Thursday, 28 de March de 2024 ISSN 1519-7670 - Ano 24 - nº 1280

Data from social networks are making social science more scientific

 

“FOUNDATION”, a novel by Isaac Asimov from the golden age of science fiction, imagines a science called psychohistory which enables its practitioners to predict precisely the behaviour of large groups of people. The inventor of psychohistory, Hari Seldon, uses his discovery to save humanity from an historical dark age.

A fantasy, of course. But the rise of mobile phones and social networks means budding psychohistorians do now have an enormous amount of data that they can search for information which might yield more modest patterns of predictability. And, as several of them told the AAAS meeting, they are doing just that.

Song Chaoming, for instance, is a researcher at Northeastern University in Boston. He is a physicist, but he moonlights as a social scientist. With that hat on he has devised an algorithm which can look at someone’s mobile-phone records and predict with an average of 93% accuracy where that person is at any moment of any day. Given most people’s regular habits (sleep, commute, work, commute, sleep), this might not seem too hard. What is impressive is that his accuracy was never lower than 80% for any of the 50,000 people he looked at.

Alessandro Vespignani, one of Dr Song’s colleagues at Northeastern, discussed what might be done with such knowledge. Dr Vespignani, another moonlighting physicist, studies epidemiology. He and his team have created a program called GLEAM (Global Epidemic and Mobility Model) that divides the world into hundreds of thousands of squares. It models travel patterns between these squares (busy roads, flight paths and so on) using equations based on data as various as international air links and school holidays.

The result is impressive. In 2009, for example, there was an outbreak of a strain of influenza called H1N1. GLEAM mimicked what actually happened with great fidelity. In most countries it calculated to within a week when the number of new infections peaked. In no case was the calculation out by more than a fortnight.

Politics, too, is falling to the new psychohistorians. Boleslaw Szymanski of the Rensselaer Polytechnic Institute in New York state studies how societies change their collective minds. By studying simulated networks of people he can predict the point at which a committed minority can convert almost everyone else to its way of thinking. For an idealised model, the size of this catalytic minority is just under 10%. Tweaking the model with data from real networks such as Twitter and Facebook, he hopes, will allow these insights to be applied to the real world.

The boldest idea of the session, though, came from Dirk Helbing of the Swiss Federal Institute of Technology. Dr Helbing is one of the leaders of the FuturICT project, the aim of which is to create a general computer model of society.

As Dr Helbing puts it: “We understand the universe much better than we understand our own societies.” But physicists do not understand the universe by tracking every atom within it. They do so by devising and combining laws (gravity, thermodynamics and so on) that each describe part of the system. Similarly, a model of society would not aim to simulate in detail every human being on the planet. Rather, by combining smaller, more specific models, of the sort outlined by Drs Song, Vespignani and Szymanski, Dr Helbing hopes he or his successors will eventually be able to describe whole societies.

The Terminus man

That is a wildly ambitious project, but there could be some useful staging posts on the way: predicting when people are likely to riot, for example, or modelling the breakdown of trust between banks and customers that causes financial crises. This really is Seldon country, although Dr Helbing is quick to point out that a crystal ball is impossible. That is because the maths underlying complicated systems like societies are exquisitely sensitive to a model’s starting conditions. Small errors can quickly snowball to produce wildly different outcomes. But, decades hence, a kind of social weather forecasting that would make reasonably specific predictions, with a reasonable amount of confidence, over short periods may not be out of the question.