A machine-learning algorithm can now predict human personality types using nothing but what people like on the Facebook social media site. A team at Stanford University in California and the University of Cambridge used data from a questionnaire filled out by 86,000 people that identified their "big five" personality traits. The results were correlated with their Facebook activity.
On the basis of between 100 and 150 Facebook likes, the team's algorithm could determine someone's personality more accurately than could their friends and family, and nearly as well as their spouse.
"In the past, my research has looked at how accurately people can judge each other's personalities," says Youyou Wu at Cambridge, one of the study's authors. "It's surprising that computers can do better using just one piece of information – likes."
In 2013, the same group found that Facebook likes can predict private personal information like sexuality.
Shallow assessment
David Funder of the University of California, Riverside, says predicting a big-five score is impressive, but it is just one component of personality. "This is a very broad way of describing human personality. It's useful, but not intimate," he says. "It doesn't really comprise understanding on any deep level."Funder's own work studying human personality uses more than 100 dimensions, not five. He points out that humans have depths that the algorithms of Wu and Michal Kosinski – Wu's co-author at Stanford – do not scratch. A spouse of 10 years will know their partner in ways that computers are still far from fathoming.
However, Kosinski and Wu's computational approach to understanding personality could be useful in hiring, health and education, where short personality tests are already used.
Their techniques aren't confined to Facebook, so any pot of data that you generate – websites you visit, phone calls you make, presents you buy – could be aggregated and analysed to figure out exactly who you are.
Work comes to you
Your suitability for a job is one of the first problems that could be tackled. Instead of people seeking and applying for specific jobs, hiring AI's would roam pools of data, seeking the right candidate for a specific task, says Kosinski. "Instead of having an 18th century way of recruiting people, where you plead with the mill for work, the work comes to you," he says. "This is the revolution."Ideally, every person would have full control of their own data and over which companies and organisations can ask questions about them. But Kosinski acknowledges that the algorithm could be used in an invasive way too, with such personal data available to paying customers. "There's a very big dark side to the technology," he says.
"With one click of the button it could be applied to the entire population of the world. We have to make sure that, in terms of policy and technology, we solve those issues properly," Kosinski says.
Journal reference
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.
No comments:
Post a Comment