For the telephone network, control does not scale. There is no way for the phone company to censor what happens in one billion daily phone calls. It can't be done (until we get adequate speech recognition, but that's not an answer that I want to see applied in this way).
Facebook, YouTube, and Twitter are like the phone company, not like newspapers and TV. Censorship does not scale.
Twitter, being text, could think about trying to automate censorship. The problem is that people do all kinds of typos, circumlocutions, innuendo, and dog whistles to get around such things. Words are really flexible, so it's a hard problem.
Facebook and YouTube are worse. They can try to have automatic assists, but pretty much the only way it can be done is via user reporting. Even that doesn't scale all that well.
A case can be made to protect someone who allows all comers, who increases the reach of the public to communicate with each other, and who extends the scope of public discourse to be protected from the legal consequences of what their users say. The same case is a lot harder to apply to an institution that curates the content they choose to publish. By way of comparison, telephone carriers, an industry that exerts no editorial control, are largely not legally liable for what customers say over the phone network. Newspaper publishers, who have absolute editorial control, on the other hand are legally responsible for what they publish.
Which of these are Facebook and Twitter? My answer is that the more they limit speech that is not legally criminal then the more liability they should have for what they allow.