without an active champion (or team) invested in a mission’s survival, politics tends to be neglected entirely. Many propositions that could otherwise make plausible AI4S missions won’t naturally be politically sexy – and that can make them vulnerable to shifting political winds. The answer is not to ignore this reality, or to retreat into academia and hope that distance from the political battlefield confers safety. Though not necessarily mission-critical, any proposal should engage seriously with the political context it inhabits. The question for each candidate is stark: how many different reasons does the government have to care about this in five years’ time?

All to play for

There is plenty of reason to doubt that these missions will come off. The budget is modest, the institutional capacity to run missions properly is uncertain, the gravitational pull of the prevailing ways of working is intense. Missions are easy to announce, and brutally hard to deliver.

These tests are demanding because missions are demanding. It’s not lost on us that many of the most successful historical examples were driven by necessity: wartime pressures, the Space Race, the COVID-19 pandemic. These missions operated on a fundamentally different scale, but it would be naive not to recognise that maintaining critical mass around a mission may be impossible without the crushing pressure of a national crisis.

And failure may be not neutral, but negative. £137 million is a modest sum for national missions, but it could go a considerable distance in driving AI adoption across the research base. The opportunity cost of squandering it on a programme that collapses into business as usual would be a serious setback for AI for science – and for the credibility of ambitious government action more broadly.

And yet. It would be bloody brilliant if a mission worked. If the government can hold its nerve – apply rigorous selection criteria, resist the incentive to dilute, build up the organisational muscle, and refuse to let the programme degrade into a collective delusion – there is a chance. Britain needs to be able to do this.

AI for science is a great arena in which to recover the habit of national ambition, even if the budget suggests a country tiptoeing where it ought to stride. A well-executed mission programme would count twice over: once for the scientific progress, and again as a demonstration that the British state can still marshal resources towards a formidable goal and see it through. With sufficient will from the right people, we’ll see.

Our thanks to Sam Currie, Alex Chalmers, Alvin Djajadikerta, Rory Byrne, and Charlie Harris.

Cover image: The First Step, by František Kupka, 1910-13

Subscribe now

(Reminder: the framework splits the R&D budget into funding for three distinct types of activity: (i) curiosity-driven, fundamental research; (ii) research that’s aligned to government or societal priorities; (ii) business-led innovation.)

In early February, UKRI CEO Professor Sir Ian Chapman appeared before the Science, Innovation and Technology Select Committee to explain the new framework and to address concerns about funding pressures at the Science & Technology Facilities Council. The session offered the first opportunity to see how the bucket framework holds up under scrutiny – not just as a set of budget tables, but as a governance settlement that the sector can understand, that parliament can hold to account, and that connects public investment to research outcomes.

This piece argues that the Explainer and the subsequent exchanges reveal a gap between what the bucket framework claims to be and what it currently does. The framework classifies budgets, but it does not yet govern research. And the consequences of that gap are already becoming visible in the research base.

In November, before the Spending Review and Explainer were published, I proposed ten tests that the bucket framework would need to pass to be meaningful. These tests weren’t meant to be gotchas; I was clear in the piece that the bucket framework has real potential to help with coherence and prioritisation in the science budget. Rather, my tests were presented as the minimum conditions for the framework to function as a governance settlement rather than just a taxonomy. In sum: does it state what problem it solves? Does it specify distinct governance modes for each bucket? Does it provide a doctrine for research transitioning between buckets? Does it create a credible mechanism for protecting discovery?

The Explainer is the first opportunity to assess how the framework has been operationalised. It engages seriously with almost none of these questions.

What the Explainer says

A fair summary first. The Explainer allocates UKRI’s £38.6 billion across four categories. Bucket 1, curiosity-driven research, receives £14.5 billion. Bucket 2, strategic government and societal priorities, receives £8.3 billion. Bucket 3, supporting innovative companies, receives £7.4 billion. A fourth category, “enabling and strengthening UK R&D,” receives £8.4 billion for infrastructure, talent, institutes and facilities.

Each Industrial Strategy sector gets a Senior Responsible Owner (SRO), drawn from the executive chairs of UKRI’s councils, who leads a cross-UKRI programme board. In addition, we get an extremely high-level breakdown of what’s going into each bucket (how much per sector, how much for QR, how much for compute, and so on.)

Should we wish to compare these numbers with what was being spent before, the Explainer is explicit: “it is not possible to directly compare these allocations to previous budgets” (and we will return to this).

What the Explainer does not contain is any governance architecture. There is no description of how the buckets operate differently from one another, nor is there any account of how boundaries between them are managed. There is no discussion of how risk appetite or accountability varies across modes, nor are we provided with a doctrine for what happens when research transitions from one bucket to another.

It is, essentially, a set of budget tables with brief contextual framing. Further strategic documents are promised for spring 2026 – for now we have the high-level numbers and that’s about it.

Subscribe now

What the select committee revealed

Chapman’s appearance before the committee is worth examining because it illuminates problems that the Explainer’s careful prose obscures.

The most important is comparability of budgets over time. The Explainer says it’s impossible – but if a framework cannot be measured against its starting point, then it cannot be held accountable for its results. Under questioning from the committee chair, Chi Onwurah MP, Chapman’s position shifted: comparison was “not possible,” then “possible, of course, but it would take a lot of effort and backwards bookkeeping,” then he offered a high-level estimate anyway – roughly 50/25/25 between curiosity, applied research, and company support (setting aside the enabling layer) – and conceded that grant-level data exists and could in principle be attributed to the new buckets. The committee asked him to return with both the high-level estimates and a fuller explanation of why more detailed comparison is difficult. He should.

The 50/25/25 estimate is itself revealing. If the historical split is “basically the same” as the new allocation, what has the reform actually changed? Chapman’s answer was indeed all about governance: coherence, accountability, and single programme owners who can be held responsible. That is a reasonable ambition that we should support. But the Explainer’s money tables don’t measure up. We may get more detail in spring 2026. Until then, we have a reform that, by the CEO’s own account, hasn’t changed allocations, and hasn’t yet specified the governance it promises. As Onwurah put it: “We are giving you a pass for where you are now.”

There was also a brief exchange about the fourth bucket – the £8.4 billion enabling layer covering infrastructure, talent, and international subscriptions. Chapman described it as something that “just pro-ratas to the other three.” But we should push back on this – a quarter of UKRI’s total allocation cannot be dismissed as a proportional residual. Infrastructure investment, doctoral training, and international facility subscriptions each have their own particular justifications and approaches, which do not distribute evenly across curiosity, strategy, and business support. As I argued in November, cross-cutting enablers need their own governance layer, not an accounting afterthought.

QR: the central exhibit

The single most consequential decision in the Explainer is the classification of quality-related research (QR) funding. All £8.9 billion of it, over the SR period, is placed in Bucket 1: curiosity-driven research.

This decision shapes everything else. With QR included, Bucket 1 is the largest category by a wide margin – £14.5 billion, nearly double Bucket 2’s £8.3 billion. Without QR, Bucket 1 would be approximately £5.6 billion: still substantial, but no longer dominant, and uncomfortably subscale relative to the other buckets.

The political utility of this classification appears straightforward. A £14.5 billion curiosity-driven budget allows the government to say it is investing generously in blue-skies research. Without tucking in QR, it’s much harder to sustain that narrative – particularly when the fastest-growing budget lines are all in Buckets 2 and 3.

But there’s a problem with this. Under questioning at the select committee, Chapman used QR as his primary example of why backward comparison is difficult. “If I said, all of the money that we spend on quality-related research,” he told the committee, “that is for sure contributing to all three buckets.” He continued: “It represents £10 billion or £11 billion of our spend. It goes in multiple different directions.”

So parliament is now told that QR contributes to all three buckets – just weeks after UKRI published a document placing all QR in Bucket 1. The UKRI CEO was simultaneously defending the classification and undermining it. The committee did not quite catch the significance of the admission. Readers of this piece should.

QR is allocated through the Research Excellence Framework. REF assesses research of all types – not just curiosity-driven research – and spends enormous effort measuring the impact of that research: its real-world effects on policy, industry, health, and culture. Impact case studies, which account for a quarter of the REF assessment, reward exactly the kind of strategic, applied, partner-engaged work that DSIT says it wants more of in Buckets 2 and 3. Universities use their QR funding to sustain research groups working on everything from clinical trials to industrial partnerships to policy analysis. Classifying all of this as “curiosity-driven” is a massive fudge.

The Explainer gets around this by never defining “curiosity-driven.” But the term is not ambiguous. Curiosity-driven research is research motivated by the intrinsic desire to understand something. The government chose that label, not “investigator-led” or “non-commissioned” or “responsive mode,” presumably because it carries a particular weight: the image of foundational inquiry, of researchers following questions wherever they lead. Having chosen the label, the Explainer then uses it to mean something quite different: research that doesn’t fall within a government-identified priority area. That is a description of where the funding sits in UKRI’s portfolio, not of the research it supports.

This might seem like I’m arguing the semantics of government comms. But if the buckets are meant to be about improving the way we actually do research in the UK, then the labels need to describe the research accurately. A framework that cannot tell the difference between “the government hasn’t prioritised this” and “this was motivated by curiosity” does not understand what it is governing.

QR is a funding stream that supports the capacity of universities to self-organise around government priorities without being centrally directed. The civic university agenda, industrial partnerships, place-based innovation strategies, and addressing societal challenges – all of this happens through institutional initiative, funded significantly by QR. If all of this sits comfortably under “curiosity-driven,” then we have hollowed out a word that means something real – about motivation, about inquiry, and about what it feels like to follow a question into the unknown – and replaced it with an administrative residual category. This is ugly. And the flip side is that the system gets no credit for strategic delivery that isn’t centrally directed, which in turn makes the case for ever more central direction look stronger than it should. This is dangerous.

How did we end up here? Through a convergence of institutional convenience that nobody has an incentive to disturb. The sector, led by the Russell Group and others, has advocated for QR as discovery funding for years, presumably because framing it as foundational research is seen as the safest political argument for protecting it from cuts. The Russell Group’s advocacy materials lead with graphene, genomics, and cosmology – science as pure as the driven snow. Additionally, HM Treasury’s internal spreadsheets classify QR as “core research”, presumably because it fits nicely into the 1980s basic research market-failure-response mindset that permeates Whitehall, but also because it’s formula-based and not centrally directed. So DSIT naturally follows suit, slotting QR into Bucket 1 because both the sector and the Treasury have already told them that’s what it should sit – no harm, no foul.

Nobody in this chain has an incentive to say: “QR funds a huge amount of strategic, applied, and commercially oriented activity – and that’s a feature, not a bug.” But the Russell Group’s own position gives the game away. Their December 2024 briefing recommends that “when new R&D commitments are introduced, QR funding should be uplifted to ensure the research base is supported to deliver on new priorities.” Wait a minute – if QR were purely about curiosity, why would it need uplifting for strategic priorities? The recommendation implicitly concedes that QR is delivery capacity for government objectives – which of course it is, but this is precisely the reason it shouldn’t be classified as though it were all blue skies.

The governance consequence of this misclassification is significant. If QR is treated as undifferentiated curiosity, there is no mechanism within the bucket framework for connecting problems in the research base – declining momentum in strategically important fields, for instance – to the funding instrument that actually sustains university capacity in those fields. The framework creates a structural disconnect between the diagnostic and the instrument that could respond to it.

Subscribe now

Quantum: the mirror image

If QR illustrates the problem of strategic research being classified as curiosity, quantum technologies illustrate the reverse. Quantum receives over £1 billion across Buckets 2 and 3 in the Explainer. There is a National Quantum Strategy, a ten-year programme track record, and an SRO in the form of Charlotte Deane, EPSRC’s executive chair. By the framework’s logic, quantum is strategic.

But much of the underlying physics – error correction, topological quantum states, fundamental materials science – is as curiosity-driven as anything EPSRC funds through responsive-mode grants. And if the National Quantum Strategy didn’t exist, I’d bet that much of this work would be happening anyway, funded through standard EPSRC grants, and classified as Bucket 1. So the bucket follows the administrative wrapper, not the character of the inquiry.

