How is ML not also an opaque algorithm? You can still bake bias in, especially if you train it on the type of content to upvote.

I wasn't trying to say ML makes an algorithm any less opaque - just that I don't like when such algorithms are used to populate content feeds.

I avoid content feeds that aren't driven by algorithms with simple, understandable rules, like the HN's voting system.

But I thought maybe there would be an application for using ML to help moderators more quickly detect posts and comments which warrent their attention. In that case, the "opaqueness" of the algorithm doesn't matter to me.

You might want to look at the current issues in StackOverflow - both sides of the dispute say that ML is not good enough to detect bad things.

Do you have a good link? I think it's an interesting subject, but all I can find are posts made by SO mods and admins which are generally positive about using bots to detect bad posts. If there is a controversy, I suppose SO would probably prefer it be hard to search for.

I've thought many times about ideas for platforms which involve user-submitted content, and I've always given up on them quickly because I couldn't think of a good way to handle moderation. It's probably just wishful thinking that ML could significantly help with that.