The UK likes to declare itself a ¡°science superpower¡±. Yet in a recent study, Paul Nightingale of the University of Sussex and UCL¡¯s James Phillips ?UK researchers are only involved in between 3 and 7 per cent of advances in the national research priority areas of quantum computing, AI and synthetic biology: a fraction of the figures such rhetoric would suggest.?

The authors blame the UK research system for being ¡°optimised for volume production, like a factory¡±, rewarding ¡°research that is regarded as good now, not research that is regarded as truly exceptional today or that can be regarded as game changing in 100 years¡±. They are right. The UK produces a reliable stream of research that is excellent by its own terms, but the Research Excellence Framework¡¯s award of an ever-rising proportion of 4* (¡°world-leading¡±) grades is just an artefact of marking your own homework. In reality, without DeepMind, the UK¡¯s share of the citations garnered by the top 100 recent AI papers, for instance, drops from 7.8 to just 1.9 per cent, while the UK is only present on one of the biggest 27 synthetic biology advances of the past decade thanks to the MRC Laboratory of Molecular Biology in Cambridge.

Why? While UK government spending on research and development has risen recently, it remains relatively low, according to the recent : 0.46 per cent of GDP, compared with an Organisation of Economic Cooperation and Development average of 0.6 per cent. What¡¯s more, overheads on UK grants are exceptionally high. Typically, more than half goes straight to central universities¡¯ coffers. Although such funds are supposed to underpin well-founded laboratories, few of us ever see any direct benefits.?

Nurse complains of overzealous ¡°management¡± of research and asks the government to ¡°replace frequent, repetitive, and multi-layered reporting and audit¡­with a culture of confidence and earned trust¡±. Indeed. While this kind of management works well in a fast turnover business, it does not in academia. It distorts not only research priorities but also timelines. In , one of us found that before every REF deadline, UK researchers produce 4 per cent more journal papers compared with the years after the deadlines.?The contrast is starker within REF submissions specifically: 29 per cent more journal papers and 60 per cent more books are produced in the submission year.?

³ÉÈËVRÊÓƵ

These deadline-pushing publications are not generally groundbreaking. They receive 5 per cent fewer citations over the next five years compared with those published immediately after the deadline (16 per cent more among the REF submissions). And the 9 per cent lower citation half-life of the journals in which they are published points to their being on shorter-term topics. They are also more prone to being retracted, raising questions about rigour.

If 4* really meant world-leading, things might be different. But it is not necessary to publish in the really top journals to get such a rating. Indeed, numerous institutions dislike their faculty trying for 4* papers and want them to aim instead for 3* papers (considered to be much cheaper to produce). Even academics at leading institutions have told us they are actively dissuaded from investing in new research areas as producing results ¡°would take too long for the next REF¡±.

³ÉÈËVRÊÓƵ

The research councils are also to blame, of course, for tying up far too much funding in cross-cutting initiatives endorsed by ministers, as opposed to discovery-led research. Nor is this decline in the ¡°disruptiveness¡± of research solely a UK phenomenon: a recent study this was true of science in general.

But many countries have of the REF, and the extra layer of scrutiny such exercises impose is clearly counter-productive. However many ¡°REFable¡± papers researchers are forced to produce each cycle, the relatively short time horizons encourage a mass of small, undistinguished and weedy research projects to flower ¨C while tall poppies that stand out globally become ever rarer.

Moqi Groen-Xu is senior lecturer in economics at Queen Mary University of London. Peter Coveney is director of the Centre for Computational Science at UCL and co-author of??just published by Princeton University Press.