A study recently published in PNAS concludes that a computer is able to more accurately predict a person’s personality than their own friends and family … from the analysis of some of the data we have left in Facebook .

The researchers conclude that, by analyzing 10 “likes,” a computer can describe our personality better than our co-workers; with 70, better than our friends or roommates; with 150, better than a family member; and with 300, better than a spouse. This shows that machines, despite not having the social skills to interpret language and human intentions, may be able to make valid judgements about us by accessing our fingerprint on the Internet .

Facebook knows you better than your own friends

A personality test based on the Big Five model was provided to 86,220 people for this research. Each of them had to fill out these 100-item forms designed to record information about the different traits that define the way we act, perceive and feel things.

In addition to the information obtained through the personality tests, some volunteers also gave their permission for the research team to analyze the “I like” they had given from their Facebook accounts. These “likes” were not those that can be given by clicking on Facebook statuses, photos or videos, but those associated with pages about movies, books, TV shows, celebrities, etc.

Later, a software found trends and relationships between personality traits and certain preferences for one or another page located on this social network. For example, it was found that people with a high score in the trait “Openness to Change” tend to show a liking for Salvador Dalí or the TED Talks, while the extraverted ones show a liking for dancing. This may be a conclusion based on stereotypes, yet there is empirical evidence to support these ideas.

While the software played at learning how human behaviour works, a group was formed with the other evaluators who were to predict the personality scores of the volunteers. This group was made up of friends, family and acquaintances of the participants who had completed the test. Each of these flesh and blood judges had to describe the personality of the tested subject by filling out a questionnaire. The results (somewhat humiliating for our species) that lead the article emerged by comparing the degree of accuracy with which humans and machines predict personality scores. Only a husband or wife can rival the computer-generated personality models from a few data obtained by Facebook.

Electronic brains

How can software speak so accurately about aspects that define us and make us unique? The biggest advantage they have over us is their access to massive amounts of personal information and their ability to put data in relation to each other and find patterns of behaviour in fractions of a second. Thanks to this, computer-generated personality models can predict certain behaviour patterns automatically, without the need for social skills and with more accuracy than humans.

As a consequence, today we are closer to knowing the features certain aspects of people’s psychology without the need to interact with them face to face , after information about movies, books and celebrities we like passes through an algorithmic kitchen. Taking into account that the average “I like” that each of us has accumulated on Facebook is around 227, we can imagine what this innovation in psychometrics means for statistics centres, recruitment agencies or even groups dedicated to espionage and social control. All this makes the website created by Mark Zuckerberg more like a market segmentation tool than a social network.

Moreover, the consequences this can have for the world of advertising and marketing are obvious. If today it is already possible to roughly estimate a person’s tastes and hobbies from their Google searches, perhaps in the future a car brand will be able to know which model can attract us most from the fact that one day we made around twenty clicks on a social network.

One of the paradoxes of this psychological evaluation methodology is that it studies qualities that make us social and unique beings without the need for social interaction and applying generic rules on human behavior. This perspective can be so appealing to organizations that the University of Cambridge already has an application that allows you to see what your Facebook profile, tweets and other forms of fingerprints say about your psychological profile. One of the supposed advantages that can be read on their website is: “avoids having to ask unnecessary questions”. How this methodology will affect privacy protection remains to be seen.

Big Data: Facebook and its database

In short, it is now possible that computers are increasingly able to infer information about us that we have never stated directly, and that this information is of higher quality than that inferred by anyone else. All this can be made possible, to a large extent, by the analysis of Big Data on Facebook : the massive processing of data (personal or otherwise) that we provide of our own free will. The team of researchers talks about this qualitative leap in the conclusions of their article:

Popular culture has come to represent robots that surpass humans when it comes to making psychological inferences. In the film Her, for example, the protagonist falls in love with his operating system. Through the administration and analysis of his fingerprint, his computer can understand and react to his thoughts and needs much better than other humans, including his girlfriend and close friends. Our research, along with advances in robotics, provides empirical evidence that this scenario is becoming increasingly possible as digital assessment tools mature.

What will computing be capable of when a computer is able to read not only Facebook pages, but also photographs and text with the same level of accuracy ? Will we be beings without any mystery under the gaze of mass-produced processors? Whether this form of understanding of the human being that machines may reach in the future reflects our essence as sentient and unique people is something worth reflecting on.

Bibliographic references:

  • Youyou W., Kosinski, M. and Stillwell, D. (2015). Computer-based personality judgments are more accurate than those made by humans. PNAS 112(4), pp. 1036 – 1040.