Earlier quoted context omitted.
The main thing I have in mind is machine learning algorithms that optimize for engagement. Those aren't necessarily biased in favor of a specific political position, but tend to amplify rumors over well-sourced reporting, demagoguery over reasoned debate, and often malicious false claims. Off the top of my head, I don't have a good way to differentiate those algorithms in legal terms. As another comment points out, e…
OK, that's at least a more reasonable thing to be worried about [0], but as you say trying to use the law there would be damn near impossible. Take the case of a law against airplanes/cats again. I'd definitely feel very, very strongly about such an effort, and want very, very much to defeat it. In the democratic system that means rallying a critical mass of fellow citizens. If it so happens my airplane/cat platform…
Now what if I hand-curate a front page? If I feature this libelous rumor on it, I'm acting like an editor/publisher rather than a platform and the chances you can successfully sue me go up. What if I pick the users who have been most responsive to that sort of content and email it to them in the hope they'll visit my site and spend more time there as a result? I see my legal risk increasing even more.
Of course it's easy to hold a person responsible when a person is making the decisions. It's harder to say exactly what criteria an algorithm can use, and not the sort of thing I'd want a politician or bureaucrat deciding. Perhaps what I really want is for people to voluntarily stop using corporate social media so much, but I don't know how realistic that is.
[0] Other technologies are better-suited to regulation; I think I wouldn't want to fly on an airplane that got only the maintenance the market demanded.