Facebook and Twitter know your profile even if you do not have an account, delete it or do not use it

Facebook and Twitter know your profile even if you do not have an account, delete it or do not use it

It is easier to predict someone's interests by their friends than by their own words, according to a new study

Facebook and Twitter always want to know who our friends are. "Find people you know, import your Gmail contacts," they say on Twitter. "People you may know," Facebook offers. Who gives this information should know that it is not only their data, but also those of their friends.

The usefulness of friends to define someone is as old as a saying: tell me who you are going with and I will tell you how you are . But social networks multiply the prediction options of our profile according to our contacts: you can predict someone's profile with 95% success options only from 8 or 9 friends on Twitter, without requiring any information from the person. The figures come from a new scientific article by professors James P. Bagrow, Xipei Liu, of the University of Vermont (USA) and Lewis Mitchell, of the University of Adelaide (Australia) published in Nature Human Behavior.

The important detail in that phrase is "without requiring any information from the person". The implications for privacy in the main networks is obvious: Facebook or Twitter can know what we are interested in without writing anything: just watching our main friends. They do not need us to open an account.

And all that only with public information. Twitter and Facebook manage, nonetheless, non-public information about us: search history and browsing patterns, for example. "With that we can go further, they are more private actions than a public tweet," says David García, a researcher at the Complexity Science Hub in Vienna and author of previous articles that described similar privacy problems in social networks.

Profiles in the shade

Academic research has shown again and again how networks can gather accurate information about our lives without our knowledge and without ever having written a tweet. These are the so-called "profiles in the shade" ( shadow profiles ): profiles of people who do not have an account in the social network. Mark Zuckerberg said last April in the US Congress that he "was not familiar" with the concept. He admitted that Facebook has data from people who are not users of the network, but not that Facebook creates such profiles.

It is possible to test how well you know us, for example Facebook. Any user can delete the account and register again only with their first and last name: another photo, another email and another phone. Apart from the fact that Facebook retains the information that we have aspired to erase, the accuracy of the "people you may know" that Facebook offers is extraordinary. But that is just a use of having "profiles in the shade".

"If they do not get caught, it can be very lucrative," Garcia says. "For example, the amount of money that political parties would pay to have access to that, and you can cross the data with other sources and find the address, mobile number or email of people to send electoral propaganda, or worse, false news to manipulate them, "he adds.

How do Facebook or Twitter manage to know your friend list if you have never downloaded or used the application? From the contact lists that your friends share and where you are. "When you install one of the Facebook or Twitter applications on your mobile, they ask you to access the mobile's contact list, Google and WhatsApp also have it, even if you're not in the network, you appear in the contacts list of many users. , and from there you can build your profile without even touching the network, "says Garcia. Even if you do not say anything, you know what you think.

Prediction tools are increasingly refined. The three authors of the article have found a metric that can tell how a future text is predictable based on your old posts.

The implications of such a tool in the hands of social networks is limited only by the imagination: "The ability to predict allows networks to develop profiles that identify and track individuals and even manipulate the exposure to information, for example, a model of language can be trained to generate new text with the 'voice' of the user and from there the profile can be derived by asking about the possibility of that user making certain type of affirmations, for example, policies ". That is, the model will know what someone will think of a new political controversy according to their opinions in old polemics.