Elon Musk wants to know how many fake accounts Twitter has, but experts say his approach is all wrong

Musk said on Friday that he was going to put his $44 billion acquisition of the social network on hold, while he looked at the number of fake accounts on the platform.

Musk continued to hammer on the issue of fake accounts even after he clarified that he remains committed to the deal. He wrote that his team would do their own analysis and expressed doubt about the accuracy of the numbers that were reported in the most recent financial filings.

In its first-quarter earnings report this year, the company acknowledged that there are a number of false or fake accounts on its platform. An internal review of a sample of accounts has shown that the average number of false or fake accounts during the first quarter of the year was less than 5% of the mDAU.

Over the past 3 years, the company has overstated user numbers by over one million. The feature that allowed people to link multiple accounts together was launched in March of 2019.

Experts in social media, disinformation and statistical analysis say that Musk's suggested approach to further analysis is woefully deficient.

The CEO of the two companies said he would look into how many fake and duplicate accounts exist on the micro-blogging site.

To find out, my team will do a random sample of 100 followers. I invite others to do the same thing and see what they discover. I am open to better ideas.

Musk said that he picked 100 as the sample size number for his study because that's the number they use to calculate the numbers in their earnings reports.

Any random sampling process is fine. It will be telling if many people get the same results. The sample size number was 100, because that's what Twitter uses to calculate fake/spam/duplicate.

When asked if his description of its methodology was accurate, he declined to comment.

Moskovitz pointed out that Musk's approach is not random, uses a too small sample, and leaves room for massive errors.

I feel like I don't trust the Twitter team to help pull the sample, is that its own kind of red flag?

In an interview with CNBC, Christopher Bouzy, the founder and CEO of BotSentinel, said that 10% to 15% of accounts are likely to be fake.

BotSentinel uses a mix of machine learning software and teams of human reviewers to independently analyze and identify inauthentic activity on social media. More than 2.5 million accounts are monitored by the company.

I think that it's not realistic to classify accounts as false and spam.

The number of inauthentic accounts can appear higher or lower depending on the topic being discussed. BotSentinel has found that more inauthentic accounts talk about politics, climate change, and covid than about non-controversial topics.

“I just can’t fathom that Musk is doing anything other than trolling us with this silly sampling scheme.”

Carl T. Bergstrom, a University of Washington professor who co-authored a book to help people understand data and avoid being taken in by false claims online, told CNBC that sampling one hundred followers of any single account should not serve as due diligence.

The norm for social media researchers is a sample size of 100. selection bias is the biggest issue Musk would face with this approach.

Bergstrom wrote in a message to CNBC that followers of the official account are not a representative sample of accounts on the platform. Maybe bots are less likely to follow this account. Maybe they are more likely to follow to be legit. Who knows? I can't believe that Musk is doing anything other than messing with us.