Automated accounts have become more sophisticated and complex in recent years. Many fake accounts are operated in part by humans and machines, or amplify messages written by real humans (what Menczer calls “cyborg accounts”). Other accounts use tricks designed to evade human and algorithmic detection, such as quickly liking or disliking tweets or posting and deleting tweets. Of course, there are plenty of automated or semi-automated accounts that aren’t really harmful, such as those operated by many companies.
The Botometer algorithm uses: machine learning It evaluates not only the content of a Tweet, but also a wide range of public data linked to an account, such as when the message was sent, and who follows the account to determine whether it is a bot. Algorithms are state-of-the-art, but Menczer notes that “many accounts now fall into the realm where the algorithm is fundamentally uncertain.”
Menczer and others say spotting bots is a cat-and-mouse game. But they are spammers Algorithms that can generate persuasive text and maintain a consistent conversation.
Since Twitter itself has access to more data about each account, it can use machine learning to better spot bots. This includes a complete history of your activity and the various IP addresses and devices you use. However deliph raoA machine learning expert who studied spam detection on Twitter from 2011 to 2013 says Twitter may not be able to disclose how this works. Because doing so may disclose personal data or information that can be used to manipulate the platform’s recommendation system.
This week, Musk had an argument with Twitter’s CEO Parag Agrawal about how easily the company could reveal a methodology for finding bots. Monday, Agrawal posted a thread It explains how complex the problem is still. He noted that personal data held by Twitter could change the calculation of the service’s bot count. He wrote in the thread, “FirstnameBunchOfNumbers with no profile picture and weird tweets may look like bots or spam, but behind the scenes we often see multiple indications that they are real people.” Agrawal also said Twitter was unable to release details about these ratings.
The deal could remain opaque if Twitter says it can’t or is unwilling to disclose its methodology and Musk won’t proceed without details. of course, Musk is leveraging this issue. to negotiate a price.
For now, Musk doesn’t seem to be satisfied with Twitter’s efforts to explain why finding bots isn’t as easy as it seems. He responded to Agrawal’s long thread on Monday: simple message It seemed a lot better suited for bots than potential buyers on Twitter. One smiling poop emoticon.
Why It’s Hard to Count Twitter Bots
Source link Why It’s Hard to Count Twitter Bots