Chapman told the select committee that quantum is his “poster child.” He described it as having “a really clear strategy,” internationally differentiated companies, and a ten-year track record of STFC, EPSRC and Innovate UK working together coherently. For what it’s worth, I strongly agree with Ian on this – if Bucket 2 is to make any sense at all, then UKRI could do much worse than looking at the National Quantum Technologies Programme as a benchmark for how to invest strategically in a field.

But analysis of UK research share across 4,516 topics using OpenAlex publication data tells a more complicated story. Quantum information and cryptography shows the UK losing significant ground to our global competitors: the UK’s share of publications in this field has dropped significantly from a high of over 8% to around 6% today.1 Publication share is not the only measure of research health, of course, but it is a signal that warrants explanation, and the bucket framework currently has no mechanism for providing one.

UPDATE 28/3: This is an updated graph. See the STFC section below for a discussion.

The obvious pushback is: quantum has a multi-billion pound programme across Buckets 2 and 3, with an SRO, a ten-year track record, and investable companies to show for it – so what exactly is the problem? The problem is that both things appear to be true simultaneously. The programme is delivering on its own terms, while the UK’s position in the global quantum research landscape appears to be deteriorating.

There are several possible explanations. The programme may have deliberately shifted effort from publications toward commercialisation, in which case declining research share is an acceptable trade-off – and someone should say so explicitly. Or the programme may be successfully translating existing research while the pipeline of new foundational work thins underneath it, in which case there is a problem that will not be visible until the pipeline runs dry. Or there could be some other explanation – I suspect we don’t know.

Whatever the reason, the buckets framework provides no way of asking the question. All we know is that the SRO owns the programme and is accountable for programme outcomes. For quantum, the SRO just so happens to be the executive chair of EPSRC, the council which funds the bulk of upstream and downstream quantum research. But even if this led to all kinds of accidental synergies, Charlotte Deane isn’t structurally accountable for the relationship between the directed programme and the broader health of the research field that feeds it – a relationship that spans the boundary between Buckets 1 and 2. As we have seen, the bucket framework has no mechanism for managing that boundary.

This matters because Chapman is presenting the quantum SRO model as the template for all Industrial Strategy sectors. If the model works by delivering coherent programmes while remaining blind to what is happening to the foundational research underneath, then generalising it across the system embeds that blindness everywhere. Because Charlotte Deane is managing quantum in both bucket 1 and bucket 2, this may end up sorting itself out in practice; but in other areas (such as advanced manufacturing or clean energy) this won’t necessarily apply.

STFC: what happens when the framework meets reality

On 28 January, Michele Dougherty wrote to the particle physics, astronomy and nuclear physics (PPAN) community with difficult news. STFC’s PPAN budget would need to fall to around 70% of its 2024-25 level. Individual projects were being asked to model flat cash and reductions of 20%, 40% and 60%, and to identify the funding point at which they become non-viable.

Four days later, Chapman published an open letter to the wider research community. It was admirably direct. STFC’s core budget is holding roughly flat – £835 million to £842 million over the SR period – but costs have risen beyond what the settlement can sustain. Energy prices and exchange-rate movements alone added over £50 million a year, and an ambitious programme from the previous Spending Review had accumulated commitments that the current budget cannot cover. £162 million in cumulative savings must be found by 2029-30.

Credit where it is due. Both letters represent a quality of directness and personal ownership that is rarer in UKRI communications than it should be. Chapman signs his letter “Ian.” Dougherty asks the community to “work with us” while being clear about the severity of what is coming. These are leaders grasping the nettle of a problem that’s accumulated over years. Chapman is explicit that “the situation at STFC is unique among the UKRI councils” and should not be confused with the broader bucket reforms.

But we should notice what the bucket framework contributes to understanding this situation: absolutely nothing.

STFC’s activities span multiple buckets. Its applicant-led PPAN grants sit in Bucket 1. Its international subscriptions – CERN, the European Southern Observatory – sit partly in Bucket 4. Its facilities and national labs seem to straddle Buckets 1, 2 and 4. Its applied and innovation work feeds Buckets 2 and 3. When STFC faces a £162 million cost pressure and must cut PPAN grants to around 70% of their recent level, which bucket is actually absorbing the pain? The honest answer is that the buckets are invisible at the point where the trade-off is actually made. The prioritisation happens within STFC, impacting activities that span across the entire framework, and mediated by the executive chair’s judgement and the Science Board’s advice.

Dougherty’s letter even acknowledges this implicitly. “These difficult funding decisions should be viewed alongside emerging opportunities including new investments in digital infrastructure and compute, and future funding through other UKRI buckets, which could benefit the PPAN community.” The buckets are being invoked as a consolation – perhaps the community can win money back through Bucket 2 or 3 – rather than as a framework that governs the decision itself.

This is the comparability problem in live action. The committee chair asked how parliament can scrutinise changes in spending if backward comparison is impossible. The STFC situation shows that even forward transparency is compromised. The actual decisions happen below the bucket level, inside councils, where the old system of disciplinary portfolio management still operates. The buckets are a reporting layer only.

And then there is the “protected” claim. Chapman’s open letter repeats the central assurance: “support for curiosity driven research is protected across the SR period, comprising around 50% of our investment.” Dougherty’s letter frames PPAN as “now operating fully within STFC’s curiosity-driven research portfolio.” So: curiosity is protected. PPAN is curiosity. And PPAN is facing being cut to 70%. All three statements are held to be true simultaneously.

The resolution is that “protected” means the aggregate Bucket 1 number holds. Within that aggregate, individual disciplines can be cut dramatically – particle physics, astronomy, nuclear physics – as long as the total stays roughly constant. The bucket protects a budget line, not the research it contains.

To be clear: this is not a criticism of the specific decision to reduce PPAN funding. That may well be the right call given STFC’s cost pressures, and Chapman is absolutely right that hard choices are unavoidable within a constrained envelope. My criticism is that the framework’s definition of “protected” is so coarse-grained that it is compatible with 30% reductions in specific disciplines while maintaining the headline assurance that curiosity-driven research is safe. Speaking at a Campaign for Science and Engineering event earlier this week, Chapman stressed that UKRI is still in a solicitation phase, gathering impact information and priorities before final decisions are taken through council and advisory routes. But given the scale of cuts under consideration, “curiosity-driven research is protected” may turn out to mean something much less reassuring than it sounds.

The bibliometric data adds an uncomfortable dimension. The STFC disciplines are precisely the areas where UK research share is already in rapid retreat. Particle physics saw its UK share fall from approximately 4.7% in the 5-year period before 2021 to 2.3% the period to 2024, with momentum reversing from slightly positive historically to clearly negative in recent years. In the ‘galaxy formation and evolution’ research topic, UK share fell from around 15.5% to 6.5%. These trends predate the current cuts. The PPAN reductions will likely accelerate declines that were already under way. Perhaps that’s OK. Perhaps the UK doesn’t need to participate as much in big physics as we once did, and the STFC cuts simply come in the wake of that hard fact. But the bucket framework has no mechanism for flagging this interaction. There is no system-level process that says: particle physics is in rapid retreat, and now we are proposing a 30% budget reduction on top – is this a deliberate strategic choice, or an unintended consequence of cost pressures elsewhere in STFC? The decision is being made inside STFC’s portfolio, not at the system level where the interaction between strategic intent and research trajectory could be examined.

UPDATE 28/3: After publishing this post, I discovered that the OpenAlex snapshot I used (November 2025) contained a data quality error: a number of arXiv preprints spanning multiple years had been given an incorrect publication date of 2022. This error in the input data inflated the denominator for all rolling five-year windows that included that year. This issue distorted the picture for a few disciplines – including the two PPAN topics I’d highlighted, producing a much more negative picture than is the case. I have now recomputed the charts with a newer OpenAlex snapshot with that error fixed.2

The updated graphs are given below. What this shows is that these two PPAN disciplines are, according to the UK’s share of (country-affiliated) global publications at least, in relatively good health. In my view, this adds another dimension to the ongoing discussions between the PPAN community, STFC, UKRI and others. I will return to this at a future date, as there have been a number of developments in the weeks since I published this piece.

I am very grateful to those in the PPAN community who got in touch about this.

Dougherty’s letter does promise that “UKRI and STFC will continue to monitor the health of disciplines and maintain flexibility to adapt investment plans.” This is welcome language. But there is no published framework for what monitoring the health of disciplines means in practice, what metrics are used, what triggers escalation, or who is accountable if a strategically important field enters terminal decline while remaining nominally “protected” inside Bucket 1.

The STFC situation illustrates a general point. The bucket framework operates at a level of abstraction that is disconnected from where prioritisation actually happens. Real trade-offs are made within councils, between disciplines, between facilities and grants, between inherited commitments and new opportunities. The buckets classify the results of those trade-offs after the fact. They do not govern them.

So what are the buckets actually doing?

At the CaSE event this week, Chapman offered the clearest rationale I have yet heard for why three buckets might matter: if UKRI stops forcing every proposal through multiple lenses at once, then curiosity-driven funding can take more genuine risk, rather than converging on a “grey middle” of projects that are simultaneously publishable, impactful, and plausibly growth-adjacent. That is a serious ambition.

And in fairness, if the buckets function simply as a way for UKRI to organise its internal budget management – knowing when to be more or less hands-on – they may still prove useful. The government never explicitly promised distinct governance modes with different logics for agenda-setting, accountability, and risk tolerance: that was the standard I argued for in November, not one that DSIT set for itself.

But the bucket framework did not arrive as a modest internal management tool. It was presented as the most significant change since UKRI’s establishment, restructuring the entire budget as the operational expression of the Industrial Strategy, with explicit demarcation and protection for curiosity-driven research. When tested against the weight of this, what we find is a series of fudges: the strategic work that QR supports gets classified as curiosity-driven, quantum is curiosity-driven research classified as strategic, and STFC disciplines are “protected” research facing a cut of 30%.

As things stand, the buckets are not really about research at all – they are about administrative control. “Curiosity-driven” means money that the state doesn’t direct. “Strategic” means money wrapped in a programme with an SRO. These are descriptions of the funding approach only, not of the research it supports. A quantum physicist working on error correction is very likely doing curiosity-driven research by any reasonable definition, but it sits in Bucket 2 because there is a programme around it. A university research group using QR funding to partner with an NHS trust on clinical data science is doing strategic, applied work, but it sits in Bucket 1 because the state is hands-off.

I do not think my outstanding questions are unfair: who manages the boundary between Buckets 1 and 2? Who monitors whether foundational research fields are strengthening or weakening? Who connects research trajectory to strategic intent?

The old system did not answer these questions either, but at least it did not claim to. The risk I identified in November was that, despite the potential for it to illuminate our understanding of the research base, the bucket framework would materialise merely as “Frascati-with-branding” – a relabelled version of existing categories that resolves none of the real tensions. Nothing we’ve seen has yet mitigated that risk.

What would have to be different for this entire thing to work as it should

So where do we go from here?

What follows is not a comprehensive reform proposal. But based on the analysis above, a governance framework that functions as more than a budget classification would need at least five things.

First, QR should be recognised as multi-bucket funding – because that is what it is, as the CEO himself told parliament. This does not mean carving QR up between the buckets, which would be both impractical and undesirable. It means the governance framework should acknowledge that block grant funding sustains capacity across all three modes, and that the Bucket 1 headline should not rest on the fiction that QR is undifferentiated curiosity. QR is, by design, the closest thing the system has to a genuinely enabling and strengthening investment, and classifying it alongside infrastructure, talent, and facilities would better reflect what it actually does. Therefore, QR should be moved to Bucket 4.

Second, the boundary between Buckets 1 and 2 needs hard-edged criteria. When does research move from “curiosity-driven” to “strategic”? Currently the answer seems to be: whenever someone in Swindon creates a programme. Without objective criteria, bucket assignment is a proxy for revealed political preference – the stuff the government has decided to wrap in a programme gets classified as strategic, and everything else is curiosity by default.

Third, someone needs to own disciplinary health. The Explainer creates SROs for each sector programme, and Chapman told the committee there is now “one person responsible” for each programme area. Good. But who is responsible for advanced materials research? For the upstream physics that feeds quantum? For the disciplines that don’t map onto Industrial Strategy sectors? Richard Jones has observed that only around £800 million a year goes directly to research councils for applicant-led proposals – out of an average annual UKRI budget of £9.6 billion. If the councils’ direct stewardship of responsive-mode research is this small relative to the total, and if the cross-UKRI programmes are absorbing an increasing share, then disciplinary health monitoring becomes extremely important.

