We are already aware that our every move online is tracked, aggregated and analysed. But you couldn't have known how much Facebook can learn about you from the smallest of social interactions ? a "like".
Researchers from the University of Cambridge designed a simple machine-learning system to predict Facebook users' personal information and traits based solely on which pages they had liked. "We were completely surprised by the accuracy of the predictions," says Michal Kosinski, lead author on the paper in PNAS (DOI: 10.1073/pnas.1218772110).
Kosinski and his colleagues built the system by scanning likes for a sample of 58,000 volunteers, and matching them up with other profile details such as age, gender and relationship status. He also matched up those likes with the results of personality and intelligence tests the volunteers had taken. The team then used their model to make predictions about other volunteers, based solely on their likes.
Super sorter
The system can distinguish between the profiles of black and white Facebook users, getting it right 95 per cent of the time. It was also 90 per cent accurate in separating males and females, gay and straight men, and Democrats and Republicans. It can even predict whether a person takes drugs or has divorced parents, say, although with lower accuracy.
Personality traits like openness and intelligence were also estimated based on likes, and were as accurate in some areas as a standard personality test designed for the task.
Mixing what a user likes with myriad other data from their real-life activities could improve these predictions even more. Voting records, utility bills and marriage records are already being brought into Facebook's information fold, where they are easier to analyse.
Online locker
Facebook recently partnered with offline data giants Epsilon, Acxiom and Datalogix, which all collate this kind of information. This move will allow even deeper insights into the behaviour of the web populace.
Sarah Downey, a lawyer and analyst with privacy technology company Abine, foresees insurers using the information held by Facebook to help them identify risky customers and perhaps tag them with higher premiums.
But there are potential benefits for users too. Kosinski suggests that Facebook could end up as an online locker for your personal information, unleashing your profile at your command to personalise things like smart cars or to help you with career planning.
Never forgets
Downey says the research is the first solid example of the kinds of insights that can be made through Facebook. "This study is a great example of how the little things you do online show so much about you," she says.
"You might not remember liking things, but Facebook remembers and it all adds up." Making personality predictions this way is interesting because likes are visible to the public by default on new Facebook profiles, she says.
A Facebook spokesperson said this kind of insight into personal preferences has been around for years, although did not address the fact that tapping into this information is now easier than ever using its new Graph Search facility. "With Graph Search it's not just whether you get a job, but whether you get a date," says Downey.
If you would like to reuse any content from New Scientist, either in print or online, please contact the syndication department first for permission. New Scientist does not own rights to photos, but there are a variety of licensing options available for use of articles and graphics we own the copyright to.
Have your say
Only subscribers may leave comments on this article. Please log in.
Only personal subscribers may leave comments on this article
Subscribe now to comment.
All comments should respect the New Scientist House Rules. If you think a particular comment breaks these rules then please use the "Report" link in that comment to report it to us.
If you are having a technical problem posting a comment, please contact technical support.
monsanto boston weather dr seuss birthday romney michigan derrick williams railgun jk rowling new book
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.