This isn't entirely true [1]:
> As of January 2015, ISIS had reconstituted its regional accounts with strong privacy settings, allowing only a small group of known ISIS supporters to follow the accounts and read their tweets. The content of the tweets—primarily news releases, videos and photos from ISIS’s various provinces—are then disseminated by a number of other smaller accounts using hashtags. After the initial dissemination, the content is more widely distributed, but at significantly reduced levels from early 2014.
> Despite this, the number of new accounts created dropped significantly after the first round of suspensions in September (section 2.5.4), and while we do not have complete data to make a positive assessment, it appears the pace of account creation has lagged behind the pace of suspensions.
Another study about the impacts of Twitter's more aggressive stance against ISIS found, "These findings support the notion that Twitter’s policy deteriorates IS followers’ ability to gain traction on the platform, likely hindering their reach to potential recruits. " [2].
A third study also concluded, "Our data showed that the costs for most pro-IS users of engaging on Twitter (in terms of deflated morale, diffused messages and persistent effort needed to maintain a public presence) now largely outweigh the benefits." [3].
Deplatforming seems to work to reduce the spread of extremist ideas, which then reduces the spread of extremist actions, based on the years of research we've apparently been doing on the topic prior to today. We can reasonably disagree on the philosophy (and this appears to largely be a think piece), but there are practical realities that cannot be ignored.
[1] https://www.brookings.edu/wp-content/uploads/2016/06/isis_tw...
[2] https://extremism.gwu.edu/sites/g/files/zaxdzs2191/f/Digital...
[3] https://www.researchgate.net/publication/319395856_DISRUPTIN...
[4] https://threadreaderapp.com/thread/1348147867739975681.html <- where I found this info (there's more here I didn't add to this comment, all credit to @AmarAmarasingam for compiling this data)