Fourth, Bucket 4 needs real governance. £8.4 billion of infrastructure, talent, and international investment cannot be dismissed as a pro-rata residual, however convenient that might be for the ongoing presentation of a “three-bucket” framework. These are shared platforms on which all three modes depend, and they require their own accountability and planning logic.

Fifth, comparison simply must be possible. Not because backward-looking accounting is intrinsically valuable, but because a framework that cannot be measured against its starting point cannot be held accountable for its results. The select committee chair was right. Chapman has acknowledged he could provide high-level mapping and has committed to audit trails when specific choices are made. Both should be delivered. The committee should hold UKRI to it.

Close

The bucket framework was a serious proposal that deserved serious implementation. Lord Vallance brought genuine intellectual ambition to the role of Science Minister, and the desire to create a clearer relationship between public money and public purpose is the right instinct. Ian Chapman has been admirably direct about the choices UKRI faces, and his willingness to own difficult decisions personally is a genuine asset.

What the framework got, however, was a budget reclassification that inflates the curiosity headline with QR, sorts directed programmes into new headings, and defers the governance architecture to documents not yet published. Meanwhile, the research base is shifting in ways the framework cannot see: strategic sectors with weakening research foundations, cross-cutting capabilities with no institutional owner, and a position among European peers that is barely net positive.

The spring strategy documents – promised by the Explainer – are an opportunity to remedy this. But they will only do so if they engage with the governance questions that the Explainer avoided: how the buckets operate differently, who manages the boundaries, who monitors what is happening to the research fields that feed the programmes.

The framework might usefully start by internalising a truth that Richard Jones articulated on BlueSky last week: “the motivation for a scientist to work on a particular problem doesn’t have to be (and often isn’t) the same as the motivation for someone to fund it.” The buckets have been constructed entirely around the funder’s motivation – directed or undirected, commissioned or uncommissioned. They are silent on what the research actually is. A governance framework worthy of the name would need to see both sides.

So, to great fanfare and no small amount of surprise, the UK has announced that we will be rejoining Erasmus. Many across the UK higher education and further education sectors are delighted about this “fantastic news”.

Why has this happened? The government wants you to believe this is about “breaking down barriers to opportunity” (Stephanie Peacock MP), “breaking down barriers and widening horizons” (Nick Thomas-Symonds MP), and “breaking down barriers to opportunity, giving learners the chance to build skills, confidence and international experience” (Baroness Smith). Lots of agreement across those three ministerial quotes!

Well, there is certainly something to be said for breaking the link between a child’s background and their future success, and about disadvantaged learners finally getting the chances that were once hoarded by the privileged few. But if we apply an opportunity lens to the Erasmus decision and actually look at the money, the distribution, and the alternatives – the story starts to fall apart.

The weight of the money

The headline figure is that Erasmus will cost British taxpayers £570 million for the first year, with the 30 per cent discount reportedly applying only in that initial year. The Times reports that future annual costs could rise towards £1 billion, and the Telegraph pegs it at potentially £8 billion over the lifetime of the agreement. For some, this sum of money is “practically nothing” for the “huge benefits” it brings. I’d suggest this reflects a certain loss of calibration about what public money means.

There are 3,166 state-funded primary schools in England located in neighbourhoods in the highest child-poverty quintile (IDACI deciles 1-2). Divide £570 million by that number and you get roughly £180,000 per school. Nobody in government is offering that trade – but it’s a useful way to feel the weight of the money. What does £180,000 mean for a primary school in Knowsley or Jaywick? It means teaching assistants, speech and language therapists, family liaison officers, after school clubs, musical instruments, books, sports equipment… the list goes on. Multiplied across the most deprived or “left behind” areas in the country, we’re talking about serious impact.

Or we can consider the government’s own fiscal choices. The Office for Students distributes around £300 million annually to widen access and participation to English higher education each year – a figure that is in decline. The new international student levy, which ministers have explicitly linked to reintroducing maintenance grants for disadvantaged students, is expected to raise around £445 million in its first year.

So: one year of Erasmus is nearly double the English WP budget and comfortably exceeds the entire maintenance grant envelope that the government is creating a new tax to fund. These comparisons are uncomfortable.

And what about the old HEFCE supplement that supported students on overseas exchanges – around £2,315 for full-year placements? This was finally scrapped in the OfS allocation for this academic year. Are we simply robbing Peter to pay Paul?

Subscribe now

The distribution question nobody wants to answer

Defenders of Erasmus rightly point out that it’s not just for university students. The scheme covers schools, FE, vocational training, adult education, youth work, and grassroots sport. The vocational and schools elements are real and potentially valuable.

But breadth of eligibility is not the same as breadth of uptake. The question is not “can Erasmus fund a hairdresser in France?” but “what share of UK Erasmus participation will actually be vocational versus university? And within each sector, what share will be disadvantaged learners?”

We know something about this from the literature, and it’s not reassuring. A JRC working paper on Erasmus participation in UK higher education found a negative relationship between the share of disadvantaged graduates at an institution and its Erasmus participation rate. In other words, universities with more disadvantaged students send fewer of them abroad.

A UUKi report on widening participation in mobility put numbers on the gap: in 2015-16, students from higher socio-economic backgrounds were sixty-five per cent more likely to be outwardly mobile than their disadvantaged peers. Care leavers participated at barely half the sector average rate.

This shouldn’t be surprising. Mobility requires capital – financial, yes, but also cultural. You need to know the scheme exists. You need to believe it’s “for people like you.” You need the confidence to apply, the resilience to navigate bureaucracy, the family circumstances that allow months away. The 2012 Riordan report on barriers to mobility found that money and language were the headline obstacles, but these were reinforced by curriculum inflexibility, patchy credit recognition, and thin institutional support. I’m sure lots of this was improving in the years before Brexit, but we shouldn’t forget that any system that makes mobility feel like an optional add-on for confident, well-off students rather than a normal part of study will reproduce existing advantages.

Yes, some working-class students do go abroad and find it transformative. I don’t doubt the individual stories – nor do I question that important work happened in non-elite institutions to coordinate Erasmus in a way that served disadvantaged students. But the existence of working-class beneficiaries doesn’t tell us about aggregate distribution. If 15 per cent of participants are from disadvantaged backgrounds and the scheme is life-changing for them, that’s still a scheme where 85 per cent of the spend goes elsewhere. The question is never “can this help some disadvantaged people?” but “what share of the spend reaches them, and is that share better or worse than alternatives?”

Which brings us to the comparison the sector doesn’t want to make.

The Turing counterfactual

The Turing Scheme was created after Brexit as the UK’s own outward mobility programme. It has been relentlessly criticised – too small, too transactional, and too much a reminder of what we left behind after we left the EU. But here’s what Turing actually delivers, according to DfE’s published data for 2025-26: of 35,248 placements, 61 per cent (21,411) went to disadvantaged participants, at roughly £2,500 per placement overall.

Crucially, those numbers are published. Turing reports actual spend, actual placements, disadvantaged share by sector, and geography. We can scrutinise it, ask hard questions about it, and see whether it’s working.

The government says Erasmus will benefit “up to 100,000” people annually. Crudely split, that’s £5,700 per head – more than double Turing’s unit cost – but we get no transparency on who those participants actually are or what share will be from disadvantaged backgrounds. Perhaps this will all come out in the wash – but right now we’re being asked to spend nearly twice the widening participation budget on a scheme whose distributional impact is both questionable and poorly evidenced.

The structural problems Erasmus won’t fix

As Jim Dickinson notes on Wonkhe, the UK’s problem was never just low participation overall – it was patterned inequality with compounding effects. Students with overlapping disadvantages had even lower rates. And the barriers are structural: tightly specified curricula, professional accreditation constraints, the UK’s general failure to do credit transfer properly, and insufficient support infrastructure in many institutions.

Erasmus membership alone won’t fix any of this. Without curriculum reform, functioning credit transfer, or requiring outward mobility to feature in access and participation plans, we’ll likely see the same skewed uptake we saw before Brexit – just at greater public expense.

Dickinson is right that if ministers really want this to work as an opportunity intervention, they need a proper strategy inside DfE, proper incentives or targets for universities, and progress on the structural blockers. Otherwise, a future government will look at disappointing participation figures – in volume or access terms – and conclude the money was wasted. Given what we know about who historically participated in Erasmus, that conclusion may not be wrong.

Why is this happening?

If the opportunity case is weak, why is the system so convinced that this is a great idea?

A cynical reading is that the people making and influencing these decisions are disproportionately the people who benefited from programmes like Erasmus. When your career has moved between Russell Group universities, Brussels think tanks, and Whitehall policy roles, European exchange doesn’t feel like a class-skewed luxury – it feels like the essential infrastructure of an educated life. Does the possibility that your own formative experience might not be the highest-return use of public money even compute?

That’s unfair, of course – but the sector’s criticism of Turing was also very revealing. It was framed almost entirely as “less money than Erasmus” and “Brexit damage” rather than engaging with whether Turing’s design principles – explicit disadvantage targeting, and clearly published outcomes – might actually be better for an opportunity agenda. Turing made the distributional choices visible and therefore contestable. Does Erasmus simply launder them through the warm glow of European solidarity?

The honest version of this debate

Rejoining Erasmus is a legitimate diplomatic and geopolitical choice. It signals alignment, repairs relationships, brings the UK back into European educational networks, and has genuine soft power value. The institutional benefits are real, and so I should not be surprised that sector leaders are delighted about all of this. But if that is the argument, let’s make it openly and let it be judged on those terms.

The opportunity case is, at present, much thinner than the rhetoric suggests. It invites an audit. On outcomes, Turing currently looks like the more effective tool: majority participants from disadvantaged backgrounds, full transparency, and clearer sectoral targeting – and at less than half the unit cost.

If ministers want to make the opportunity case for Erasmus, they should commit now to the same transparency standard: actual spend, actual placements, disadvantaged share, sector split, and geography. They should require outward mobility to appear in access and participation plans. And they should explain why they cut the existing overseas mobility supplement in the same year they signed a £570 million cheque to Brussels.

Otherwise, this whole thing is a gift to politicians who want to reason, in a motivated way, about whose opportunities our systems and institutions really care about.

The government has started to talk seriously about putting public R&D funding into three “buckets”:

1. “basic curiosity-driven, investigator-led research”;

2. “applied research which should be aligned to Government ambitions”; and

3. “helping the transition from start-up to scale-up” and “support[ing] our big R&D-intensive companies”.

Starting life as comments made by science minister Lord Vallance, this framing recently made its way into the Post-16 Education and Skills White Paper. While this framing may initially sound like a useful shorthand for capturing the breadth of our R&D system, the formalisation in the White Paper shows it is much more than that; it’s part of a wider effort to connect research, skills, industrial strategy and economic renewal.

That gives the buckets extra weight: positioning them as an organising device for how ministers intend to steer the entire research-innovation-skills chain. The three buckets could even frame how UKRI allocates its budget, how departments think about their R&D portfolios, how universities and other research organisations position themselves, and how companies interact with the state.

The three buckets are also an explicit intervention in the long-running Haldane Principle conversation: the boundary between ministerial direction and academic autonomy.

There are reasons to welcome clarity. An explicit ‘discovery’ bucket could stop the recurrent tendency of governments to meddle in areas where light touch is both appropriate and politically defensible. A government-aligned bucket creates space for more coherent strategic action. A business-led bucket could signal that innovation policy matters at least as much as research policy.

But there is an obvious risk. This could become nothing more than Frascati-with-branding: a relabelled version of the basic/applied/experimental development split that the OECD has used for half a century. Critically for our discussions, Frascati is descriptive, not normative: it tells you what you have, not how it should be governed. If the UK simply maps existing programmes into three headings, the result will not resolve any of the real tensions: who sets priorities, how you govern complex institutes like LMB or Turing, how you protect discovery-led research at Spending Reviews, or what “business-led” even means when most relevant levers sit outside DSIT.

