, Physics Today
Pages: pp. 104
Abstract—When you vote for your favorite song, what is the context behind that decision? Here, the author discusses what data mining and network analysis reveal about how we choose our loyalties for voting and collaboration.
Keywords—data mining; network theory; network analysis; national voting patterns; scientific computing
AS PART OF MY JOB AS PHYSICS TODAY'S ONLINE EDITOR, I BROWSE THE ARXIV E-PRINT SERVER IN SEARCH OF INTERESTING PAPERS TO POST TO THE MAGAZINE'S FACEBOOK PAGE. AMONG MY MOST FRUITFUL HUNTING GROUNDS IS THE SECTION OF ARXIV CALLED PHYSICS AND SOCIETY.
When I wrote this column's first draft, I found papers entitled "Terrorist Network: Towards an Analysis" and "Uncovering the Wider Structure of Extreme Right Communities Spanning Popular Online Networks," among others. Physics and Society is also where I found a paper by David García and Dorian Tanase, of the Swiss Federal Institute of Technology in Zürich. Its topic: changes in national voting patterns in the annual Eurovision Song Contest.
I first watched the popular music contest in 1974, when I was 11 and living in North Wales. Although I didn't know it at the time, that year's winner—"Waterloo" by ABBA—represented the now-57-year-old contest's musical peak. Most entries, whether they're from Belgium or Belarus, are simultaneously earnest and trite.
Still, despite its musical shortcomings, the Eurovision Song Contest is entertaining. International sporting events bring Europeans together, but the contestants' nationalities are typically manifest only by their uniforms. Watching the Eurovision Song Contest, however, you can hear songs in up to 40 different languages.
Winners of the Eurovision Song Contest are chosen partly by expert jury and partly by popular vote. It was the popular vote that interested García and Tanase. They subjected the past 15 years of digital voting records to a type of network analysis that identifies nodes of affinity—in this case, of international kinship or its opposite, international animosity. Intriguingly, they found evidence that the euro crisis caused viewers in the hardest-hit countries to become more inclined to vote for contestants in other hard-hit countries and less inclined to vote for Germans.
Not long after encountering García and Tanase's e-print, I came across a similar analysis in the journal Science. 1 Alessandro Chessa of the Institute for Advanced Studies in Lucca, Italy, and his colleagues crunched through millions of patent applications and hundreds of thousands of scientific papers that originated in the EU. Their algorithm identified the EU's principal nodes of collaboration—that is, the research centers that host the largest number of extramural collaborations. The algorithm also determined whether those collaborative networks reached across national borders.
With the exceptions of a Nordic node in Copenhagen and a Benelux node in Eindhoven, most nodes did not extend significant connections across national borders. Interestingly, whereas the British Isles, France, and Italy each had one single node—Cambridge, Paris, and Milan, respectively—Germany had seven!
Given that the EU spends billions of euros to promote cross-border collaborations, the findings of Chessa and his colleagues reveal what looks like a disappointing and expensive failure. Despite financial inducements, the EU's scientists evidently favor working with people in their own countries or regions.
There could, however, be a glimmer of encouragement for the Eurocrats in Brussels—or at least a less parochial explanation for the findings. Citizens of the EU are free to work anywhere within the federation. The scientists linked together in the Munich node, for instance, might not all be Bavarians, but a diverse international mix.
And in a way, it's also a positive manifestation of European unity if Irish, Portuguese, and Spanish viewers vote together for an Italian song in the Eurovision Song Contest.