Why is it so hard to count bots on Twitter?

- Advertisement -


This is Twitter Account @ElonMask bot? One of the best algorithms to identify fake accounts thinks it might bewhich shows how difficult it is to quantify the proportion of fake accounts on a social network.

- Advertisement -

The bot count on Twitter has become the subject of controversy in Elon Muskcurrent $44 billion acquisition of Twitter. Billionaire last Friday tweeted that it has “temporarily suspended” its purchase until the company provides details in support of its claim (as stated in his latest SEC filing) that less than 5% of Twitter’s “monetizable daily active users” are spam or fake. Musk also mentioned plan count the bots yourself, which required a sample of 100 @Twitter subscribers to see how many bots there were and said approach suggests that more than 20 percent of the accounts are fake.

- Advertisement -

But accurately determining the percentage of bots on Twitter, according to experts, is much more difficult.

Finding them is not difficult if you know where to look. Some accounts, including Musk’s, seem to attract many of them. “If you just mention Elon Musk on Twitter, you will immediately get involved in a lot of cryptobots,” he says. Chris Baleprofessor of sociology at Duke University who studies social networks.

- Advertisement -

Twitter is not the only social network that is cracking down on fake accounts. facebook removes billions fictitious accounts Every year. But it’s hard to say for sure that a Twitter account is a bot, as legitimate users may have few followers, rarely tweet, or have strange usernames. It is even more difficult to estimate the number of bots running on the platform as a whole.

To test what Musk suggested methodology, V.ayThe AI ​​firm, which previously identified bot activity among accounts spreading misinformation about voter fraud in the US, looked into 100 accounts that spy on Musk’s car manufacturing company. Tesla on Twitter.

An algorithmic account check on Tuesday found that more than 20 accounts out of 100 have a high probability of being bots. A manual check of the same 100 concluded that more than half could be bots. And an analysis of the topics discussed by these accounts found no evidence that any of the suspected accounts were ads. But many of those accounts also disappeared soon after, which suggests that Twitter is catching bots pretty quickly. Vince LynchCEO of IV.ai, says identifying questionable accounts is also inherently subjective and involves a degree of uncertainty.

“It’s a very difficult problem,” he says. Filippo Menzerprofessor at Indiana University who led the development Botometer algorithm, which gave Musk’s account a relatively high bot score. Mentzer says looking at 100 accounts won’t be representative of daily active Twitter users, and different samples will produce wildly different results. “I want to hope it was a joke,” Mentzer says of the methodology.

In recent years, automated accounts have become more sophisticated and complex. Many fake accounts are operated in part by humans as well as machines, or simply amplify messages written by real people (what Mentzer calls “cyborg accounts”). Other accounts use tricks to avoid human and algorithm detection, such as quickly liking and unliking tweets, or posting and deleting tweets. And, of course, there are plenty of automatic or semi-automatic accounts, such as those run by many companies, that aren’t actually harmful.

The botometer algorithm uses machine learning to evaluate a wide range of public data associated with an account – not only the content of tweets, but also the time of sending messages, account subscriptions, etc. – to determine the likelihood that this is a bot. While the algorithm is state of the art, Mentzer says that “many accounts now fall into a range where the algorithm is basically not very accurate.”

Mentzer and others say that bot detection is a game of cat and mouse. But they add that this could become much more difficult in the future as spammers use algorithms that are better able to generate persuasive text and have coherent conversations.

Twitter itself is better equipped to detect bots through machine learning because it has access to much more data about each account. This includes a complete user activity history as well as the various IP addresses and devices they use. But Delip Raoa machine learning expert who worked on Twitter spam detection from 2011 to 2013 says the company may not reveal how it works because it could expose personal data or information that could be used to drive a recommendation system platforms.

This week, Musk also fell out with Parag Agrawal, CEO of Twitter, over how easily the company can reveal its bot-finding methodology. Monday Agrawal published a topic explaining how challenging the task remains. He noted that private data held by Twitter could change the calculation of the number of bots on the service. “FirstnameBunchOfNumbers without an avatar and strange tweets may seem like a bot or spam to you, but behind the scenes we often see several signs that this is a real person,” he wrote in the thread. Agrawal also said that Twitter could not disclose the details of these estimates.

If Twitter is unable or unwilling to reveal its methodology, and Musk says he won’t go ahead without details, the deal could remain in limbo. Certainly, Musk uses the problem as leverage negotiate a price down.

For now, Musk seems unhappy with Twitter’s attempts to explain why finding bots isn’t as easy as he thinks. He replied to Agrawal’s long thread on Monday: simple message it seemed much more appropriate for a bot than for a potential Twitter buyer: one smiling poop emoji.

.


Credit: www.wired.com /

- Advertisement -

Stay on top - Get the daily news in your inbox

DMCA / Correction Notice

Recent Articles

Related Stories

Stay on top - Get the daily news in your inbox