There is also the danger of creating a framework that simply gives future Chancellors a more legible way to shift money out of discovery and into short-term priorities. The question is how to secure the benefits of this approach safely.

To do this, ‘bucket theory’ needs to work as a governance framework: distinct modes of agenda-setting, accountability and instrument design. That is, admittedly, a high bar; most government restructurings fail this test. So to help us, we need to ask a sharper question: what conditions must be met for ‘bucket theory’ to be meaningful? What would have to be true for buckets to improve the functioning of the UK’s R&D system, rather than add another layer of classifications and acronyms? And what capabilities and constraints distinguish one bucket from another in practice?

If buckets are to become a meaningful settlement, they must pass some hard tests about governance, institutions, incentives and political economy.

In the remainder of this piece, I set out the tests I think matter most. They are not exhaustive, but in my view they are necessary. If the three-bucket framework cannot meet them, it risks becoming an elegantly structured way to misdiagnose, and potentially worsen, the UK’s long-standing challenges in research and innovation.

Subscribe now

Test 1: Does bucket theory solve anything ministers actually care about?

Before you can judge the usefulness of the three buckets, you need to know what problem they are meant to solve. In my interactions, I perceive at least four different motivations:

1. Government cannot direct research towards national priorities with enough grip.

2. Business investment in R&D is low and must driven up significantly.

3. The system is too messy: nobody can describe what sits where or why.

4. Government knows it is funding lots of blue-skies research, but given the state of the economy, it has no idea whether it’s funding the right amount.

Each of these is a legitimate problem. But each implies a different remedy. If the concern is that missions need more coherence, the question might be not be about buckets at all – it is about how to get as quickly as possible from hypothecating the funding to running effective departmental R&D governance. If the concern is weak business R&D, the answer probably lies in improving tax policy, procurement, standards and regulation, none of which fall neatly into Bucket 3 unless you radically enlarge the definition of “public R&D”. If the real concern is that our university research system is too large or insufficiently steered, then bucket theory is all about giving ministers a neat mechanism for shifting money from Bucket 1 into Buckets 2 and 3 (and no need to pretend otherwise).

A bucket framework that does not clearly state the problem it is designed to address will quickly become everyone’s favourite justification tool: every spending review, every ministerial push, every institutional turf battle will try to claim the buckets support their preferred narrative.

A serious bucket settlement must therefore begin by answering one simple question: what specific system failures or political dilemmas is this intended to fix? Unless that is stated upfront, no other part of the framework can carry weight.

Test 2: Do the buckets correspond to genuinely different modes of governance?

The second test follows immediately from the first. If bucket theory is meant to change how decisions are made, each bucket must embody a distinct mode of governance. That means each must have:

• a different logic for agenda-setting (who decides the “what” and “why”);

• a different accountability structure (to whom, through which forums, and with what consequences);

• a different tolerance for risk and failure;

• a different suite of instruments and operating rhythms or tempi.

Without these differences, the buckets are nothing but “purposes”: Bucket 1 “exists to expand the frontiers of knowledge”, Bucket 2 “exists to advance national missions”, or Bucket 3 “exists to support business innovation”, or whatever the wording is. But those are statements of intent, not operating models. The real distinction lies in how power flows:

• In Bucket 1, is government genuinely hands-off on content, intervening only in the architecture of the system (portfolio balance, infrastructure decisions, public appointments)?

• In Bucket 2, can ministers articulate outcome-based missions and hold programme directors to account without drifting into micromanagement or political signalling?

• In Bucket 3, is the state clear about which market failures justify its participation – and which instruments align with those failures?

A useful thought experiment here is the Alan Turing Institute, which has recently been the subject of very Bucket 2-style intervention. But the Turing is an independent institute, funded by UKRI – whose website asserts their decision-making independence from government (putting Turing in Bucket 1). So into which bucket does Turing fall? Its activities arguably straddle all three buckets. If the only reliable way to classify it is “where ministers want more control vs. where they fear political blowback”, then bucket theory has failed the mode test. Our buckets become a map of revealed political preference, not a structure that disciplines it.

Or perhaps we should look at the REF. The White Paper implies that institutions will be rewarded for bucket-aligned behaviour – demonstrating clarity of purpose, alignment with government priorities, or measurable impact (one might read this as a sector-friendly version of the three buckets). This is indeed a mode shift of sorts! But it is one that leaves open major questions on how this will work to differentiate university activity (and consequent funding) across the buckets. No small challenge.

The severity of this second test is intentional. If buckets do not lead to visibly different behaviours – in programme design, accountability, tempo, risk appetite, and system oversight – then the reform is cosmetic. If they do, then bucket theory becomes a way of hard-wiring the differences that most people in the system already recognise, and we might be able to right-size our efforts accordingly.

Test 3: Can the framework cope with research moving between buckets over time?

A further conceptual challenge arises when we try to apply the three buckets to real research trajectories, when today’s discovery becomes tomorrow’s mission priority and next year’s commercial opportunity. Genomics, quantum technologies, AI, new materials, and mRNA platforms have all travelled this path. A static allocation framework built around project types may therefore be inherently unstable.

This matters because it forces a deeper question: what exactly is being ‘bucketed’? Are we classifying projects? Institutions? Funding streams? Capabilities? Patrick Vallance drifts between talking about buckets as priorities and broad categories of research activity; officials mainly think in budgets; and universities might wisely assume it must be about institutional or other politics. These are not compatible views.

If the buckets apply to projects, then most significant research programmes will, by construction, cross multiple buckets during their lifetime. If the buckets apply to institutions, then the framework implies a much more radical restructuring of the landscape than is currently being acknowledged: the Research Councils, PSREs, Catapults, ARIA, national research infrastructure, and major university institutes would need to be designated into specific modes and held to them. And if the buckets apply to budgets, then regular reclassification becomes unavoidable as fields mature, making long-term planning harder.

None of these problems is fatal, but they require explicit design choices. At a minimum, bucket theory needs a doctrine (and protocols) for how research should transition across modes, who adjudicates those transitions, and what happens to the funding and governance arrangements when it does. Without that, the buckets will either become rigid categories that distort the evolution of research, or porous categories into which anything can be moved for convenience – in which case they cease to have meaning at all.

The test here is simple: does bucket theory contain a credible account of how research moves between modes, and what institutional machinery supports that movement? If not, it risks operational incoherence from the start.

Test 4: Can the buckets be mapped onto existing institutions without breaking them?

Even if the conceptual model works, it must still survive contact with the machinery of UK research and innovation – and this is where difficulties could multiply. The country’s institutional landscape is not designed around cleanly separated purposes or modes. Almost every significant organisation straddles two or three of the proposed buckets.

ARIA is discovery-oriented but was explicitly created outside Haldane with unprecedented political risk-taking and, through its Activation Partners, is now engaging in firm formation and scale-up. Catapults are business-facing yet majority publicly funded, often operating pre-competitively, and pursuing goals that are part-technological, part-government mission. Research England’s functions sprawl across all three buckets: QR might be ‘core research’ in the Treasury R&D stack, but in reality props up significant amounts of government and industry funded research. REF’s increased use over the last few cycles as a powerful lever for policy change makes it look ever more mission-oriented. And in a post-HEFCE world, HEIF has inched ever closer to being exclusively a mechanism for business engagement and economic growth. Even the research councils have mixed portfolios that run across discovery grants, strategic programmes, translational pathways, industry partnerships, and major facilities.

To cut through this tangle, bucket theory must specify whether institutions are expected to align to buckets or span them. Both choices carry significant consequences. Alignment implies major structural redesign, where existing organisations are carved into bucket-specific units, or where new machinery is built on top of what we already have. Span implies matrix governance, with explicit rules for which decisions belong to which mode. Neither approach is straightforward, and both risk unintended incentives: institutions forced into a single bucket may lose capabilities; institutions allowed to span buckets may default back to the familiar governance patterns that bucket theory was meant to clarify.

A workable settlement therefore needs to answer at least three questions. First, how will bucket assignments interact with the governance of our existing funders and institutions? Second, where do cross-bucket entities such as large facilities, shared talent funding, PSREs and departmental R&D programmes sit? And third, what prevents institutions in politically attractive buckets from absorbing functions by default?

Unless bucket theory can be mapped onto the real institutions of the UK system without producing internal contradictions or perverse incentives, it will have no operational purchase. That is why institutional mapping is a crucial early test.

Test 5: Can discovery be protected without becoming either illegitimate or an easier target?

Bucket 1 is vulnerable. Missions have ministers. Business-led research has firms, investors and industry groups. Discovery has a diffuse constituency whose political salience is low and whose work is often presented as remote from everyday concerns. Without explicit guardrails, Bucket 1 becomes the easiest source of funds for anything urgent or electorally resonant. The fact we have a minister who ‘gets it’ on discovery research is a telling exception that proves the rule – and even Patrick can’t stop the research community from descending into blind panic at every fiscal event.

One answer might be to introduce a floor or band for discovery spending. But why should frontier research enjoy protected resources when other important public services do not? Can we show the compelling public-interest rationale without indulging in self-declared privilege? I wonder.

And here is the paradox. Creating a formal Bucket 1 does not necessarily protect discovery; but it may make raiding it politically easier. Once clarified and ring-fenced on paper, discovery spend becomes a discrete line item that can be targeted in future fiscal consolidations, particularly if a government faces incentives to shift resources into applied missions or industrial support. A well-defined bucket might prove easier to cut than a diffuse and conceptually interwoven system – and while ministers are clear that discovery research will be protected and, when the economy allows, grown, this is protection-by-assertion.

The science minister himself has illustrated the tension. In recent testimony, Vallance framed Bucket 1 as “knowledge creation and the cost of doing business,” urging that we “not have false promises around economic growth from that in the short term.” That’s fair enough, and loosely echoes a familiar sector talking point about not throwing the golden goose out with the bathwater. But it is also an attempt at protection through lowered expectations: if discovery isn’t held to growth metrics, it can’t be judged to have failed them. This framing cuts both ways, of course, as when budgets tighten, costs must be cut.

There is a related danger in treating Bucket 1 as “set and forget”: a space where government steps back, leaving the research community to steer priorities through a mix of peer review, QR allocation and institutional strategy. This ignores where important governance questions sit. The state already shapes the architecture of discovery: the mix of instruments, the balance between disciplines, the approval and siting of major infrastructure, the incentives for interdisciplinarity, and the metascientific rules that govern quality and reproducibility. These are not questions that can be delegated wholesale to an undifferentiated “academic community” – researchers hold divergent interests, and many of these decisions involve public interest considerations that aren’t synonymous with institutional preference.

The most obvious of these is disciplinary balance. How much of the national research effort should sit in STEM versus AHSS? How much should flow into fields with large spillovers versus fields with high public-value content but fewer market-facing impacts? For ministers responsible for stewarding national research capability, these choices cannot be elided. But our system rarely confronts these questions explicitly, because our sector operates on a shared belief that all disciplines are created equal. Bucket theory does not resolve this problem; if anything, it might obscure it further.

A performatively insulated Bucket 1 may also invite attack rather than deflect it. If discovery is placed behind an institutional curtain and presented as exempt from scrutiny, it becomes an obvious target for future populist governments intent on demonstrating value for money or punishing perceived ideological bias. Trump’s “Restoring Gold Standard Science“ Executive Order is instructive: reproducibility, replication and bias are no longer internal metascience debates but matters of national political attention. A discovery bucket with opaque governance and weak public explanation of quality assurance will be left very vulnerable here.

Thus the real test is dual: can Bucket 1 be protected in ways that are politically sustainable and publicly defensible, while also being governed explicitly enough to address both disciplinary balance and quality – all without collapsing into either politicisation or abdication? Protection cannot be achieved by structural boundaries alone. It requires a publicly defensible account of why discovery matters, how it is governed, and how its quality is assured. Without a visible metascience agenda – covering reproducibility, transparency, risk appetite and evaluation – the bucket will lack the legitimacy needed to withstand fiscal pressure or ideological policing. Any protection that rests on insulation rather than legitimacy will fail at the first encounter with a hostile Chancellor or a populist critique.

Test 6: Can the boundary between Buckets 1 and 2 be drawn in a way that resists political opportunism?

