Automated accounts have become more sophisticated and complex in recent years. Many fake accounts are partly operated by humans as well as machines, or simply amplify messages written by real people (what Menczer calls “cyborg accounts”). Other accounts use tricks designed to evade human and algorithmic detection, such as liking and not liking them quickly, or posting and deleting tweets. And of course, there are many automated or semi-automated accounts, such as those run by many companies, that aren’t really harmful.
The Botometer algorithm uses machine learning to evaluate a wide range of public data linked to an account, not just the content of tweets, but when sending messages, who follows an account, etc., to determine the likelihood that it is a bot. . Although the algorithm is state-of-the-art, Menczer says, “many accounts now fall into the range where the algorithm is basically not very secure.”
Menczer and others say that detecting robots is a cat-and-mouse game. But they add that it could be much more difficult in the future, as spammers use algorithms that are better able to generate compelling text and maintain consistent conversations.
Twitter itself is better equipped to detect robots that use machine learning because it has access to much more data about each account. This includes the complete activity history of a user, as well as the different IP addresses and devices they use. But Delip Rao, a machine learning expert who worked on spam detection on Twitter from 2011 to 2013, says the company may not be able to disclose how this works because doing so could reveal personal data or information that could be used. to manipulate the platform recommendation. system.
This week, Musk also had a discussion with Twitter CEO Parag Agrawal about how easily the company could reveal its methodology for finding robots. Monday, Agrawal posted a thread explaining how complex the challenge is still. He noted that Twitter’s private data could change the calculations on the number of bots in the service. “FirstnameBunchOfNumbers without a profile picture and weird tweets may sound like a bot or spam, but behind the scenes we often see several indicators that this is a real person,” he wrote in the thread. Agrawal also said Twitter could not disclose details of those assessments.
If Twitter can’t, or doesn’t want to, reveal its methodology and Musk says it won’t proceed without details, the deal may remain in limbo. Of course, Musk is using the problem as leverage to negotiate the price down.
For now, Musk seems dissatisfied with Twitter’s efforts to explain why finding robots isn’t as easy as he thinks. He responded to Agrawal’s long thread on Monday with a simple message this seemed much more appropriate for a bot than for a potential Twitter buyer: a unique, smiling poop emoji.