In vague terms, the idea is to suffer less from 'how large group dynamics on the internet work (by default)' while still having enough eyeballs per front-page submission to have a discussion.
Now how would we operationalise that? A simple measurement is to check whether engagement per submission is having a longer, fatter tail. But that would be merely something that's easy to measure, not something we directly care about.
You'd need to have some proxies for drawbacks of 'large group dynamics on the internet'. Perhaps check civility of discussion or so?
> [...] but mostly people got pissed off that they were seeing random stuff on the front page.
I guess if you'd want to check again, you'd either have to educate people better (ie better PR) or you'd have to be more sneaky.
An idea for the latter: instead of restricting users to 10% of submissions, as a test run you can reduce them to 80% of submissions. That way the front-page would still look pretty similar to before and you wouldn't drive people too far into the long tail of submissions. Of course, any effect you could measure would also be weaker.
What did you measure (or hope to measure) when you ran this experiment a few years ago?