The hardest line to draw in the entire framework is the one between discovery-led research (Bucket 1) and mission/priorities-led research (Bucket 2). The two are not cleanly separated in the real world. Much discovery occurs in institutions that have implicit strategic intent; much mission work depends on fundamental capabilities; and many organisations span the boundary as a matter of good design.

This makes the boundary acutely vulnerable to political opportunism. If the criteria for classifying research as “mission-led” are vague, then almost any activity can be placed into Bucket 2 when ministers want more grip, and almost any activity can be placed into Bucket 1 when ministers want to avoid ownership or accountability. Without hard-edged criteria, bucket assignment becomes a proxy for revealed preference: which programmes ministers want to direct, which they want to shield, and which they want to disown.

That is precisely the pathology a governing framework is meant to prevent.

A workable settlement therefore needs objective tests for Bucket 2 status – tests that apply equally whether the activity is fashionable or embarrassing, electorally attractive or politically toxic. At minimum:

• a clearly articulated, outcome-based mission;

• a time-bounded charter;

• a named political owner; and

• a theory of change linking research to mission outcomes.

If a programme cannot meet these tests, it should not be designated as mission-led. If it can, it should not be allowed to retreat into Bucket 1 when results are disappointing. The same logic applies to institutions: if an organisation claims it is “mission-oriented”, it should be accountable for mission progress; if it claims it is discovery-oriented, it should be insulated from mission-driven interference.

The test here is whether bucket theory can stabilise the 1/2 boundary, preventing both disguised control and disguised abdication. Without that, the framework will merely provide a more elegant vocabulary for the politics it claims to rationalise.

Test 7: Is this a whole-of-government R&D strategy or merely a UKRI reorganisation?

An ambiguity runs through bucket theory: its scope. Is this a framework for allocating the DSIT R&D budget that sits under Lord Vallance, or is it a design for the UK’s entire public R&D system? The distinction matters enormously, because the two approaches imply radically different institutional machinery and very different claims about what the framework can achieve.

The immediate context is the government’s Industrial Strategy, which, due to its centrality in Whitehall discussion, must be both the conceptual jumping-off point and the organising principle for R&D investment. Growth is explicitly Labour’s primary mission – Bucket 2 by definition. Yet the Industrial Strategy is led by DBT, structured around eight sectors, and fundamentally business-facing – which sounds like Bucket 3. The government’s own Digital and Technologies Sector Plan discusses AI, engineering biology, and quantum: technologies that span discovery research, government missions, and commercial application simultaneously. The government might reasonably argue that Industrial Strategy is big enough platform to give the overall directional logic for all three buckets. But that only sharpens the question: if bucket theory is meant to operationalise Industrial Strategy priorities, can it do so from within DSIT and UKRI alone?

The answer is obviously no. The research councils and Research England account for well under half of the government’s total civil R&D spend (excluding fiscal measures like R&D tax credits). The rest flows through departmental budgets, public sector research establishments, and various arms-length bodies. Defence and health research mainly flow through separate bodies. If Bucket 2 is the home for mission-led research but excludes most departmental R&D, it cannot fulfill its stated purpose.

The same applies to Bucket 3. Innovation policy is not primarily about Innovate UK grants or Catapult infrastructure. The largest levers shaping business R&D are tax credits, procurement rules, standards, IP regimes, competition policy, and access to finance. Most sit with HMT, DBT, or regulators. If Bucket 3 is defined narrowly as “DSIT’s business R&D programmes”, it is a minor component presented as the main story.

So, if bucket theory is intended as a whole-of-government framework, then it must be designed accordingly. That means specifying how departmental R&D budgets, PSREs, tax instruments, procurement, and regulatory policy map onto the buckets. And it means creating effective cross-departmental governance and coordination mechanisms for Bucket 2 missions, including explicit machinery for aligning DSIT, DBT and HMT around industrial strategy priorities. At the risk of stating the obvious, this is very difficult to do.

The test is therefore simple but consequential: if bucket theory drives the whole-of-government R&D settlement, does the institutional design measure up? If the framework is meant to operationalise industrial strategy but is built only on DSIT and UKRI machinery, it will fail. The buckets cannot carry more weight than the institutions that underpin them.

Test 8: Can a three-bucket framework handle cross-bucket dependencies without collapsing under its own contradictions?

Even if the buckets are conceptually clear and institutionally mapped, the real system is knitted together in ways that resist neat separation. Much of the UK’s research capability is sustained by shared platforms: large facilities, expensive instrumentation, data infrastructures, compute resources, doctoral training pipelines, technical professions, place-based clusters. These are neither purely discovery nor purely mission nor purely commercial: they are enablers that sit underneath all three.

Trying to allocate these platforms to specific buckets creates immediate distortions. Does Diamond Light Source become a “mission-led” asset because so much of its science supports national priorities? Does exascale compute belong in Bucket 1 because frontier researchers need it? Are Catapult facilities “business-led” even when they serve as translational infrastructure for publicly funded missions? And how do we classify doctoral training programmes whose graduates feed universities, PSREs, government labs and industry simultaneously?

If bucket theory forces all shared infrastructure into one category, it will misrepresent the system. If it spreads infrastructural responsibility across buckets, it risks fragmentation, inconsistent investment cycles and duplicated oversight. And if it attempts a proportional allocation (X% discovery, Y% mission, Z% business), we are back to accounting cosmeticism rather than governance design.

The stable solution must be to acknowledge explicitly that cross-bucket platforms require their own logic, governance and investment model, sitting alongside the buckets rather than inside them. Without such a platform layer, the buckets cannot function operationally. With it, they begin to look less like a whole-system architecture and more like discrete modes nested within a broader infrastructure regime.

Thus the test is: can bucket theory accommodate the horizontal dependencies – infrastructure, talent, data, compute, or regulation – that make the vertical buckets possible? If not, the framework breaks the system into conceptual parts it cannot operate.

Test 9: Can Bucket 3 be made coherent without redefining “business-led innovation” beyond recognition?

Bucket 3 is, on paper, the simplest of the three: research led by firms, focused on commercialisation, scaling, productivity and growth. In practice, it might be the most internally conflicted. Truly business-led research is already occurring in the private sector – and not only does this spending not appear in public R&D budgets, it actually dwarfs it. What instead appears in public accounts is the relatively small amount of research that government co-funds because of specific market or system failures: spillovers, coordination failures, finance gaps, or public-good characteristics.

In this context, a credible Bucket 3 therefore requires the government to articulate:

1. which failures justify public involvement;

2. which instruments address each failure (grants, loans, equity, guarantees, procurement, standards, data regimes); and

3. how those instruments interact with the tax system, which remains the largest single lever shaping business R&D.

Without such clarity, “innovation” risks becoming a rhetorical wrapper for an eclectic mix of policy interventions – some pre-competitive, some mission-driven, some industrial-policy-driven, and some regionally motivated – whose only common feature might be that they are not “discovery”. That is neither analytically coherent nor operationally stable.

There is a further accountability issue. If Bucket 3 is truly market-driven, who is accountable for outcomes? Ministers cannot credibly claim credit for things built by the private sector (though they will give it a go!). But nor can they disclaim responsibility when public money is involved. Innovate UK sits uneasily in this space: they must answer to ministers for strategy while claiming to serve business demand.

The test here is whether the government can produce a credible theory of business R&D, grounded in identified failures and matched instruments, that explains why these activities belong in the public R&D line at all. If it cannot, Bucket 3 will remain an unstable hybrid: part industrial strategy, part economic development, part fiscal instrument, part corporate subsidy. And, over the years, this unstable settlement has repeatedly shown itself to be vulnerable to attacks from within and without.

Test 10: Does bucket theory change how decisions are actually taken?

The final test is the most important. After all the conceptual work, institutional mapping and political economy analysis, bucket theory must alter behaviour – otherwise it is just a set of headings. The UK R&D system has seen multiple reforms over the past two decades, many of which have not had the expected results because they merely rearranged reporting lines while leaving decision-making cultures untouched.

The risk is clear: the buckets will become a new way for finance teams to classify spend, a new annex in the DSIT annual report, or a new slide in UKRI staff presentations – while councils, departments, institutes and firms continue to operate exactly as before. Missions are still run through ad hoc committees; discovery is still shaped by opaque system-level decisions; business support is still fragmented across Innovate UK, DBT, HMT, and procurement rules with little integration.

For bucket theory to matter, it must create operational consequences. Programme design templates would need to differ between modes. Accountability mechanisms would need to change. Risk appetite statements would need to be explicit and enforced. Departments would need to adapt their R&D governance. UKRI structures would need to reflect the distinction between modes rather than the accident of disciplinary history. Evaluation would need to be bucket-specific rather than one-size-fits-all.

This is not a minor undertaking– a system redesign on a scale greater than anything attempted in recent memory. Which is precisely why this test is necessary: bucket theory must demonstrate that it is capable of changing decisions, not just relabelling budgets. If nothing about programme governance, institutional behaviour or accountability practices changes, then bucket theory fails – even if the taxonomy looks coherent on paper.

Conclusion: what would it take for bucket theory to succeed?

The three-bucket proposal is not trivial. If handled intelligently, it could clarify the roles of discovery, mission-led research and business-led innovation in a way the UK system has struggled to articulate for decades. It could give ministers a clearer mandate in areas where political direction is appropriate, protect discovery from episodic interference, and bring some discipline to the sprawling domain of business support. It could even become the spine of a more coherent national research and innovation strategy.

My tests set a high bar – deliberately so. Government is dealing with a messy inherited system, imperfect information, and real political constraints. No framework will resolve every tension or eliminate every trade-off. But bucket theory is not being proposed as an incremental adjustment or a tactical relabeling. It appears in the White Paper as a potential organising principle for a £20 billion system. If it is to function as so claimed, then it must meet standards appropriate to that ambition.

The purpose of my tests is not to try to argue the framework out of existence, but rather to insist that it be taken extremely seriously. For government, that means:

• identifying clearly the problems that the framework intends to solve;

• treating the buckets as distinct modes of governance, not categories of research;

• managing the inevitable transitions and boundary cases;

• mapping the framework coherently onto real institutions;

• giving discovery research legitimate and stable governance while protecting it from attack;

• stabilising the boundary between discovery and missions to prevent political opportunism;

• embracing this as a whole-of-government affair, not merely DSIT arcana;

• acknowledging the cross-cutting infrastructure on which all three modes depend;

• producing a credible theory of business R&D grounded in identified market failures;

• changing real decision-making, not just spending classifications.

If the government can meet these conditions, then bucket theory could form the basis of a more stable and intelligible R&D system – one in which autonomy, missions and commercial innovation have clearer mandates and less destructive ambiguity. If it cannot, the risks are obvious: another round of structural reform that promises clarity but delivers only reclassification, leaves the core institutional dilemmas untouched, and exposes the system to future interference.

In that sense, the three-bucket proposal is a revealing moment. It forces us to confront, openly, questions that have long been managed implicitly: who sets priorities, who governs discovery, who owns missions, who speaks for business-led R&D, and what the state is ultimately trying to achieve with its R&D investment. The framework will stand or fall on whether it answers those questions directly.

Most of the paper refers to England, as education and skills are devolved. But since UKRI’s remit is UK-wide, all universities in the UK should pay attention.

Overall

Like many white papers, this one is opinionated – and, in being so, it makes a few political moves very adroitly. But it’s not clear that all of those opinions agree with each other.

And, crucially, opinions ≠ policy. The government might have expressed a view on what shape the sector should take once it has dragged itself out of its current financial mire, but it hasn’t really set out where and how it plans to throw the sector some rope.

After 70 pages, the cumulative effect is that the paper commissions a great deal while reforming almost nothing. Repeatedly, one finds government waving the baton in expectation of music – but the score is blank, and the orchestra isn’t sure it’s being paid.

Positively, the white paper appears to have been a genuine joint effort between DSIT and DfE officials. This is not the Montagues-versus-Capulets of previous attempts to marry higher education and research policy. Whether this reflects genuine collaboration or simply the blitz mentality of joint case-management in a financial crisis, officials deserve credit for joining the dots.

