The takeover of Twitter by Elon Musk had the effect of a tsunami. Then in a few weeks, nothing. After the storm, the calm. Now the richest man in the world is even threatening not to make his $44 billion purchase. His excuse? The excessive presence of fake accounts, “bots”, on the social network. “20% fake accounts, four times what Twitter claims, could be much higher,” the Tesla boss said on his Twitter account on Tuesday. Before alerting: “My offer was based on the accuracy of Twitter’s statements”.
But why do fake accounts create such a rift between the social network and Elon Musk? 20 minutes tries to answer it in five questions.
Alright, what is a bot?
Far from the image we have of the robot destroying humanity, a bot is quite simply a fake account that rages on social networks, and more widely on the Internet. On Twitter, for example, you can chat with a certain “Sandrine Dupont”. Your interlocutor displays a completely realistic profile picture and her writing leaves no doubt that a real person occupies this account. Except that Sandrine Dupont does not exist and the bot is in fact managed either by another human or by an artificial intelligence. “Spam is not just ‘binary’ (human/non-human). The most advanced spam campaigns use combinations of coordinated humans and automation,” explains Twitter director Parag Agrawall, in a lengthy thread posted on his account.
But why create a fake account? Companies, for example, can use this system to influence consumers. But they are certainly not the only ones setting up such a game. Soon, bots were also used to manipulate citizens, including by widely inflating trending topics online. It is not uncommon, for example, to see fake accounts relaying the words of a candidate during an election. The goal ? Make your words more audible, and thus give them more importance.
Ok, but are bots really that dangerous?
By definition, bots are dangerous when they influence or even manipulate Internet users. “They compromise the real accounts, then use them to advance their campaign,” says Parag Agrawall, who also believes that these fake accounts remain “sophisticated” and “difficult to catch”. Indeed, the bots learn from their mistakes and are constantly changing to counter their future blockages. “The adversaries, their goals and their tactics are constantly changing – often in response to our work! You can’t build a set of rules to detect spam today and expect them to work tomorrow,” Parag Agrawall points out.
But couldn’t we regulate them, these bots?
It’s written in black and white in the rules and policies of Twitter “You may not use Twitter’s services in any way that seeks to artificially amplify or suppress information, or engage in behavior that manipulates or disrupts users’ experience.” Among these prohibitions, we find “spam for commercial purposes”, but also “inauthentic engagements which attempt to make accounts or content appear to be more popular or active than they actually are”.
There is therefore indeed a regulation on Twitter, which would go from the request for identification of the account until its definitive suspension. Except that, as we mentioned earlier, fake accounts are much more sneaky than simple regulation and still manage to slip through the cracks. “We suspend over half a million spam accounts every day, usually before you even see them on Twitter. We also lock millions of accounts every week that we suspect are spam – if they can’t pass human verification challenges (captchas, phone verification…),” the Twitter CEO wanted to reassure. tuesday.
However, the limit of regulation arises when what appeared to be a fake account was not one. “A difficult challenge”, according to Parag Agrawa. Who adds: “Some of the spam accounts that are actually the most dangerous – and cause the most harm to our users – may seem totally legitimate at first glance. »
Why can’t we count them?
Since Elon Musk’s backpedal on his takeover, the presence of bots on the social network seems to have become impossible to quantify. According to Twitter, the number of fake accounts does not exceed 5%, but the boss of Tesla is convinced that this figure rises to 20%. Why such a discrepancy? First, because many fake accounts are still invisible today. But not only. The two camps do not seem to have chosen the same sampling to measure the magnitude of bots on Twitter.
“Our estimate is based on multiple (replicated) human reviews of thousands of accounts, which are randomly sampled, consistently over time […]. We do it every quarter, and we have done it for many years, ”explains the boss of Twitter. For his part, Elon Musk announced a completely different methodology, which had the merit of awakening the mockery of budding statisticians. “My team will take a random sample of 100 followers from the social network’s Twitter account,” he said on Twitter on May 14.
What to react the professor at the University of Washington Carl T. Bergstrom, who declared on CNBC: “There is no reason to believe that the subscribers of the official Twitter account constitute a representative sample of the accounts on the platform. shape. Maybe bots are less likely to follow this account to avoid detection. Maybe they are more likely to follow to seem legit. Who knows ? But I just can’t imagine Musk doing anything other than trolling us with this idiotic sampling scheme.”
Why does this worry Elon Musk?
In general, a fake account is an account that cannot be targeted by advertising. Except that Twitter mainly derives its revenue from advertising, largely paid for by advertisers. But they want to talk to consumers and potential buyers, not robots.
Worse still for Twitter, advertisers pay according to the number of users, a figure which could therefore be inflated by the number of bots. If their proportion turns out to be really close to 20% rather than 5%, advertisers would risk fleeing little by little and creating heavy economic losses for the social network… Enough to cool Elon Musk.