Facebook has a "parallel Facebook" full of bots that simulate malicious user behavior

Must Read

Brian Adam
Professional Blogger, V logger, traveler and explorer of new horizons.

A constant fight of social networks like Facebook or Twitter is against the users who make spamcheat through phishing or just trolle. Facebook now claims to have a new tool to detect them: a parallel Facebook where bots with artificial intelligence imitate this type of users. With this, they can study their behaviour and experiment with measures to see if their activity diminishes.

The other Facebook

To carry out this simulation the company has essentially created an alternative Facebook. Known as ‘WW’ (referring to it being a minor version of the WWW), this simulator works with the same code as the original Facebook used by real users. But in it there are no real users, just a bunch of bots with artificial intelligence that roam freely and imitate users.

An example of how this simulator works. Facebook describes that they trained some “bots bad “for selling illegal products on the platform to others”bots good. “This is a relatively common practice on real Facebook, where it is often about selling items that are prohibited such as weapons or drugs. bots They carry out all the activities that a real user can do: send messages to others, search, visit pages, leave comments … Analyzing how “bots bad guys “were behaving and trying to sell those illegal items, they started to apply different restrictions to see if this prevented these sales from occurring, such as limiting the number of messages sent between strangers.


Once the measures taken show that they can be effective, they will begin to be implemented on real Facebook and real users. The interesting part of this is that being an alternative and simulated Facebook there is no interaction with real users.

Now, as much as it is a simulated Facebook, it is still not exactly like the real Facebook and also cannot simulate all real-world factors. In the end, as much as the bot be smart, cannot predict the exact behaviour and intentions of a human. However, it is an interesting way to test measures to be taken safely and with the certainty that it will not affect end users.

Considering that the bots are equipped with AI, in principle they can serve more than just mimic known human behaviours. In other words, what Facebook engineers really do now is program bots to do specific tasks such as trying to sell an illegal product. But you can also simply give him a goal and find for himself the way to achieve it, if so you could, for example, detect possible behaviours and methods that real users have not yet used and prepare in case they are going to do so.

The project is part of Facebook’s artificial intelligence department, Facebook AI. From here have come out the most curious things, such as a robot that moves without a map, artificial intelligence that solves complex mathematical equations or another that changes the voice of the singers.