Subscribe now

Differentiation and specialisation

The white paper has a lot to say about how institutions need to accept that they can’t be good at everything – and that this extends to research.

“We will incentivise this specialisation and collaboration through research funding reform… Ultimately, we anticipate that over time there will be fewer broad generalist providers and more specialists. This will include specialists in teaching only, specialists in research, and some institutions which specialise in teaching with applied research in specific disciplines.”

“By incentivising a more strategic distribution of research activity across the sector, we can ensure that funding is used effectively and that institutions are empowered to build deep expertise in areas where they can lead. This may mean a more focused volume of research, delivered with higher quality, better cost recovery and stronger alignment to national priorities.”

A sectoral reshaping is already happening in response to the funding crisis. When ministers say they want fewer broad, generalist universities and more distinct institutional missions, they are leaning into the crisis, not trying to resolve it. Government has accepted that the sector cannot return to a “great-at-everything” model.

During a funding crunch, it makes sense to argue for specialisation. Broad-based institutions that try to protect everything risk failure – which is precisely what government fears. What gets lost is system breadth, but that is happening anyway.

What’s new is that government now says it will use research funding to guide where the axe falls. There is no detail on what that will look like, beyond calls for joint research grants and shared equipment (operationally difficult and not transformative). That leaves a great deal of uncertainty – and potential for heat rather than light.

For universities and funders in the devolved nations, this may grate. UKRI appears to be handed a latent mandate to reform funding in support of an England-only agenda, with no clear constitutional mechanism for doing so.

REF reset

Will the REF “reset” be part of this reshaping? Probably not.

“We will reform the research assessment system to better incentivise excellence and support the government’s vision for the sector. We will review the weightings for the three elements of the framework to ensure we retain and sustain our focus on excellence… The timetable for REF 2029 remains on track, following a pause of no more than three months.”

The pause is not new, but the white paper makes a performance of it. I read this as groundwork for tweaks already expected: adjustments to weightings, a small-provider route, and probably a quiet retreat from the People-Culture-Environment elements that have caused friction and reputational damage.

A three-month “reset” cannot deliver genuine reform. Expect a tidied-up REF and a quick return to business as usual. The wider research-assessment ambitions – recognising outputs beyond publications, using better data, and building a longer-term model – remain distant. DSIT will likely continue to explore alternatives in the background, focusing on improved visibility of the research base as it stands today rather than in 2029.

Costs and sustainability

If part of the agenda is to ensure research is better supported, that cannot rely on government funding alone. Here the white paper points toward the charity sector.

“We will work with charities to improve cost recovery and ensure the sustainability of charity-funded research. We will also work with other funders to improve use of the Transparent Approach to Costing system for annual reporting and assurance.”

Everything here is gestural. “Working with charities” on full economic costing is a well-worn line that solves nothing. Charities will, with good reason, continue to fund as many projects as they can – it looks good in annual reports to have funded 500 projects rather than 400. And almost nobody in HE has the appetite to tell an academic to reject a grant on cost-recovery grounds.

Charities will continue to push for increases to the charity-indexed stream of QR (“Charity Research Support Fund”, as they like to call it). But without a stick – for example, linking Charity QR to actual recovery rates – the subsidy cycle will continue. Situations like these contribute to the sense of unreality that surrounds the whole sustainability issue.

AI for Science and digital infrastructure

“We will publish an Artificial Intelligence for Science plan this autumn and continue investment in Isambard-AI and the new Edinburgh supercomputer.”

AI for Science is the one genuinely live initiative. If it builds on the TBI report’s proposals around dark data and better software tooling, it could have real substance.

Knowledge exchange and commercialisation

“We will review Higher Education Innovation Funding (HEIF) to ensure it supports growth priorities and aligns with the new Local Innovation Partnerships Fund, which will invest up to £500 million to grow high-potential innovation clusters.”

This is the first time HEIF, the Local Innovation Partnerships Fund and cluster support have been tied together into a single narrative sweep. But without reform of the HE-BCI dataset, this linkage cannot be operationalised. And, again, the devolution wrinkle has been ignored.

Direction of travel

Overall, there is plenty here that sets the government’s direction and makes its position on HE reform clear. But it’s a finger pointed toward the horizon, not a detailed map.

The real action may be inside UKRI. DSIT’s three funding “buckets” – curiosity, missions, commercialisation – are becoming the organising doctrine, and the direction of travel is toward a smaller, simpler, more legible funding system aligned to explicit outcomes. The white paper should be read as a political overture to that process rather than a plan in its own right.

A question: When did you last see a government-funded science project that made you catch your breath?

The kind of awe you might have felt watching a gigantic SpaceX booster land itself in the tiny arms of a crane?

Public science in the UK and Europe has congregated around a central orthodoxy of prudence and discipline. We've become excellent at justifying our existence through market-failure logic and efficiency metrics. But we’ve neglected the magnificent public spectacle – the pageantries of competence that build deep, lasting public legitimacy.

This has created a vacuum which is now being filled by tech billionaires, philanthropies, and authoritarian states who understand that pairing visible feats with central patronage is a powerful source of legitimacy.

The solution isn't to abandon rigour, but to learn from a surprising source: the playbook of the Counter-Reformation. In a new, long-form essay for Better Science, I argue that we must rediscover a form of "demonstration statecraft" to make public science awe-inspiring once more.

Read "The Counter-Reformation of Science" here:

Better Science Project

The Counter-Reformation of science: how to make magnificence democratic

Recently in Rome, amid the Baroque splendour of Bernini and Borromini, I was struck again by a central paradox of the Counter-Reformation. Faced with fierce competition from Protestantism, the Catholic Church understood that discipline alone would not suffice to secure their positi…

Read more

8 months ago · 1 like · Ben Johnson and Better Science Project

I've been thinking about this while reading Richard Hamming's book, which is a repository of amazing insights into the textured experience of working in scientific research. Hamming famously talked about the value of “open doors”: 

Working with one’s door closed lets you get more work done per year than if you had an open door, but I have observed repeatedly that later those with the closed doors, while working just as hard as others, seem to work on slightly the wrong problems, while those who have let their door stay open get less work done but tend to work on the right problems! I cannot prove the cause-and-effect relationship; I can only observe the correlation.

This quote resonated with me, because it captures something we all recognise about how our institutions of knowledge actually work. 

Most researchers are trained along what David Guest calls T-shaped career paths.2 This means gaining a broad grounding in the literature and methods of a field (the horizontal bar), then a deep probe into increasingly esoteric knowledge formation (the vertical spike).

The T-shaped career is a sensible and easily replicable model that provably produces serious scholarship and builds genuine expertise. 

But what are the connections between all these verticals? A office with an open door? A department, faculty, or an entire university? A journal editorial board or an academic conference?

And what happens when the questions we care about – particularly the big and messy ones – don't sit neatly within a single T or collection of similar Ts? What happens when keeping the door open isn’t enough?  

That's where interdisciplinarity is supposed to come in. But the problem is that different disciplines don't just use different methods – they often have completely different ideas about what truth looks like. What counts as evidence? What counts as success? Try getting a clinical trials researcher and a design ethnographer to agree on the word "data" and you'll see what I mean. 

So we end up with a curious tension. Our system trains people to dig ever-deeper wells when society is crying out for aqueducts. 

The hard truth is that interdisciplinary research is genuinely difficult to do – it’s hard to fund, hard to undertake, hard to assess, and hard to publish. So what passes for interdisciplinarity might be merely institutionally performative: the talking shops and coffee mornings disguised as ‘embedding strategic themes’, or the funding application that we claim is interdisciplinary because it superficially involves someone from another faculty.

The questions that have been nagging at me are: what does interdisciplinary actually mean? What does good interdisciplinary research actually look like in practice? And if we look at the actual corpus of research, can we tell the difference between genuine knowledge integration and institutional box-ticking?

Moving beyond shibboleths

So what do we mean about interdisciplinarity?

Perhaps we know! After all, it is certainly an important and widely accepted idea! Every university has interdisciplinary themes. Every funding agency has interdisciplinary calls. Every major societal challenge apparently requires interdisciplinary solutions.

In certain academic circles, incanting the term has become a kind of entry requirement for serious discussions about policy and strategy.

But ask what it actually means and you get institutional answers. Cross-faculty collaboration. Mixed review panels. Joint supervision. Team science.3 All of this might be very valuable, but it tells us nothing about whether the research itself is genuinely connecting different knowledge domains together. 

If there is a performative element to interdisciplinarity, then I believe it’s well-intentioned at least. At its heart may be a sense of duty to honour the social contract of research: our shared sense that publicly funded science ought to be tackling problems that matter in the real world – many of which do not fit into our internal structures. In this way, interdisciplinarity can serve as a sort of "impact wrapping": a signal that our research agendas are connected to broader purposes.4

That isn’t bad or wrong per se, but there’s another very different way of looking at all of this. In discussions about the "edginess" or “novelty” of research, and particularly the scope for greater support for “disruptive” research, interdisciplinary research often gets positioned as the opposite of boring “mainstream” research. Where most research is supposedly “conservative” or “incremental”, interdisciplinary work can be entrepreneurial and boundary-pushing.

This is heady stuff, if true, and it might be uncomfortably Dionysian for some. But if we probe this intuition, we might indeed uncover more concrete notions of interdisciplinarity – where novel combinations of research activity, including through radically new institutional forms, can and do lead to breakthroughs at the intersection of fields. 

After all, some of the most exciting recent advances have come from genuine integration across disciplines. AlphaFold is the obvious example: real breakthroughs at the intersection of biology and computer science, making advances in both fields. 

But how much of what gets labelled "interdisciplinary" is actually like that?

Measuring what actually moves

If we can't define interdisciplinarity structurally, we're stuck with institutional proxies and good intentions. So I decided to approach this differently: what if we looked at the actual structure of research and tried to identify papers that are functionally interdisciplinary? 

The approach I developed models the research literature as a giant citation network – papers as nodes, citations as edges – and then asks: which papers are acting as connectors between otherwise separate communities?

The first step is identifying those communities. On a sample of OpenAlex publications metadata, I used the Leiden algorithm to partition the network into clusters based on citation patterns. This gives us a data-driven map of the intellectual landscape: which papers cite which others, which groups tend to be internally connected, and which bits of the landscape are more isolated. 

Then I built what I call a Bridging Score to identify papers that connect across these boundaries. It's a composite measure that weighs four factors:

• Connectivity (40%): What proportion of a paper's citations reach outside its home community? If most citations are cross-cluster, it's acting as a connector. If it's mostly reinforcing its own field, it's not. 

• Diversity (30%): How many different communities does it connect to? Bridging two fields is one thing; bridging five is something else entirely. 

• Centrality (20%): How often does this paper sit on the shortest paths between other papers in the network? This captures whether it's playing a "broker" role in knowledge flow.5

• Impact (10%): A small weighting for citation count – partly as a quality filter, and partly because if nobody's building on it, it's probably not doing much bridging in practice. 

The result is a metric that identifies papers that aren't just interdisciplinary by intention or label, but functionally interdisciplinary in the structure of knowledge flow. In other words, papers with a high bridging score are moving ideas between fields. 

Now, this isn't perfect. The community detection algorithm makes choices about granularity. The weightings are arbitrary and could easily be adjusted. Citation patterns don't capture everything about intellectual influence. And I haven’t controlled for some disciplinary effects (e.g. the effect of generous funding in certain fields).

But the Bridging Score is an attempt to move the conversation from institutional definitions to structural ones – from "what do we call interdisciplinary?" to "what does knowledge integration actually look like?"

Who's funding the bridges?

I ran this analysis on about 500,000 papers, using fractional counting to handle the co-funding problem (i.e. if a paper has three funders, each gets one-third credit). The question I wanted to answer was: which funders are systematically supporting work that scores high on genuine boundary-crossing?

The results were illuminating – and I think surprising. 

The first graph shows bridging performance at a funder level, where each dot is a funder in my sample (hover over each dot for more information). The further to the right a funder is, the greater the average Bridging Score of the work it funds. Bigger funders are at the top, and bigger dots support larger numbers of bridging papers (though not necessarily at a higher intensity, as you can see).

The second graph aggregates this data by funder type. (Again, hover over each dot for a breakdown).

What does this all mean?

