Our recent research on international networks within UK higher education questions the claim that UK scientists will be able to forge wider collaborations with peers in the USA, China and the Commonwealth after Brexit
Much has been said about the likely consequences of Brexit on the UK research infrastructure within universities, policymaking circles and the media in the UK. These include the substantial shortfalls in research funding hitherto provided by the EU, difficulties in recruiting and retaining world-class researchers and scientists, and the weakening of vital research links between its institutions and European ones. Often, these statements are not based on a clear understanding of the nature of the international science collaboration network and of the UK institutions’ position within it. In a recent study, we approach this task from a complex network perspective.
There exists a widely-shared consensus that the international scientific collaboration (ISC) network is self-organising and scale-free. A scale-free network displays a unique feature: only a few of its ‘nodes’ – these can be authors, countries or research institutions/universities – are highly connected to other nodes, while many others have only a few connections. In other words, a scale-free network is strongly hierarchical because few hubs enjoy many more connections than the large majority of other nodes. It is a self-organising network in the sense that it follows its own logic based on rules of preferential attachment: the network changes and grows as new entrants seek to attach themselves to the better connected hubs which, in turn, increase their connectivity faster than less connected nodes.
We mapped the ISC network based on data on co-authored articles drawn from the Thomas Reuter’s WebOfScience for 2000, 2003, 2006, 2009, 2012 and 2015, totalling more than 10 million articles in 170 science and engineering disciplines. Using the information of authors’ institutional affiliations in our database, we were able to identify 7,083,075 cross-border collaborative pairs for 2015 involving two authors in two different international institutions. International scientific collaboration today constitutes a sizeable phenomenon: indeed, the share of international co-authored articles in 2015 was in absolute terms the same as the whole of global scientific production in 1993.
A key dimension of the network became immediately clear: in 2015, it had 50,889 institutions with only one link and one institution with 54,813 links, confirming that the international scientific network is highly hierarchical. Upon further investigation, we uncovered some interesting insights about the position of UK research institutions/universities in this network. The most interesting revelation was that the node with most links – 54,813 in 2015 – was the University of Oxford and it was followed by the University of Cambridge with 50,481 links. In other words, these two institutions are the most important hubs of the international scientific collaboration network. Two other UK institutions appear in the top ten nodes: UCL in the fifth position with 45,773 connections and the University of Edinburgh is 11th with 44,087. The most connected US institution is Harvard University (47,100 links) in the fourth place.
Analysing the institutions and countries with which the UK institutions tend to interact the most, there appears to be little evidence in support for pro-Brexit supporters’ claim that UK scientists will be able to forge wider collaborations with peers in the US, China and the members of the Commonwealth once the UK leaves the EU. Our data suggests that while the US appears to provide most frequently collaborators for UK scientists, if EU countries are bought together, as indeed they are under the EU Framework Programmes, the EU surpasses the US by far (over double the level of interactions with the US, totalling 490,540 interactions).
The structuring effects of the EU research policy on the ISC network are notable; the European countries certainly show higher rates of international collaboration than the three other leading producers of scientific knowledge, USA, China, and Japan. Beyond the EU, our data shows that the UK institutions tend to collaborate often with scientists from Canada and Australia; India ranks in a relatively low position with 8,732 links.
What would happen to the UK research institutions/universities as and when the UK leaves the EU? This is a difficult question to answer; there are statements that safeguarding the UK’s reputation as “a global science superpower” will be a government priority post-Brexit, but no clear policies to evaluate. But insights can be drawn from what we know about the nature of complex networks like the ISC one. The two main types of attacks these networks experience are random or coordinated. In the former, random nodes are disturbed by altering their state or by depleting their connections, whereas in the latter knowledge about the network structure is deliberately used to attack a special subset of nodes. Because free-scale networks are hierarchical and have a few, extremely well-connected nodes, they tend to be quite robust against random attacks, which is why, despite frequent router problems, experiences of internet outages are extremely rare.
However, such robustness comes at an extreme price: these networks are exceptionally vulnerable to intentional attacks on – e.g. the selection and removal of – their key nodes. It is usually an informed agent, seeking to damage the network, that will preferentially target its most important nodes – the hubs – that play a vital role in maintaining the network’s connectivity. When the network’s hubs are removed, the ability of the remaining nodes to communicate with each other decreases, potentially leading to a drastic change in the network’s topology and, in some cases, to its collapse into many isolated clusters. This feature is very useful for those designing drugs that target the key nodes of a metabolic network, but hardly so in the case of social networks like the international scientific collaboration one.
It would be far-fetched to suggest that the UK leaving the EU constitutes an intentional attack on the ISC network by an informed agent. Likewise, it is difficult to imagine the key nodes of the network representing UK institutions – e.g. the University of Oxford, University of Cambridge, UCL and the University of Edinburgh, to name the main ones – cut off the network completely. Although these hubs’ connections are considerably stronger with the EU institutions, they are also connected to many other nodes across the network. However, hubs need not be so severely attacked to effect changes in the network; even altering their state would generate changes in the way the network behaves. This need not be bad news for the network itself, but it is likely to be so for the leading UK institutions. Only time will tell whether this came to pass; what we do know is that no agent is ever likely to master the ability to bring a complex social network to its pre-altered state.