[buddapalm]

> The problem is that the abstractions for working out what gets attention/promotion on social media favour the lowest common denominator.

IMHO you've identified the root problem. it's simply very easy to optimize a feed around quick interaction signals as a way of surfacing what's important, and still get the quality wrong.

> we are effectively trying to automate an editorial process for deciding what is currently the most important thing in the world, in the meantime separating what is poor content from what is merely controversial, and ranking an image macro against a 10,000 word essay. This is a herculean task, and the methods we are using are manifestly not up to the job.

what's needed IMO are user satisfaction signals. I wonder if social networks could invite users to report content that makes them mad or they find useless. not for abuse, simply to identify what different users like and start understanding what groups of people prefer