Elon Musk has wrong approach to count fakes, spam on Twitter: experts

Tesla Sent by CEO Elon Musk twitter Stocks falling on Friday when he said he was going to put him $ 44 billion acquisition of “pending” social network He is investigating the percentage of fake and spam accounts on the platform.

Musk later revealed that he continued to commit to the deal, but he continued to hammer the fake account issue. He wrote on Twitter that his team conducted their own analysis and questioned the accuracy of the numbers that Twitter reported in its latest financial report.

among them First quarter financial report of this year, Twitter has acknowledged that there are numerous “fake or spam accounts” on the platform, in addition to legitimate monetizable daily active use or users (mDAU). “We conducted an internal review of account samples and estimated that the average number of fake or spam accounts in the first quarter of 2022 was less than 5% of mDAU during the quarter,” the company reports. bottom.

twitter We also admit that we have overestimated 1.4 to 1.9 million users over the last three years. “In March 2019, the company launched a feature that allows you to easily switch between accounts by linking multiple separate accounts.” “Because of the error at that point, the action taken through the primary account counted all linked accounts as mDAU.”

Mask may be curious, of course, but experts in social media, disinformation, and statistical analysis say his proposed approach to further analysis is terribly inadequate.

This is SpaceX Tesla The CEO said he would do it to identify the number of spam, fake and duplicate accounts on Twitter.

“To find out, my team creates a random sample of 100 followers on @twitter. Repeat the same process and invite others to see what they discover. He clarified his methodology in subsequent tweets, “Choose an account with a lot of followers”, “Ignore the first 1000 followers and then choose every 10 days. Better ideas We are accepting. “

Musk also said that he chose 100 as the sample size number for his study because it is the number Twitter uses to calculate revenue report numbers without providing evidence.

“Any wise random sampling process is fine. If many people get similar results on their own for the percentage of fake / spam / duplicate accounts, you know. For Twitter to calculate <5. I chose 100 as the sample size number because I want to use it.% Fake / Spam / Duplicate. "

Twitter declined to comment when asked if the methodology description was accurate.

Facebook Co-founder Dustin Moskovitz considered this issue through his Twitter account and pointed out that Mask’s approach isn’t really random and there’s room for big errors because there are too few samples to use. Did.

He also feels that he “doesn’t trust the Twitter team to help pull out samples.” It’s a danger signal of its own kind, “he said.

Bot Sentinel Founder and CEO Christopher Bouzy said in an interview with CNBC that 10% to 15% of Twitter accounts were fake, spammers, scammers, malicious bots, duplications, “single-purpose”, according to his company’s analysis. “It is highly possible that it is not genuine.” A “hate account” that usually targets and harasses individuals, along with others who knowingly disinformation.

Supported primarily through crowdfunding, BotSentinel uses a combination of machine learning software and a team of human reviewers to independently analyze and identify malicious activity on Twitter. The company currently monitors over 2.5 million Twitter accounts, primarily English users.

“I don’t think Twitter really categorizes’fake and spam’accounts,” Bouzy said.

He also warns that the number of fraudulent accounts can increase or decrease in various corners of Twitter, depending on the topic being discussed. For example, BotSentinel found that more fraudulent accounts tweeted about politics, cryptocurrencies, climate change, and covid than accounts discussing indisputable topics such as kittens and origami. ..

Co-authored Professor Karl T. Bergstrom of the University of Washington Books that help people understand the data Also, to avoid getting caught up in false claims online, sampling 100 followers on a single Twitter account to CNBC acts as a “due diligence” for a $ 44 billion acquisition. He said it shouldn’t be.

He said the sample size of 100 is orders of magnitude smaller than the standard of social media researchers studying this kind of thing. The biggest problem Mask faces with this approach is known as selection bias.

In a message to CNBC, Bergstrom said, “There is no reason to believe that followers of an official Twitter account are a representative sample of accounts on the platform. Perhaps bots may follow this account to avoid detection. Will be lower. Probably they are more likely. Follow to look legitimate. Who knows? But other than the mask trolling us with this ridiculous sampling scheme I can’t understand what you’re doing. “