Monday, December 19, 2016

Mathematicians can now predict your social networks

Mathematicians have developed a formula that can describe our interactions with the people around us.

The discovery could improve models to predict human behaviour, which can improve our understanding of how epidemics spread or improve town planning.

“For the first time in history, we can now study social networks on a large scale and learn more about them,” says co-author Sune Lehmann from the Department of Applied Mathematics and Computer Science at the Technical University of Denmark (DTU).

“Until now, no one had really gotten down to the mathematical core of it--but now we have, and it opens up for new modelling options,” he says.

The results are published in the scientific journal PNAS.

Thousands of articles on social networks

We all take part in social networks. It could be in a football club, at the work place, meeting old school friends, and so on.

There are many reasons why scientists want to understand the mathematics behind social behaviour. For example, by understanding patterns of behaviour, scientists can understand how quickly infections can be transmitted and ultimately how epidemics spread.

But until now, maths had fallen short of being able to describe these complex networks.

To change that, Lehmann devised an experiment involving 1,000 undergraduate freshmen from the Technical University of Denmark.

They all received a mobile phone, which tracked their whereabouts for 2.5 years. Even more, it recorded who the students spoke to, who they socialised with on Facebook, and who they met out in the real word.

All of this was possible with the telephones inbuilt apps, GPS, and Bluetooth.

Lehmann and his team amassed a gigantic dataset, which contained data on all of the students’ social interactions.

“The idea was to see if we could learn something from looking at all of these data channels simultaneously, and not just what people did on Facebook, or who they spoke with on the phone,” says Lehmann. “We have all of this in the dataset alongside data on when people met each other face to face.”

Time is important

Using all of this data, Lehmann and colleagues were able to mathematically describe the students’ social network dynamics.

They discovered that time resolution of the data was an important factor to understand the complex networks.

Without this temporal framework, the data becomes a mass of social interactions, which does not yield much useful information. So the scientists broke the data up into 15 minute time intervals.

In doing so, they could see how one person interacts with their friends at a given time of the day, without having to analyse everyone’s social connections at the same time.

“It’s been a hard nut to crack because we were studying the entire network over long time intervals. So the picture was too muddy. We find that you don’t need to search after all these groups in the entire network. If you just look at the data in short snippets of 15 minutes, then it all falls into place, and you can observe the groups directly,” says Lehmann.

The results also show how each individual student forms a part of different social groups.

It could be specific courses, reading groups, pub meet ups, or sports teams. Often these groups overlap with one another.

Sj√łgren, Kristian. 2016. “Mathematicians can now predict your social networks”. Science Nordic. Posted: October 20, 2016. Available online:

No comments: