Automated accounts have become increasingly sophisticated and complex in recent years. Many fake accounts are operated partly by humans, as well as by machines, or only amplify messages written by real people (what Menczer calls “cyborg accounts”). Other accounts use tricks designed to avoid human and algorithmic detection, such as quickly liking and unliking tweets or posting and deleting tweets. And of course, there are plenty of automated or semi-automated accounts, such as those run by multiple companies, that aren’t actually harmful.
The Botometer algorithm uses machine learning to evaluate a variety of public data related to accounts — not just the content of tweets but also when messages were sent, who follows the account, etc. — to determine the likelihood that the account is a bot. While the algorithm is state-of-the-art, Menczer says, “a lot of accounts now fall into a range where the algorithm is fundamentally uncertain.”
Menczer and others say bot detection is a game of cat and mouse. But they add that it could become significantly more challenging in the future as spammers use algorithms are better able to produce persuasive text and hold coherent conversations.
Twitter itself is better equipped to detect bots using machine learning because it has access to more data about each account. This includes a user’s entire history of activity, as well as the various IP addresses and devices they use. But Delip Raoa machine learning expert who worked on spam detection at Twitter from 2011 to 2013, said the company may not disclose how this works because doing so could reveal personal data or information that may be used to manipulate the platform’s recommendation system.
This week, Musk also argued with Parag Agrawal, Twitter’s CEO, over how easily the company can reveal its method of finding bots. On Monday, Agrawal posted a thread explains how complex the challenge remains. He noted that the personal data Twitter holds could change the calculations for the number of bots on the service. “FirstnameBunchOfNumbers has no profile picture and quirky tweets may sound like a bot or spam to you, but behind the scenes we often see many signs that it is a real person,” he wrote in the thread. subject. Agrawal also said that Twitter could not disclose the details of these reviews.
If Twitter is unable or unwilling to disclose its methodology, and Musk says he won’t proceed without details, the deal could remain in limbo. Of course, Musk can use this problem as leverage to negotiate the price down.
For now, Musk seems unhappy with Twitter’s attempt to explain why finding bots isn’t as easy as he thought. He replied to Agrawal’s long thread on Monday with a simple message that seems much more appropriate for a bot than a potential Twitter buyer: a single smiling poop emoji.
