Why I think is hard it's because multiple rules can be made to make it impossible to spread ideas: talking loudly in the street => you disturb the neighbors; you send mails with pamphlets => it's spam; want to make an add on TV => extremely expensive. And so on.
Most of them have a vision for their platform, i.e. town hall for Twitter, family and friend conversations for Facebook etc.
To adhere to this image they filter out spam etc. now, filtering out obnoxious content just becomes one more rule and thus the slippery slope begins.
But to answer your question: for me, any kind of interference such as deleting/hiding content or algorithmically influence which content is shown is censorship on social media platforms, and the user should be responsible for applying such censorship.
I.e. provide a UI which let the user configure their own preferences. But actually nailing such a feature with a good UX ain't easy, and how to actually implement it isn't either, so that's just a pipedream, realistically speaking.
Maybe the network should also limit interaction and exposure. It's fine if you get more interaction than you could do in real life, but I find worrying to have one person followed by tens of millions ... (and even if it was the case before with newspapers, I don't think it was ideal either)