Well, half a million papers sounds like a lot, but it’s only a small fraction of the full ~250 million-strong OpenAlex graph – and we can clearly see some sampling noise for smaller funders in the first graph. But I think we have some signals here, which can be understood as emerging insights that now need to be stress-tested at scale, with greater attention to sensitivity analysis, confounding factors, and so on. 

So with those caveats in mind, here’s what I think the data tells us: 

Insight 1: Mission-driven funders consistently outperform government bureaucracies.

International philanthropic foundations achieve high bridging rates (i.e. the percentage of funded papers that had a bridging score ≥ 0.5). UK medical charities come a close second. At the top end, you see funders like the Chan Zuckerberg Foundation (28.4%), Howard Hughes Medical Institute (19.7%), and Alzheimer’s Research UK (17.6%). 

These are charitable organisations, but I am calling these mission-driven funders because it seems important that these organisations are focused on specific problems that happen, perhaps, to require integration across fields. 

Insight 2: UK government funders lag substantially.

UK central government funders manage a bridging rate of just 5.9%. Even accounting for the broader scope of government funding, this is a horrible finding. If these results hold up across the full network, we need to draw important lessons from it, including about the vital importance of other types of funding – particularly charity and philanthropic funding – for enabling UK researchers to undertake risk-taking, boundary-crossing research. 

Insight 3: Universities are unusually good at supporting interdisciplinary research from their own funds.

University internal funds score unusually highly – certainly much better than government project funding. 

When universities back their own researchers' judgement about promising directions using their own limited resources, they get more boundary-crossing work than when government funding channels research through disciplinary review panels and administrative structures. 

This is a striking finding for current UK policy discussions! It would appear to provide remarkable evidence for the value of quality-related research (QR) funding, and the dual support system more broadly. If universities are more willing to support interdisciplinary research themselves than the government is funding them to do directly, then we should think very differently about how the public might best support the work that will solve problems for society.

Insight 4: China is clearly doing something right.

The National Natural Science Foundation of China achieves a strikingly high bridging rate of 12.5% – over double the UK government average. I initially wondered if this could be a data artifact or sampling bias – but it doesn’t appear to be. If anything, coverage limitations probably understate the effect. 

This is hard to square with commonplace narratives about Chinese research being conservative, or following rather than leading. Whatever structural features of their funding system produce this effect, they're worth understanding further. At the very least, this finding suggests that different institutional arrangements at a national level can systematically encourage different kinds of research. Though I suspect the implications could be a lot sharper than that.

The collaboration curve

I also flipped this analysis on its head – is there a relationship between the number of funders a paper recognises, and its interdisciplinarity? Grouping papers by the number of co-funders, I found a clear but non-linear relationship:

Bridging scores rise with collaboration, peak at around 7-8 co-funders, then start declining. This indicates that there’s a sweet spot for collaborative funding – enough diversity to bring together different perspectives and resources, but not so much that coordination costs and consensus-seeking start to dominate. 

This would appear to have immediate implications for how we design interdisciplinary funding programmes. There is a benefit to collaboration, likely naturally arising at the level of projects, rather than funder-to-funder. But building ever-larger partnerships and consortia may be counterproductive beyond a certain point. Perhaps very large collaborative efforts introduce bureaucratic overheads that actually reduce the risk-taking and boundary-crossing we think we're encouraging. 

We must take these particular findings with a pinch of salt! The funder attribution metadata is patchy, and while I’m confident in the individual funder attribution, this ‘counting up’ is likely weakened by poor metadata coverage. But it’s interesting to see confirmation in the data that funder diversity and plurality seems to give rise to interdisciplinarity (up to a point). If nothing else, it reinforces the broader point that structure matters more than good intentions. 

What it means for how we think about research

These emerging patterns reveal something important about the gap between rhetoric and reality in research funding. We've built systems that talk interdisciplinarity but structure around disciplinary boundaries. 

Despite integration within a single UKRI umbrella, UK government funding seems to remain unusually focussed on silos of activity, rather than integrative impact. UKRI is on a journey here (with very good early steps being taken), but it’s a bit of a shame we’re not further on.

More broadly, many researchers still insist that their environments continue to reward contribution within disciplines, not between them, where depth of impact is prioritised over breadth of reach. It’s clear from this data that knowledge integration is happening – but it is happening despite the overall architecture and intent of UK government policy, not because of it.

The success of mission-driven funders suggests that problem orientation naturally produces disciplinary integration. This makes sense: when you're trying to tackle Parkinson's disease or arthritis, you don't care whether the promising approaches come from neuroscience, engineering, or computer science. You care whether they work.

The performance of university internal funds suggests that researchers know how to identify boundary-crossing opportunities when given discretion to take risks. When universities invest their own limited resources, they seem to be better at supporting work that connects fields than when government funding filters research through expert review processes. One important lesson for the UK? Telling a crisis-stricken university base to focus on ever-decreasing circles of ‘specialisation’ is a path to impoverishing us all.

The Chinese pattern suggests that different funding structures and environments might systematically be encouraging different kinds of research behaviour. Of all of the findings, this is the one that merits further study outside of my methodology, as it suggests that the geopolitical implications of funding structure choices may be more profound than we’ve yet recognised. 

Conclusion

Not all interdisciplinarity is good. Much of it is probably performative. But some is structurally real: papers that genuinely move knowledge between otherwise disconnected communities. 

I return to AlphaFold as a test case. It's genuinely interdisciplinary in the sense that most matters: work that makes genuine advances in multiple fields, not just applications of one field's methods to another's problems. It required mission focus, technical depth, and the kind of institutional support that doesn't fit neatly into traditional disciplinary categories. 

If we think that kind of knowledge integration matters for complex societal challenges, for scientific breakthroughs, or for system efficiency, then we need to understand what produces it. 

The Bridging Score is one approach to making this distinction. The institutional patterns it reveals suggest we have more to learn about how funding structures shape knowledge production. The emerging data shows that mission-driven funders, flexible university funds, and problem-oriented institutions outperform disciplinary-facing bureaucracies at fostering genuinely interdisciplinary research. 

This is a system feature we could design for, if we were serious about wanting the real thing. 

Let’s leave the door open to that! 

Acknowledgements: I am grateful to Ian Chapman, Ben Steyn, Stian Westlake, Christopher Smith, Helen Cross, Laura Ryan, Sam Roseveare, Pedro Serôdio, Sam Currie and Jim Coe for being enthusiastic about this work and/or encouraging me to share this with you all. All mistakes are entirely my own (and there will be some!).

Below is a summary of the piece + a postscript.

Summary

For decades, the US has been the most connected node in the global research system. It has benefited from serious investment, has a broadly liberal and democratic society, and is home to many of the world's highest performing research institutions and labs. It has acted as a primary anchor of norms, collaborations, standards, and trust. This has made it a reliable partner for other research-intensive nations.

But under Trump this reliability is now in question. The NIH faces a 37% cut. The NSF has terminated hundreds of grants. The NEH and NEA face total elimination. The FDA is laying off thousands of staff. For US universities, the pressures are financial and political. International student admissions are being weaponised. Overheads are being slashed. Campuses have become political flashpoints.

When the most entangled node in a network falters, the effects are never contained, and those with the closest ties to the US will feel the largest effects. It's therefore deluded to think this faltering is an upside-only 'opportunity' for the UK – or for anyone else in the West, for that matter.

Meanwhile, China marches on. It now dominates the Nature Index, with 8 of the top 10 institutions for research output. Why? It is not just investment. China’s system is aligned: universities are backed to deliver national goals in tech, talent and competitiveness.

Western knowledge systems lack that clarity. Are our universities meant to pursue critical inquiry or serve the public interest? Collaborate globally or build national capability? When consensus breaks down, institutions are left fighting for basic legitimacy.

The problems run deeper, though. The US is not just cutting science but redefining it. Trump’s new executive order (Restoring Gold Standard Science) is a deeply political move, talking the talk of narrow standards and accountability, while walking the walk of increased political oversight of research. But, unlike in China, political control over US science seems likely to accelerate fragmentation rather than restore confidence.

When institutions in a divided democracy depend on political protection, international partners have good reason to hedge their bets.

The UK, despite strains on its system, retains international research credibility. Despite the odd skirmish with politicians, our institutions still enjoy relatively widespread public support. But international leadership must include achieving greater clarity on what our institutions are actually for – and how to reconcile global connectedness with UK interests. For this, we need a research policy that marries democratic freedom with seriousness, coherence and institutional resilience.

This means confronting tough questions about where and how research should be done, how to find 'purposeful freedom', and how to focus our resources properly. Here are a few starter-for-10 questions:

• Where should cutting-edge research happen—in universities, independent institutes, or industry?

• How much should research funding prioritise investigator-led free inquiry versus directed national priorities?

• What does research leadership look like in a world where no single country can dominate?

• How selective should we be about research priorities when resources are finite?

• How do we protect and enhance important research capabilities when the system is in flux?

I cover this in more detail in the piece. Please take a read – it’s free!

Postscript

My piece focuses on lessons for HE because in the UK that is where the bulk of publicly funded science is undertaken. But it can also be argued that a more fundamental problem lies in public confidence in the entire system and enterprise of science itself.

It's interesting to consider the focus on both outcomes and process here. On outcomes, Trump’s desire to fund trade schools is obviously not only about Harvard (though some argue it is a personal vendetta). While we may not like it, it’s bound to be quite a popular move – because when the outcomes of science funding are contested, other routes might more clearly promise societal impact.

On the other hand, the Trump EO is interesting because it seems to reframe the public interest question in terms of process concerns – where research can no longer necessarily be trusted to deliver high-quality, reproducible results.

It is worth working through what this emphasis on process concerns means. In the UK, we are still talking about S&T in terms of achieving outcomes – and how to incentivise this better. Research process changes in the UK have rarely been driven by politicians (outside of a few areas like national security). Politicians have instead broadly left the research community to manage things, providing occasional cover where needed on contentious issues (like open access).

This is a favourable position to be in. But – and it's a big but – some community-driven process concerns (REF, EDI, disciplinary parity of esteem) do sometimes bubble up to the political level, and occasionally this has put the research community at odds with political imperatives.

Sometimes explosively so.

It's worth considering the topic of research bureaucracy here, which on the face of it is entirely about process – but which has had real political bite in the UK in recent years. “Slashing red tape for researchers” is something which straddles politics and academia, and while it's riddled with conflicting imperatives, it has a certain kind of universal appeal, and is unlikely to blow up in the same way as more contestable issues.

So if the UK is going to head off RFK-style challenges to research process integrity and legitimacy, confronting research bureaucracy might be a sensible place to start, building on the excellent Tickell Review.

But as I say in the piece, there can be no ducking the fundamental questions about outcomes.

Subscribe now

I have created this as a microsite, designed (in a meta way) to explore why the microsite has become the native form for serious AI discourse in 2025.

Microsites take AI discourse beyond the confines of social and traditional media. They perform seriousness, engineer credibility, and collapse the distance between fact, forecast, and fiction.

In the piece, I argue that:

• Microsites update the tradition of “grey literature” for a fragmented, web-native world

• Their design – clean, singular, self-contained – preloads a sense of authority and inevitability

• They blur explanation and persuasion into immersive campaigns

• They work less by telling us what to believe and more by shaping what we remember

The essay itself is designed for a smooth scroll, and ends with a tongue-in-cheek “model card” – a nod to the very conventions it critiques.

You can read it here:
👉 scrollthefuture.ai

Would love to hear what you think!

• Nightingale and Phillips cast doubt on the UK’s claims to be a science superpower because UK-authored publications are largely absent from the very cutting edge of key fields.

• If we are becoming globally irrelevant then that’s a big problem. Debating how to concentrate our resources is a useful route to finding solutions, such as improving our connectivity with the global research system and experimenting with different funding approaches.

• The citation-counting approach that N&P used is inherently flawed, but it’s the best we can currently do. Rather than deny the importance of cutting-edge work as a core component of a globally relevant research system, we should improve our understanding of research performance with increased support for the field of metascience.

Last week Paul Nightingale from the University of Sussex and James Phillips from UCL published a paper with some interesting new insights into the UK’s status as a leading science superpower. There has been some good discussion around this, and if you haven't read the paper, then it is well worth doing so. James Phillips includes a helpful bullet point summary on his Substack page (linked above), but the full paper is not a challenging read.

