Cross-border computer science collaboration ‘lags in Europe’

The full results of the THE World University Rankings 2021 by subject

Continental Europe has often been seen as having a natural advantage when it comes to international collaboration in research due to its geographical, cultural and political links.

However, the latest data from Times Higher Education’s subject rankings suggest universities on the Continent may be lagging on international cooperation in a key area: computer science.

For instance, among the world’s major developed research nations, some of the highest average scores for international research collaboration are seen in Switzerland, the Netherlands and Sweden. But on cross-border cooperation in computer science, the Netherlands and Sweden are overtaken by the UK and Australia, while Germany – a country that also performs solidly on overall international research co-authorship – drops about 20 points.

Holger Hoos, professor of machine learning at Leiden University in the Netherlands and co-founder of a campaign to boost pan-European collaboration on artificial intelligence, said the figures could reflect historical ways of working in computer science over the past 30 years.

“In a nutshell, I’d say substantial progress was achievable [in the past] within small, relatively weakly connected groups,” he said, which meant computer science research in Europe, with its different working cultures and languages, was often more localised.

However, he added, “the potential for localised progress is decreasing – in AI, this is happening as we transition to approaches that bring together expertise from different areas of AI – and beyond – and sometimes also require far more resources than used to be the case, in terms of computing power and amounts of data”.

It is one of the key reasons why Professor Hoos has been pushing, through the Confederation of Laboratories for Artificial Intelligence Research in Europe (Claire), for greater integration of Europe’s AI research community.

He said the co-authorship data did “provide strong evidence that pan-European mechanisms and initiatives are needed to increase collaboration between researchers” but the potential for the sector was high if this could be achieved.

“The diversity in research culture and expertise found in computer science across Europe that has created some barriers in the past is actually an incredible asset, once it’s properly leveraged.”

As well as in computer science, the ranking data also show the other subjects where some countries’ ranked institutions appear to collaborate more or less compared with their overall co-authorship profile.

For example, German universities score comparatively highly for international collaboration in life sciences, Swedish institutions do well in engineering and in the US there appears to be relatively more cross-border working in the physical sciences.

There are also countries that have a more even profile across all the subject areas for international co-authorship – Switzerland and Canada are two examples of nations that score well in most disciplines.

Meanwhile, the data also suggest that China’s ranked universities do not tend to collaborate internationally as much in some subjects, with life and physical sciences being two stand-out examples.

Simon Marginson, professor of higher education at the University of Oxford, said in physical sciences this could reflect longer-term patterns of Chinese scientists in the area working in small groups and only engaging internationally on a bilateral basis, often with US institutions.

Medicine and life sciences could also now be following these patterns, he said, after receiving “much less emphasis in China” but were now becoming “the fastest growing and fastest improving areas of research” in the country.

simon.baker@timeshighereducation.com