New technology predicts your next tweet

Since the appearance of social networks and the person himself and his personal data have become their main resources, people have been inventing more and more new ones capable of maintaining anonymity. The most radical is to delete your accounts or not write anything on the Internet at all. Recently, however, an international team of researchers in the field of social engineering has developed a system for predicting human actions, for which a few tweets of his acquaintances will suffice.

The study was based on the concept of entropy, which, when applied to tweets, means the ratio in the text of random words and expressions to ordered and weighted ones. The experts analyzed several thousand random accounts with the number of subscribers from 50 to 500 and more or less noticeable activity. The entropy data was plugged into the Fano equation and calculated the probability of guessing the first word in the next tweet from the user. It was 53%.

The researchers then added pooled entropy data to the equation for the 15 most active followers of the account, and the prediction accuracy increased to 60%. And then they removed the user's data, leaving only his followers, and got a 57% chance. This means that in theory we can predict what a person will write in a certain situation, knowing only the tweets of his followers. We do not need access to his personal channel to make such a forecast, he may not turn on the computer at all, and the conclusions will already be drawn.

The scope of the methodology is not limited to Twitter, it is applicable outside the Internet, it is just that it is more difficult to collect and analyze information there. In a way, this is a rethinking of the saying "tell me who your friend is, and I will tell you who you are." Suddenly, it turned out that there are many markers, from people's behavior in games to the tendency to put dislikes under photos in order to predict the personality traits of their acquaintances, the worldview of work colleagues or the political predilections of neighbors. This is a completely new level of processing information about a person, which has yet to be mastered.