HN works very poorly, and worse by the year. I've been using HN since 2009, I've been on a wide variety of discussion platforms going back to usenet, HN has had its moment in the sun and that has largely passed.
Voting on HN barely has an effect, and I suspect that the average votes per comment on HN has gone way down year over year. People just don't vote on posts as often as you'd think, not anymore. A related problem is that commentary doesn't go on for very long. In the usenet days you could have a good thread that would last for months and months that would continue to spawn good and interesting commentary, a flash in the pan thread might only last a few days. On HN the window of commentary for a post is rarely more than a day and typically only a matter of hours. It's just people strafing comments into the void and then disengaging. Long comments typically don't get read, and don't get upvoted, don't get commented on, etc, for example.
> ...I suspect that the average votes per comment on HN has gone way down year over year
OK, I'll bite! A cursory look at the data shows a clear increase in average votes on comments from 2007 until 2012, which is the only year with a dip, followed by steady growth until the present all-time high.
I wonder if (or how) one should take population into account, too. We might figure that a majority of the people read only threads that are on the frontpage, so, if the number of people on HN has doubled, then each thread will get viewed by 2x as many people, and, if the new crowd has the same likelihood of voting on each comment as the old crowd, then you'd expect 2x as many votes, assuming no change in comment quality. Instead you might want "votes per comment, divided by number of users".
Of course, there are lots of "all else being equal" implicit assumptions there. First, if the population doubles but stories move off the frontpage in 0.7x the time, then you'd only get 1.4x as many votes—and this is one of InclinedPlane's points. Second, the newer crowd could be significantly more, or significantly less, active. To control for these two things, the measure you might use instead is "votes per comment per pageview", or "votes per comment per second a user spends on the page". Third, there might be more comments posted—well, duh, it would be weird if the new users never posted any comments.
Fourth—and I think this another thing InclinedPlane wants to focus on—comment quality could have changed. Comment sorting is relevant too, because I'm sure lots of users don't read everything. If we suppose that, due to an increase in population, we get 2x as many comments but they have the same quality distribution, and if we suppose the best comments always go to the top, then the average quality of the top n comments should increase; you can see something like this in extremely popular Reddit threads, where the top several highly upvoted comments are clearly optimized for something (often clever jokes). If we suppose a decent population of users only read the top n comments, and always use the same function that maps "quality of a comment" to "probability of upvoting", then, when the set of comments doubles and (by assumption) the best rise to the top, we'd expect these users to generate more upvotes overall, and hence "average votes per comment viewed" should go up. (It's also possible that people's standards would rise. But I think people's changing standards would lag behind the changes in what they're viewing.) That said, for the comments that aren't in the top n, the fraction of people that view them (and consequently might comment on them) would go down.
The question of how long threads sit on the frontpage is relevant, both for comment exposure and for InclinedPlane's point about conversation longevity. (There are also pages like "new" and the no-longer-linked-at-the-top "best".) I wonder how best to quantify that... perhaps "the frontpage tenure of the thread with the longest tenure of all threads on that day".
Voting on HN barely has an effect, and I suspect that the average votes per comment on HN has gone way down year over year. People just don't vote on posts as often as you'd think, not anymore. A related problem is that commentary doesn't go on for very long. In the usenet days you could have a good thread that would last for months and months that would continue to spawn good and interesting commentary, a flash in the pan thread might only last a few days. On HN the window of commentary for a post is rarely more than a day and typically only a matter of hours. It's just people strafing comments into the void and then disengaging. Long comments typically don't get read, and don't get upvoted, don't get commented on, etc, for example.