So, funny thing, but I was just going through dang and sctb's comment histories to get a stronger sense of what makes HN tick. Despite (or because of) limited features, with strong guidance from the top (Guidelines, Paul Graham's "How to Disagree"), and a community that's evolved over the course of about a decade, the discourse here is pretty good. Not perfect, but the curation and commentary are usually informative, occasionally excellent. Topical breadth is limited, and there's no categorisation of posts, but what is covered is pretty good.
I'm contrasting that with Imzy, advertised as a "kinder, gentler Reddit", but which in my experience has been among the more vile, toxic, abrasive, and hateful hellholes on the Internet. I'm not a connoiseur of such spaces -- /b/, 4chan, and the like -- but I've seen what does work.
This set of guidelines from Google+ (unofficial AFAIU) are also useful. Communities with similar tools and principles tend to do better:
https://plus.google.com/+JohnSkeats/posts/F86Hv3L6kE3
In a series of investigations trying to assess the size, scope, and vitality of discussion at G+, I hit on the model of searching for names of significant authors or other authorities as a proxy for meaningful discussion. Foreign Policy has a list of "Global 100 Thinkers" published annually which provides a pretty broad-based, multi-ideology, and current basis for finding same, and a set of Google searches (looking at site match totals) along with an arbitrarily selected string to indicate potential conversations not of interest ("Kim Kardashian"), gives an FP:KK index.
https://www.reddit.com/r/dredmorbius/comments/3hp41w/trackin...
As earlier comments have noted, Metafilter has a staggeringly high FP:KK index, though low overall traffic. Reddit is also quite high, and blogs, most particularly Wordpress.com, perform quite well.
This is the sort of tracking I'd like to do over time and across more sites -- it would be particularly useful to have a web search which would simply return the top n domains and hit counts rather than having to parse results and insert search delays (about 45-60s seems to be sufficient) to avoid Captcha prompts.
Conversely, you could find some good search tokens for topics of interest and see where results tend to pop up.
Looking at the detailed results (which I've not posted, the tables simply become too large for convenient presentation), some pretty clear patterns of distribution across both sites and identifiers turn up. Richard Dawkins and Pope Benedict, for example, were both highly popular, though on different sites.
There's a huge amount of public Web activity at several private educational institutions, in particular Harvard, MIT, Stanford, Cornell University, University of Michigan, and Johns Hopkins. By comparison, the University of California system, which I'd have expected to be more significantly represented, is tiny. European educational institutions are also sadly anaemic, with largely UK schools represented, also Maastricht and Unibo (Italy).
I did a follow-up search through major Libertarian organisations which was pretty disappointing. Having discovered virtually all are part of the Atlas Network might improve my search space a bit, and might make an interesting follow-up.
I've also become fairly convinced that good conversation scales poorly. I'm experimenting with a set of smallish groups, probably ~30 - 300 people, though that remains to be seen via empirical results. These are "knitting societies" in an effort to both be un-cool ("cool" and "good discussion" seem inversely related), and in the sense that these might be societies that knit together, both internally and externally. We're somewhat looking for a good home. (Imzy had been on the candidate list. It's no longer.)
The other option is to seek out the specific people you're interested in discussing things with. I've been taking to simply emailing people directly with either questions or information I'm hoping will be of interest. I got a response yesterday (after a week or two) from a Famous Authority in an Area of Interest who not only appreciated the obscure detain and evidence I'd turned up but invited more. Not bad for a space alien cat / 1950s sci-fi character.
The signal's out there, but you've got to hunt for it, and it is very easily scared off.