Researchers have found that computer predictions of personality traits based on Facebook 'likes' are more accurate than those made by participants' own Facebook friends.
The "big five" psychological traits were measured in more than 86,000 volunteers: openness, conscientiousness, extroversion, agreeableness and neuroticism.
By analysing just 10 'likes' the software used was more accurate at judging someone's character than a work colleague. It took 70 'likes' to beat a friend, and 150 to outperform a parent or sibling.
It's estimated the average user has racked up 227 'likes', but more than 300 were required to surpass a spouse's character assessment.
What's more, computers were better at predicting all but one of 13 real-life occurrences that the Cambridge University academics analysed - things like political leanings, depression and physical health - and with four of those they were even better than individuals at predicting their own outcomes, including substance abuse.