Elon Musk’s Full-Court Press On Twitter’s CEO

Twitter CEO, Parag Agrawal, put forward a 16 tweet thread yesterday addressing Elon Musk’s claim that more than 5% of Twitter’s user base is comprised of inauthentic accounts:

In this post, we will dissect several arguments presented by Twitter’s CEO.

Conflicting Interests

Agrawal begins by asserting what he claims to be ‘obvious’ — namely, that Twitter is ‘strongly incentivized to detect and remove as much spam as [they] possibly can, every single day.’

Nearly 90% of Twitter’s revenue comes by way of advertising (the remaining 10% is split between data licensing and other).

And what does spam provide to Twitter? Increased monetization opportunities!

In addition to *literally* creating more advertising inventory/opportunities for Twitter to monetize, spam accounts—including ones that engage in what Twitter refers to as ‘artificial amplification,’ i.e. single-purpose accounts that function solely to retweet and like predetermined accounts/narratives to increase visibility — make Twitter seem very lively. The perceived liveliness of Twitter is likely among the consideration set of many who purchase ads on Twitter’s platform.

In other words, spam has both direct and indirect impact on Twitter’s bottom line, such that properly mitigating against it would adversely impact Twitter revenue.

Another flavor of Twitter spam involves running a script to follow en masse accounts that have a script set on their side to follow back any follow request they receive. It’s a common tactic utilized by nefarious actors to make their accounts appear more popular than reality reflects (i.e. they opt to inflate their followers count).

Despite this (follow-for-follow) behavior being explicitly banned by Twitter Rules, Twitter’s Trust and Safety team cannot seem to get a handle on suspending the accounts engaging in it.

Here’s an example of one such “Resistance” account (@ArresthimNow):

Among the 27.1K accounts being followed by “T*NA … #trUmpIsGuilty!!!”, more than 96% of them (26.1K) are also followers of the @ArresthimNow account. This does not happen organically, and is indicative of 25K+ spam accounts being used to make an account appear more popular than reality reflects.

And even if Twitter were to suspend accounts like @ArresthimNow, unless they were to suspend the much larger underlying network of inauthentic accounts inflating its followers count, it is mind-numbingly easy for nefarious actors to create new accounts that similarly appear to have large followings.

The 24.6K tweets from this spam account present additional monetization opportunities for Twitter; hence, despite what Twitter’s CEO says, this is the very reason that Twitter is NOT ‘strongly incentivized to detect and remove as much spam as [they] possibly can, every single day.’

For a more detailed view into the follow-for-follow behavior that violates Twitter Rules, please see this post.

Grading Their Own Homework

Agrawal continues by highlighting that spam can be ‘sophisticated and hard to catch,’ yet anyone who spends material time on Crypto Twitter, for example, remains puzzled at how Twitter — a ~$30bn market company — has been so ineffective at protecting their users against verified accounts being hijacked/sold and utilized in phishing scams (that involve spamming Twitter users at rates of hundreds of accounts per minute!):

Twitter’s CEO then highlights that Twitter suspends ‘over a half million spam accounts every day,’ as though he and his team deserve some sort of reward for said feat.

Reality: suspending more than a half million spam accounts every day doesn’t indicate you are dealing with inauthentic accounts effectively; rather, it means you have a gaping problem with your account creation process

Agrawal further paints a picture in which the detection of inauthentic accounts/activity is extremely difficult, while also asserting that, as a result of ‘private data’ that only Twitter is privy to, that Twitter is the only entity capable of making such determinations:

Effectively, what Agrawal is implying here is that *only* Twitter is in a position to grade their own homework/assess the authenticity of accounts/activity on their platform.

Researchers (myself included) have been pointing out for years that Twitter’s Trust and Safety Team has proven to be ineffective. We have done so without requiring the private data that Twitter speaks of, and while coming from a place where we do not have the same conflicting interests that plague Agrawal and the rest of his executive team (i.e. effective mitigation of platform manipulation linked to inauthentic accounts/activity adversely impacts Twitter revenue).

Advertising Fraud

In reply to Agrawal’s assertion that it’s impossible for outside parties to know which accounts Twitter is counting as mDAU (number of monetizable daily active Twitter users), Musk astutely pushes back by questioning ‘So how do advertisers know what they’re getting for their money?’:

The reality is that advertisers continue paying for advertising inventory created by inauthentic accounts, while Twitter laughs all the way to the bank, and while claiming that only they can assess the legitimacy of users on their platform!

As an example, below you’ll see Rolex advertising on the profile of an inauthentic QAnon focused account (full post here):

While the @QAJF_bot account has since been suspended, it took more than 25K tweets from the account for Twitter to suspend it (and after Twitter had already publicly made a strong commitment to ridding QAnon from their platform!):

OK, So What’s Your Point?

Simply stated, due to conflicting interests, we cannot allow Big Tech platforms (Twitter included!) to grade their own homework, opaquely determining the authenticity of accounts.