It sounds like the problem implicit in your framing then is "algorithms", not extremists--a stance I agree with--which would imply the disinformation isn't coming from big tech's users, it is caused by big tech itself, and maybe it is the algorithms that should be limited and regulated instead of the discussions.
I doubt this, because it was possible to 'go viral' to a limited degree way back in the days of Usenet, with any algorithmic selection or amplification going on at all. preferential attachment is a social phenomenon which businesses leverage for profit (not just int he era of social media, but since the establishment of brands and before that, the awareness by politicians and theatrical promoters of how compelling parasocial relationships can be). Algorithmic manipulation is a problem, but you could switch it all off in the morning and you would still have the same issues. Any system can and will be gamed.
