According to Cambridge University researchers, the content that users tag with a ‘like’ on Facebook can be used to glean far more information about them than is strictly comfortable.
The researchers’ report, published on PNAS, claims a high degree of accuracy in predicting sexual orientation, political affiliation and even ethnicity based around which items subjects have ‘liked’ on Facebook. Aspects of a person’s personality have also been deduced.
The study made use of 58,000 volunteers who provided access to their Facebook Likes as well as their demographic information and psychometric test results. The Facebook information was fed into a logistic/ linear regression model. The results were somewhat surprising in some cases, as what was Liked on the social network was not always an explicit indicator of a given person’s personality. Also, Liking curly fries on Facebook appears to have been tied to intelligence, according to the report’s author David Stillwell.
The algorithms used for the test were able to predict a man’s sexual orientation with an accuracy of 88%, political leanings (between Democrat and Republican) with an accuracy of 85% and it was able to distinguish between Christians and Muslims with an accuracy of 82%. It was also able to detect relationship status and whether a user was given to drug or alcohol abuse with lower accuracy totals.