Social media is truly social in the sense that it features incredible pressures to form in-groups and out-groups and then to conform to your in-group. Unless you like and admire Cotton and Pompeo and want to be known to the world as a follower of Cotton-Pompeo Thought, it is not very compelling to speak up in favor of a minority viewpoint among scientists. Why spend your day in nasty fights on Twitter when you could be doing science? Then if you secure your impression of what "the scientists" think about something from scanning Twitter, you will perceive a consensus that is not really there. If something is a 70-30 issue but the 30 are keeping their heads down, it can look like a 98-2 issue.

I do not know a lot about science, so I will not opine how generally true this may or may not be.

But in economics, which I do know well, I think it's a big issue. If someone tweets something you agree with, it is easy to bless it with an RT or a little heart. To take issue with it is to start a fight. And conversely, it's much more pleasant to do a tweet that is greeted with lots of RTs and little hearts rather than one that starts fights. So I know from talking to econ PhD-havers that almost everyone is disproportionately avoiding statements they believe to be locally unpopular in their community. There is just more disagreement and dissension than you would know unless you took the time to reach out to people and speak to them in a more relaxed way.

My strong suspicion is that this is true across domains of expertise, and is creating a lot of bubbles of fake consensus that can become very misleading. And I don't have a solution.

Here is his full Substack post, I am very happy to be a paid subscriber.  The broader question of course is what we can do to limit these problems.  More pseudonymous tweeters and writers?  More grumpy old people who don't care so much about their reputations?  More who write for Substack?  Other?