Transcript

Predicting your social interactions | Sune Lehmann | TEDxAarhus

[Music] [Applause] imagine seeing a university campus from a helicopter but with detailed data about every single individual so how they move around who they're hanging out with and also all the online communication mapped out that's what you need to really understand human social networks now as a young researcher I worked at US universities like Harvard and Northeastern crunching data about phone calls and we were trying to find out how to analyze the network of all the phone calls made by an entire nation alright so that's millions of individuals and it's billions of phone calls but those billions of phone calls they weren't enough for me and the reason is that I've become obsessed with creating the best data set about networks in the world I had this idea that we could take our understanding of social networks to the next level if we could just zoom out and collect data about an entire social system I imagined this social system as a kind of massive anthill teeming with activity you see the thing is that as wonderful as it sounds with all those billions of phone calls they don't say much about how humans really connect to one another only a fraction of our communication happens over the phone so to make this obsession of mine a reality I found this group of individuals with lots of social activity between them all the freshmen at my university the Technical University of Denmark I got started on mapping out all their communication and all their networks and to do that I went out and I purchased 1,000 identical top-of-the-line smartphones and we installed custom software on each one to measure the networks and as in the side that is a lot of phones to install custom software on of course these phones that I gave to the students they would give me information about the phone calls and the text messages but the smart phone also knows who you're hanging out with what's going on in your Facebook feed and much more and as another side can I tell you something at this point I wasn't sleeping very well and the thing that was keeping me awake was the weight the sheer responsibility of this huge machine that we were building to observe this human anthill ok the budget was tens of millions I put an entire team of people to work countless hours of effort and collaborators depending on me and frankly I had no idea what I was doing because how could I no one had ever done anything like this on a scale like this before and the most scary thing was that I had to come up with some kind of fantastic scientific insight and in research the result is the only novel if you don't know what it is ahead of time so I was seriously worried that I made too big a gamble so one day as the data is rolling in and we're sitting and we're working on something boring and technical like Bluetooth connection strength something like that we more or less accidentally plot a five-minute snapshot of all the people hanging out in the real world this humble picture and this is basically just all the people that are hanging out between noon and 5 minutes past noon on some Tuesday but when we saw this picture we looked at each other and we realized that something significant had just happened and immediately we jumped up started drawing on the whiteboard and trying to sort of figure out what this would mean and by the way you know jumping up and drawing on the whiteboard that is the nerd version of hugging and high-fiving stuff so let me try and give you a sense of why this is amazing why that five-minute time window was the key and so the reason is that if you take all the connections over a full day you don't get something simple you get this undecipherable green hair ball shown here all right so the the the blue Network takes advantage through this super detailed data that we had collected but no one had ever seen what five minutes of a social system looks like or even thought about what it would look like and let me also emphasize this point the green network that's real data and it's for the same people and it's what researchers had looked up up until then and that's why the patterns that my team and I had found were invisible to them so because we were observing the network the system for the right amount of time that five-minute time window the network fell into bits and pieces and each little piece was a group of people hanging out so simply by collecting better data we had solved this fundamental problem of finding groups in social networks which is something researchers around the world are working on as we speak okay and that was just the beginning because we could now compare each five-minute window to the one right next to it so noon to five past five past to ten past and so on and that means that we could study the patterns of how groups form and to solve over time but if people are truly friends they don't just meet once they meet again and again over days and weeks and months and we can think of a group that meets again and again has a kind of social context for the people in that group right so the group could be our study group soccer team drinking buddies or something else entirely and now we began to wonder well how do people jump from group to group over time and remarkably this idea of measuring how people transition from social context to social context it gives us a whole new way of thinking about how people move through the social space we don't need the whole network just the context and technically behind the scenes that also meant that we didn't need the whole complicated Network and we could write down the relatively simple equations of moving from social context to social context and because the math was simpler we had much better conditions for understanding important real world processes like the spread of epidemics the spread of information and how opinion spread that was exactly the type of reefs of insights that I had hope for when we started the project a breakthrough and how we think about social networks and one that could only happen because we were able to zoom out and observe the entire anthill so it's difficult for me to even describe how relieved I was when we were sitting on a fantastic insight an insight that would give other people other researchers across many fields of science access to a new mental toolbox a new way of thinking about problems within their own fields and that sense of relief set us free to be even more creative because there had recently been a very very cool paper that showed how you collect people's future interactions based on where they had been in the past and because we were thinking about contexts and not networks we realized that there's this deep similarity between how we move through social space and how we move through geographical space just like we moved from place to place in geographical space we move from context to context in social space and because the problems are essentially the same that means that we could take these methods of location prediction and transfer them directly to our social data so now we had built a machine that could predict people's future encounters so I thought it was pretty sweet and believe it or not there's one last wrinkle to my story and so that is that you have to remember that we also had the data about how people move from place to place so that means that we could calculate both kinds of predictability okay and we first showed that people's social life is about as predictable as their movements but what we found when comparing this social predictability to our spatial predictability was a kind of delicate balance so in general people tend to go to the same places and meet with the same people in some weekly pattern but one important exception is clear from the data and that is that sometimes humans go exploring we go out to see new places and our movements become unpredictable and our research shows that precisely when our movements are unpredictable that's when our social life is most predictable and I think that the explanation is that when we challenge ourselves to do new things we tend to do that with the same people so if I go hiking in some dangerous mountainous terrain I don't bring some random stranger with me I go with my friends right and so I thought about it and I think almost all of you in the audience are out exploring today by being here you're doing something unusual and that means that according to my theory you should be here with someone you know pretty well so so so let's put it to the test if you're here with someone you know pretty well like your partner family member close friend please raise your hand ah pretty good and now we should also test the opposite perspective so please raise your hand if you're here on a first date I think that we just did a little bit of science so so I don't know if we proved or disproved my theory but for sure we do have another data point we have a new anthill thank you [Music]