Your phone may soon know where you're going before you do
By Jan Belezina
July 12, 2012
Phones obviously already know where we are and where we have been, thanks to GPS and other clever positioning technologies. Now, thanks to an algorithm developed by researchers at the University of Birmingham, your smartphone may soon be able to make accurate educated guesses as to where you’re going to be in 24 hours time. And here’s the dirty trick responsible for the algorithm’s future-telling powers: it spies on your friends and connects the dots where necessary.
To be more precise, not only does it analyze your own mobility patterns, but it also extrapolates from similar data collected from the people in your social circle to identify any divergences from the routine. The assumption here is that there is a strong correlation between the habits and mobility routines of friends, friends being defined as people who have each other on their contact lists.
The exhibited prediction accuracy seems to suggest that the assumption is valid. In a sample of 200 people, the system was on average less than 20 meters (66 feet) off when trying to predict where a person is going to be in 24 hours time. When the same system was stripped of the social component, the average error grew to 1 km (0.6 miles).
Say you usually go jogging on Thursdays and you leave home at 6 pm following a certain route. A regular algorithm would assume that that’s what you’re going to do this Thursday, but this time it would be wrong since you’ve decided to visit the mall instead. However, the socially aware algorithm knows that your closest friends are not going with you to the mall, so it’s probably not a social occasion and you’re likely to catch up with your jogging routine once you’re done shopping (as opposed to joining your friends on a party and ending up somewhere totally unexpected the next day).
Mirco Musolesi, the head of the team behind the study, points out that the 200 people who were willing to be tracked over an 18-month period may not reflect the general population. However, Mirco sees the benefit of the study in that it “exploited the synchronized rhythm of the city” for enhanced predictive capabilities.
Should the algorithm be developed beyond the research prototype stage, it may become a tasty morsel for mobile networks, who are used to carrying out complex data mining operations on the vast amounts of information gathered from their users. For one thing, such an algorithm could allow them to serve up better tailored recommendations and advertisements on your phone. The researchers have obviously spotted that potential, since they’re planning to create a dedicated platform for developers.
Should this route of monetization work out, Musolesi’s group can look forward to earning much more that the €3000 (US$3,650) they received from the Nokia-sponsored Mobile Data Challenge.
Although the ingenuity of this solution is admirable, it does create a whole new level of privacy related dilemmas. For example, is information about something you haven’t done yet still private? However we deal with these kinds of questions, one hopes the algorithm will be used for something more interesting and useful than serving up better targeted ads. Dynamic traffic control and clever power distribution grids spring to mind.