The paper is intended to be provocative, and this blogpost is therefore an attempt to formulate a response.

First, a summary of what I took from the paper. The headline is that Nightingale and Phillips (N&P) cast doubt on the UK's continued claim to pre-eminence as a scientific nation. They are concerned that the UK is falling behind in key fields, and the paper sets out reasons why this may be the case (discussed below). They recall COVID-19, noting that while the UK was outstanding in vaccine development and deployment using existing technologies, we had long missed the boat on the new technology of mRNA – a technology which promises much greater transformative impact through, for example, cancer vaccines.

Having dismissed other claims of the UK’s research prowess as unevidenced – e.g. because the REF outcomes are founded on marking our own homework – their paper zeroes in on one very overused data point: the UK’s impressive 13% share of the top 1% most highly cited research publications. The UK government (and other institutions) use this statistic liberally as evidence of the UK punching above its weight. On the surface of it, this is an impressive stat with abundant comms value for any science ministry!

N&P take issue with this. They argue that, in the global megaindustry that is modern research, 1% of the world’s output represents a huge stack of around 18,000 papers published annually. Such a large corpus cannot realistically represent the edge of the art – the truly blockbuster papers. Their analysis instead focusses in detail on the top 100 most cited papers in a small number of prominent fields, arguing that these 100 papers will more accurately reflect the most important work in a field than the many thousands of papers in the top 1%.

As I said, the paper itself is not a challenging read, and the data are laid out clearly, so I won’t repeat the results in full here. N&P’s key finding is that, at the narrow end of the bookshelf that is the top 100 most highly cited papers in the fields of AI and engineering biology, the UK sadly doesn’t much feature at all. Only by setting very generous parameters do N&P find a modestly better performance in the field of quantum. And where we do feature in these fields, we are clearly dependent on a tiny handful of high-performing institutions with peculiar characteristics, such as the privately owned DeepMind or the sacrosanct Laboratory of Molecular Biology.

Is it a problem if the UK does not feature heavily in the top 100 most highly cited papers? On this point, N&P point to a 2022 paper by Michael Nielsen and Kanjun Qiu that sets out how ‘outlier’ results may be driving scientific progress to a far greater degree than is captured by current comparative evaluation systems (in a research funding context). The Nielsen/Kanjun paper is worth reading in full, if only because it’s one of the most beautifully written metascientific analyses we’ve seen for many, many years! But I also commend that paper for offering new and interesting ways to think about research funding choices, identifying and working through the pros and cons of various scientific funding modes. This is particularly valuable because it helps to point to some ways forward from the situation identified by N&P.

The Nielsen/Kanjun paper also includes some neatly argued takedowns of quantitative comparative analyses of research per se:

Perhaps most fundamentally, it's not possible to neatly quantify the value of scientific discoveries in a commodity-like way. Indeed, arguably it's a mistake to attempt to quantify their value at all. Such discoveries are not fungible, nor do they add up. You can't measure the value of new scientific discoveries in units of milli-CRISPRs (say), and pile up three dozen 30 milli-CRISPR discoveries to get something more important than CRISPR. How many milli-CRISPRs was General Relativity? More or less than 27,000? It's all a little ridiculous.

This recalls some of the limitations with citation analyses that James Wilsdon et al. found when writing The Metric Tide in 2015. It was striking to me, when working on that project, to realise just how poor citations are as a measure of research quality:

• Papers themselves may not reflect the true extent of a research project’s findings and impact, particularly due to terrible tendencies in academic culture that value volume over quality and prestige over originality; 

• Citations themselves may not be reflective of a paper’s long-term scientific importance, with many important papers remaining unnoticed for years, or bad work becoming highly cited in refutation;

• The raw counting metrics such as Journal Impact Factor and h-index are so hopelessly skewed by a small number of trendy papers to be largely meaningless as an aggregate measure;

• The citation indices themselves have big holes in them where large numbers of citations, publications, or even the output of entire disciplines are missed by the scraping software;

• The citation indices do not always agree with each other on the number of citations any given paper has – with sometimes very wide variations in citation counts between different providers;  

• Some amazing research is never published, sometimes for dumb reasons, and sometimes for very good reasons.

It is of course entirely possible to apply such a critique to the work that N&P have done in their own paper. Does a handful of the most popular PDFs really represent the most important advances in a field? How confident are we in the publishing processes and organisations that have quality-assured and disseminated these PDFs? And how about the social infrastructure of journal editors, peer reviewers, issue article limits, article processing charge budgets, exhausted grad students doing literature reviews, disciplinary or departmental citation expectations, citation ‘Matthew Effects’, local citation clubs/rings, and so on – are we expected to believe that this complex and flawed mix of academic-cultural phenomena is reliably and robustly sending every piece of transformative work straight into the Billboard 100?

Clearly not, but the flaws inherent in a Billboard 100 approach should also be recognised as applying to that overused 13% of 1% stat. At least N&P are careful to caveat their analysis, and to corroborate it where possible. They are also clear that trying to address this situation by chasing an increase in highly cited papers would likely backfire, and I agree.

It would nonetheless be easy to dismiss all this sort of analysis as being based only on the flimsy evidence base of citation counts – a garbage-in, garbage-out interaction that therefore signifies only what we choose it to, and which further reinforces that famous streetlight effect where what gets measured is what counts, even if it doesn’t reflect the truth of the matter. (I’m not being particularly clever or original by highlighting these problems – I’d wager this discussion takes place regularly in every research and knowledge exchange committee in every university in the world.)

It isn’t my intention to dismiss N&P’s paper on methodological grounds, because, despite the limitations and caveats, their paper indicates something very important about the UK’s claims to be a scientific superpower. To appreciate this, we need to accept that any definition of research excellence must include the great leaps in knowledge and understanding that can have profound and far-reaching impacts.

Consequently, today’s scientific superpowers are those which can lead with a very sharp cutting edge of research, a cutting edge of excellence that can repeatedly and reliably create a bow wave of transformative change throughout the global academic community. This should never be the sole definition of research excellence, but it is a vital component that we ignore at our peril.

Moreover, this is about the UK’s relevance as a global research player. The false precision of citation counting may rub us up the wrong way, but let’s not pretend that there is no link at all between citations and relevance. The N&P paper indicates we’re becoming irrelevant, and that’s very frightening. 

Nightingale and Phillips state that, if the UK is to regain our position as a leader, greater attention (and funding) is needed at the earlier stages of discovery research. This seems a sensible conclusion to draw – but how we approach this task is not straightforward. I couldn’t help being drawn back to some of the reflections on breakthrough discoveries that Nielsen/Kanjun put forward in their paper. In this, they note:

There's a romantic version of the history of science in which it's all about the great discoveries – Galileo and his telescopes, Darwin on the Beagle, and so on. It makes for fun and perhaps informative stories. However, some people retort that focusing on the great discoveries is actually misleading. Such "outlier" discoveries usually come out of a rich surrounding context of "minor" discoveries – small increments or even errors or false trails – which are later forgotten by all but historians, but which were crucial to the evolution of humanity's understanding.

Conversely, Richard Jones asked on Twitter if we should “support fewer researchers, but fully fund their work, rather than relying on cross-subsidies, and provide them with proper technical and engineering support”. This builds on a similar line of thinking in the recommendations N&P make. Jones states this is an uncomfortable discussion, but one worth having – and I agree on both counts.

Fans of the genre will note with a wry smile that we have landed on an old, familiar dilemma in research policy. Do we help the UK system to become more competitive by concentrating resources in fewer people or institutions, or does doing so put our research base at risk? Taking concentration to an extreme, do we risk hermetically sealing our top talent in a UK system that supports them in a more sustained way, probably for years past their prime, and running our research base into the same sort of dependencies that made us a world-leader in non-replicating viral vector vaccines but relatively nowhere on mRNA?

Well, that depends to an extent on how hermetically sealed the UK system becomes. To be a bit grandiose for a moment, we cannot escape our place in the international Republic of Letters – and nor should we shirk our responsibilities to it. I’m tempted to make a facile argument about the importance of UK association to Horizon Europe – but let me instead say that international connectivity is a permanent and inescapable feature of our shared global research culture, of which intra-European collaboration is just one component part.

The above argument is obviously not a satisfactory answer to Richard Jones’ question, nor does it really address the quoted point about ‘minor’ discoveries. But I hope it shows how thinking about how to concentrate our resources can be a useful exercise in helping us focus on what’s important. If the UK’s relevance to global research is, in part, a function of our connection to it, then perhaps we need to prioritise increasing the international connectivity of our research base.

Similarly, we might also alight on arguments about the importance of taking risks on fresh thinking, rather than overfishing the same pond of overfed researchers with a high h-index. Or we might conclude from the Nielsen/Kanjun work that we need much more experimentation with diverse funding models, or be willing to test some highly disruptive new approaches – as the UK is doing with ARIA. We might also decide to pioneer new approaches in AI-driven research, given the rapid advances we’re seeing in that area.

N&P propose a number of recommendations here, many of which merit further examination on their own terms. Some of these don’t necessarily follow from their own citation analysis, as N&P take a more wide-ranging view of what’s needed to support excellent people and institutions. I have no problem with this, and would add that there is no shortage of other viewpoints on what it will take to turn the UK into a genuine scientific superpower, all of which also need examining on their own terms. I have my own views on what we need to do, but I didn’t want to turn this blogpost into yet another indexed list of headings and priorities for UK science policy.

Nightingale and Phillips have held up a mirror. It is admittedly a distorted carnival mirror because it exaggerates certain features. But it’s only a more distorted version of the carnival mirror that we’ve been staring into for years – that hackneyed 13% of 1% stat, the REF outcomes, and so on.

If we do lack a proper understanding of how the UK system is performing, then there is at least a simple solution: we should prioritise a radical increase in our support for the field of metascience. The Nightingale and Phillips paper also states the importance of improving how we assess research performance, and yet I couldn’t help thinking that their paper is the most interesting thing to emerge in a while from the UK’s research-on-research scene.

That’s interesting in itself, because what Nielsen and Kanjun are doing via their Science++ initiative is on another level. Other US players like Open Philanthropy and Stripe Ventures have been getting increasingly active in this space. I have many further thoughts on this which I’d like to expand on in future, so I’m just collecting a few salient points here to put a few markers in the ground:

• I’ve noted calls from some areas to radically rethink our notions of ‘excellence’, and I’ve often heard people argue that research excellence itself is a flawed concept – or at least that it is contested. I worry that a possible endpoint of these debates may be to move us even further away from the cutting edge, or that the debate is used as a way to avoid making hard choices about where we need to improve.

• Excellence in applied research may not be sufficiently captured by our current metascientific approaches. Perhaps as a consequence, applied research is insufficiently supported by the current funding environment, and over the years I have seen that (with some exceptions) how the current research culture doesn’t welcome this sort of work readily.

• I’d also assert that our current evaluation practices (and administrative processes) may not be set up to cope with truly ambitious programmes that bring together multiple disciplines and partners. Yet these kinds of programmes are probably exactly what we need to be pursuing if we are to solve problems in a focussed way. I have a lot of sympathy with calls for new institutional structures to address these shortcomings, but we also need to fix the issues within the existing evaluation system.

• It is possible that the next wave of great advances in machine learning, cryptography, surveillance, bioengineering, robotics, or materials science will now take place in the private sector. (Deepmind publishes its work, but many businesses rightly will not.) A nation that can harness its knowledge-creating power to support a cutting-edge private sector will be able to gain significant strategic advantage – yet we currently can’t effectively track this, so in some ways it doesn’t count.

To conclude, the N&P paper forces us to reflect on what it means to be a great research nation. Maintaining academic relevance by publishing impactful work at the very cutting edge is a key part of this – and it would be highly perilous to ignore this dimension. But it’s clear that we need to explore the whole problem space much more deeply, including so that we can integrate other important dimensions of excellence into our metascientific worldview.

Understanding what greatness looks like in research is in itself a deeply interesting intellectual problem, and one that we should now prioritise. Those that crack this puzzle will be at a significant advantage.