Jointly run between Stanford University and the University of Cambridge, the study sought to find out how a computer's answers to questions about an individual would compare to those of their friends and family. A total of 86,220 volunteers provided answers to questions about themselves regarding the five basic personality dimensions of openness, conscientiousness, extraversion, agreeableness and neuroticism. A standard, 100-item long personality questionnaire was used to gather the data.

Friends and family of each volunteer were then asked to provide their judgement of the individual's personality based on their existing knowledge of the person, via a 10-item questionnaire. The same questions were put to a computer model that answered them based on the analysis of what articles, videos, artists and other items the volunteers had "liked" on Facebook.

The relationship of human respondents to the volunteers was categorized under titles including colleague, friend, roommate, family-member or spouse. It was found that the computer model was able to more accurately predict answers of the volunteers than a colleague by analyzing 10 likes, than a friend or a roommate with 70 likes, than a family-member with 150 likes and than a spouse with 300 likes.

"This is an emphatic demonstration of the ability of a person's psychological traits to be discovered by an analysis of data, not requiring any person-to-person interaction," the researchers noted. "It shows that machines can get to know us better than we'd previously thought, a crucial step in interactions between people and computers."