Computers 'judge personality better than friends

  • Published
Facebook likeImage source, Getty Images
Image caption,
What do your Facebook "likes" say about you?

Computers can be better at predicting our personality than our friends and family, an experiment with tens of thousands of volunteers has indicated.

By analysing "likes" on Facebook, a computer model deduced a person's character on five key traits better than brothers, mothers and even some partners.

The Cambridge team acknowledge that personality is more complex than this.

But they say the results show computers can outpace humans.

And the findings, in the journal PNAS, suggest some interesting associations - Facebook fans of Dr Who tend to be shy, while fans of Big Brother are conservative or conventional.

Digital footprint

Image source, GIPhotoStock/SCIENCE PHOTO LIBRARY

The University of Cambridge and Stanford University researchers had already said that Facebook including sexual orientation and political leaning.

In this study, they wanted to go a step further and pit man against machine to see which would perform better at making judgements about human character and personality.

Dr Youyou Wu and her colleagues ran the data of 70,520 willing Facebook users through their computer system, which linked "likes" to five core personality traits:

  • agreeableness
  • conscientiousness
  • extraversion
  • neuroticism
  • openness

The Facebook users completed a personality questionnaire and were asked to get their colleagues, friends and family to act as character witnesses by filling out a survey.

The researchers then compared all the results to see how the computer model fared in summing up a person's self-reported character.

Given enough "likes", the computers came closer to a person's self-reported personality than even their closest allies.

By looking at just 10 Facebook "likes", the computer was able to outperform a work colleague.

Analysing 70 "likes" let the machine rival a friend or flatmate.

Using 150 "likes", it trumped a family member, but not a partner - it took 300 likes to match or outdo their deductive power.

Given that the average Facebook user had about 227 "likes" on their profile, this was ample for a computer to go on, said Dr Wu - a visiting intern at Facebook and a PhD student at the University of Cambridge.

"It may seem surprising because people are generally good at judging personality.

"It's intuitive to think that people close to us know us very well, so it's even more impressive that computers are able to beat us at our game."

She said that people now spent a lot of time online, which gave computers the edge.

"Friends and colleagues might only see us in a limited number of situations, so they have less information to go on.

"Computers can access a lot of relevant information about us and have a huge memory capacity.

"And they can make judgements in a consistent, systematic way, whereas humans may be biased."

Alan Redman, a chartered psychologist with a specialist interest in personality and psychometrics, said advertising companies already used our digital footprints to build a picture of who we were.

And potential employers might want to check social media sites before seeing a candidate.

"We do need to be careful," he added.

"We have little control over the data that is being collected about us."

Related Internet Links

The BBC is not responsible for the content of external sites.