You can ban speech based on non-content criteria such as the volume of sound, but when you ban speech based on the content, it is a violation of the principle of free speech.
Whether private companies should not be able to discriminate speech based on content is still up to debate, but don’t compare this to shouting.
> when you ban speech based on the content, it is a violation of the principle of free speech
This is the sealion's argument, that you have to be willing to engage with them at all times in all contexts. It implicitly denies freedom of association. For example, if I am running doomermetalchats.com do I have to tolerate K-pop fanatics with their relentless positivity and sunny optimism?
Run a decentralized system akin to a modernized mailing list with a nice UI, where every person accessing the list gets to pick how their view of the list is filtered, choosing from competing filters or creating their own. Filters could be machine-based, competing human-moderator based, whatever.
The doomers and the stans can be talking right alongside each other and never see each other; effectively, the doomer 'site' only exists virtually, as a mutually visible consensus among a subset of the network.
This has been tried repeatedly, and it turns out nobody likes it. People either self-segregate into echo chambers following very simple patterns of assortativity, or complain about being 'shadowbanned' if they can't reach the audience they want to.
There are a few different motivations for social communication, both cooperative and antagonistic. People often want to be able to communicate freely with their peers and exclude non-members of their peer group (socially, algorithmically, or by fiat), but sometimes also to be able to push into and perhaps take over network territory of rival groups or that they consider should be common.
Whether private companies should not be able to discriminate speech based on content is still up to debate, but don’t compare this to shouting.