How does the data you share unintentionally allow Facebook to know you better than you?

When we talk about privacy and personal data, we soon remember data such as address, telephone, ID number, marital status, photographs, employment, and other data like this. These data are concrete, cadastral, their uses are objective, their documents can be cloned, criminals can open companies and commit crimes with them, their photos can be manipulated, someone can surprise you at the door of home or work, but all this is well in the material world of the twentieth century.

We can not continue thinking about the 21st century with the mind of the last century, these personal data are the least relevant, as the consequences of its use by third parties are concrete and objective, it is possible to measure the damages establish a certain contingency and / or repair.

Subjective data

Most of the data we produce is subjective, and they are also the data that most tell us about ourselves. Your ID number will not reveal your psychological profile, your likes and interests, for example, but your daily use of social networks will.

A click, a like, a love or a share, are data that you produce about your subjectivity, you should not imagine, but these are your most precious data. This data allows Facebook or Google for example, to know you better than yourself, and to make billions of dollars with them.

In the article “Computer-based personality judgment are more accurate than those made by humans“, the authors Wu Youyou, Michal Kosinski and David Stillwell, did a study with more than 80,000 volunteers. They  concluded, as described in the graph above, that a little more 10 likes allow Facebook to know you better than your co-worker, 70 likes more than a friend, with 100 likes he can know you better than the average human, and 300 likes allow Facebook to know you better than your or your life partner.

Perhaps you have not yet realized the right size, so imagine that your wife or husband know how to cheer you, or irritate you, even what kind of stimuli may be more or less intense in this endeavor, know how to persuade you, or when you are trying to persuade her. It is a very deep level of knowledge of the other, in general we take years of joint life to develop it. But for Facebook, a mere 300 likes are enough.

These are data that Cambridge Analytics had access to, but not only then, those countless Facebook tests like : “What would it look like if it were the opposite sex?” That ​​companies that offer these tests on Facebook may be selling your psychometric profile there, imagine now in the elections.


If you’re still not convinced, then I suggest you take the Cambridge University Psychometrics Center test, you can do it without fear, it’s a serious academic project with a strict privacy policy, and that will surprise you.

How can my likes say so much about me?

The data that you produce voluntarily and involuntarily, along with the data produced in the same way, by billions of other people configure what is known as big data. The more data we produce, the more accurate the profiles and assessments that the big data mining produces on us, the sooner we will have a level of accuracy in the human sciences next to the exact ones.

This process of data mining is called modeling, different from statistics, modeling requires a sample at least 10 times higher, in fact the larger the sample the better is the result. As researcher Fernanda Bruno points out, data capitalism shuffles the boundary between the laboratory and real life.

The data capitalism that imposes itself with the social networks, shuffles the boundaries between the laboratory and the social, political and subjective life. (Fernanda BRUNO, 2018, our translation)

Modeling is done initially by testing and mapping known data, even by paper and pen tests. Explaining in a very simplified way, the researcher performs the test of the five factors (neuroticism or emotional instability, extroversion, pleasantness, conscientiousness and openness to the experience) with thousands of volunteers.

These five elements allow to define the personality traits of the individual with great precision, although with a focus on stereotypes. Once obtained the result of thousands of tests done directly with the volunteers, it is the time of machine learning. Through a set of algorithms (computer programs) perform numerous analyzes of the data in our Facebook example, searching patterns for each trait and personality factor, comparing the patterns found in the network with the results of the five-factor test.

Contrary to what one usually imagines patterns do not follow any predictable human logic, one is fooled who thinks that a person who liked the page of the right or left candidate has the same tendency, often a like on a page that does not has no relation with politics can say much more. That’s what computer scientist Jennifer Golbeck, who demonstrates in this TED talk, you have no idea what these little potato chips can reveal …

Once the patterns are discovered and confirmed, the reverse process is possible by connecting to your Facebook account, the algorithm in seconds analyzes your likes pattern compared to the patterns it has, and gets your psychometric profile.

But the question does not end here, this subject is deeper and more complex, my research now goes far beyond what has been said, and I am sure that I have not even come halfway.

I hope it has been enlightening for you, more texts like this will be posted here, do your part, spread it, let people know how their privacy can be so easily invaded.


ADALI, S.; GOLBECK, Jennifer. Predicting personality with social behavior: a comparative study. Social Network Analysis and Mining, [s.l.], v. 4, no 1, p. 159, 2014. ISSN: 1869-5450, DOI: 10.1007/s13278-014-0159-7. Disponível em:

BRUNO, F. A economia psíquica dos algoritmos: quando o laboratório é o mundo – Nexo Jornal. Nexo Jornal. 2018. Disponível em:íquica-dos-algoritmos-quando-o-laboratório-é-o-mundo. Acesso em: 12/jun./18.

WU, Youyou; KOSINSKI, Michal; STILLWELL,David. Computer-based personality judgments are more accurate than those made by humans. PNAS, 112 N4: 1036–1040, 2015. doi: doi/10.1073/pnas.1418680112. Disponível em:

If you are a native English speaker, I would like to hear some criticism and suggestions about my translation, and in advance thank you for your cooperation.


Leave a Reply

Avatar placeholder

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.