Tesla CEO Elon Musk said, “Twitter said I leaked the robot sample size and violated a non-disclosure agreement.”
In its earnings report for the first quarter of this year, Twitter acknowledged that there were many “fake or spam accounts” on its platform. “We conducted an internal review of a sample of accounts and estimated that the average number of fake or spam accounts in the first quarter of 2022 was less than 5% of our mDAU for the quarter,” the company reported.
Twitter also admitted to inflating its user numbers between 1.4 million and 1.9 million users over the past 3 years. “In March 2019, we introduced a feature that allows people to link multiple separate accounts together to easily switch between accounts,” the company said. Twitter disclosed that “there was a bug at the time, so action through the master account resulted in all linked accounts being counted as mDAUs.”
While Musk may have reason to be curious, experts on social media, disinformation and statistical analysis say the analysis he proposes is severely lacking.
Without providing evidence, Musk randomly selected 100 accounts as the sample size for his study, arguing that this calculated the percentage of fake/spam/duplicate accounts.
Twitter declined to comment when asked if his description of his approach was accurate.
Facebook co-founder Dustin Moskovitz weighed in on the issue via his own Twitter account, noting that Musk’s approach was not actually random, used too few samples, and left room for a lot of errors .
BotSentinel founder and CEO Christopher Bouzy said in an interview that his company’s analysis suggests that 10 to 15 percent of accounts on Twitter are likely to be “inauthentic,” including fakes, spammers, scammers , malicious bots, duplicate and “single-purpose hate accounts” that often target and harass individuals and others who knowingly spread false information.
Backed primarily through crowdfunding, BotSentinel uses a combination of machine learning software and a human moderation team to independently analyze and identify inauthentic activity on Twitter. The company today monitors more than 2.5 million Twitter accounts, mostly English-speaking users.
University of Washington professor Carl T. Bergstrom said a sample size of 100 is orders of magnitude smaller than the norm for social media researchers studying this sort of thing. The biggest problem Musk faces with this approach is called selection bias, the inability to understand what Musk can do other than fool us with this stupid sampling scheme.