Does Deplatforming Collapse the Cost of Misinformation?
Recently, Twitter has taken the unprecedented step of deplatforming users en masse largely in response to the attack on the Capitol last week. We at IPM have long monitored the cost of maintaining manipulation campaigns on popular social media sites. As they have grown in prominence, and in reaction to, deplatforming or threats of deplatforming, we’ve increasingly turned our vulnerability engine towards offshoot social media platforms such as Gab, Parler, TheDonald.win, and so forth. In our investigations, we have found an issue that deserves a bit more attention - deplatforming collapses the cost of establishing and maintaining misinformation campaigns.
As we’ve seen in multiple articles covering the practice, it’s clear that Deplatforming works. The “works” here refers to the ability to combat and minimize extremist behavior - refusing to give it oxygen helps create a less toxic Internet. The problem with deplatforming that’s not being discussed however, is that when users congregate on smaller, niche sites after being deplatformed, they typically congregate on services that have much less security in place, and thus, are much more exposed to coordinated manipulation efforts.
As an example, we identified the rough cost for establishing and maintaining an account on two platforms affected by the recent deplatforming efforts, Twitter and Parler. Twitter’s signup flow requires 27 distinct touch-points - Parler requires 5. While creating an account on Twitter, users must verify access to an external account by inserting a verification code into the signup form that is retrieved by email or by text message. Parler does not do any verification of email address, and recently stopped requiring phone number verification. As a result, the relative cost of establishing an account is much cheaper on Parler as compared to Twitter. Further, Twitter runs active sweeps of the platform to identify suspicious accounts post signup - we estimate that as many as 85% of programmatically generated accounts are ultimately identified and suspended or temporarily restricted within the first 90 days of account creation. To the extent that these processes are occurring on Parler, we have not found any evidence of them.
Further, as more “infrastructure” companies such as CDNs, cloud providers, and identity verification services refuse to work with sites like Parler, the capacity for these platforms to maintain best practices while keeping those features of their system online is greatly diminished - it’s not unrealistic to imagine this to be a reason that phone number verification could be ultimately removed, for example. As these services withdraw from platforms like Parler, the cost of ongoing attacks is even further reduced. At the time of writing, we determined that the cost of operating a single account on Parler is orders of magnitude less as compared to Twitter - we expect that politically motivated actors have taken note of this fact as well.
