One of the most noticeable things is trying to skim a discussion with thousands of comments. On Slashdot, you can set your threshold to +5, and see only the very most insightful comments, no matter where they are in the thread. On HN, there's a lot more digging required to get the gems.
That said, I do think HN's "story-selection" algorithm is an improvement. What shows up on the front page relies a lot less on the quirks of the editors, and seems to result in a much more diverse set of posts.
As others have mentioned, I think the top-level moderation is the most important factor keeping HN worth reading. It's dang intervening and setting the tone which does the most for HN at the moment. We need to keep him happy; and I hope that he has a "succession" in place in case he's ever unable to continue the work.
How hard is it to ban the same post being made repeatedly, over and over?
Slashdot stopped trying.
They grew up. They kept their opinions and attitudes, as people do. You can grow out of graffiti, but you will still like the style. Getting older does not magically make you stop liking nazi if you liked them as teenager.
Instead, you seek to pass your values to teenagers that follow you.
All the more so when they are used to, in effect, remove comments by making downvoted comments more difficult to read.
Now, I can understand the idea of making it easier to navigate through many comments by crowd-sourcing a metric of each comment's 'value', but in practice this is extremely difficult to do while preserving the diversity of views. It usually ends up being the "tyranny of the majority".
Same with 9gag btw. It was funny, then funny with occasional eye roll and then "I dont have to look at this". I dont know whether it changed back or not.
I think the majority of the enjoyable community moved on a long time ago.
For example, if you just mention you’re a happy Microsoft user, you’re accused (often in a racist way) of being a schill and the cause of everything wrong in the world.
That particular aspect of the moderation system sounds good on paper, but was rather useless in practice.
Essentially, what "matters" is that people are upvoting or downvoting a particular post. Theoretically, it's good to have a bit of additional metadata telling us why they're doing it: i.e., do they find the post Funny, Insightful, or Informative?
In practicality, it's not that useful, or at least it wasn't put to any practical use by Slashdot when I stopped reading it back in the 2000's. At a moment's glance, it's pretty trivial to see why people might've upvoted a particular post - it's clearly meant to be funny, or it has a lot of informative links, etc. So the metadata of "why" was just redundant.
On the other hand, I suppose the metadata could have been put to some practical use after all. For example, it would have been perhaps useful to filter out all of the "Funny" [1] and "Insightful" [2] replies on Slashdot and to focus on the most "Informative" [3] ones. I don't remember the interface allowing you to do that. Perhaps they added it later, or perhaps it's something you might explore in your own system.
------
[1] I mean, one can only read so many "Imagine a Beowulf cluster of THOSE!" jokes before one wants to jump off of a bridge.
[2] In my experience, 99% of the "Insightful" comments were just those that forcefully/eloquently regurgitated FOSS talking points that resonated well with established Slashdot groupthink. "Insightful" upvotes tended to mean "I agree with the author of this comment" and not "wow, actual new insights are presented here." Perhaps this might have been avoided if there was a separate "I agree with this" upvote method, distinct from "Insightful."
[3] It was easy to farm for karma this way; if you stumbled upon a recent Slashdot article without many replies, you could throw a few easily-Googled relevant links into a reply and be assured of getting a bunch of "Informative" votes. Nonetheless, while imperfect, the replies deemed "Informative" had a decent signal-to-noise ratio relative to the rest of the replies.
The interface did indeed allow you to do this. At least, I remember it being there in about 2001.
Additionally, the metadata on voting was particularly important for meta-moderation which other platforms don't really have. Users were (at random) shown a comment that you've voted on and your vote. They then said whether your vote was fair or not, based on what they see.
Stepping back a bit - the most important difference between the Slashdot system (vs something like Reddit) was that votes were much more limited on Slashdot. Not everyone had modpoints (votes) and you only got a handful at a time. This allowed meta-moderation to actually have some impact - if you continuously abused modpoints by downvoting someone as troll just because you disagreed with them then you wouldn't get banned - just you'd get less votes in the future. The system was a little opaque and I've not used the site in years but that was my understanding back then, it may have changed since then.
While it's a bit complex I think the fundamental idea is very good. Ok making a distinction between informative vs insightful never seemed that useful to me, but certainly allowing people to hide "Funny" content definitely has value because funny one liners tend to crowd out other content by sheer volume.
It's certainly very interesting to talk about, at least for me!
Starting with the results and working backwards, I didn't find the results of moderation+metamoderation on Slashdot to be superior to subsequent sites like Reddit and HN that skipped the metamoderation portion.
Why? I think metamoderation on Slashdot was ultimately redundant. The metamoderators were also the moderators. So your moderation was being moderated by the moderators that moderated the posts. The metamoderation was not really of a higher quality than the moderation, so it didn't really add much.
Metamoderation was good at flagging grossly incompetent moderation, such as somebody seeing an ASCII swastika and upvoting it. However, as long as the community as whole is relatively healthy, that sort of poor moderation would be rendered moot anyway; such a post on Reddit would receive perhaps 98% downvotes and 2% upvotes, it's not clear to me that preventing that tiny 2% minority from voting in the future would be demonstrably better than simply drowning them out which is what happens by default anyway.
Metamoderation could perhaps be useful if the metamoderators were guaranteed to be trusted, high-quality users.
Slashdot tried to do something like that; IIRC you got the opportunity to meta-mod after you were up-modded yourself, so theoretically "better" posters were doing the meta-modding, but in practice I think this just added an extra layer of echoes to the echo chamber.
You're right that the pool of moderators + meta-moderators was the same, but I think there's value in a kind of "peer review" for voting. On Reddit almost everyone (who expresses an opinion) complains that people use downvotes to disagree but I suspect that even the people complaining about it do the same thing from time to time (even if they rationalise it). Sometimes it's helpful for someone not involved in the argument to judge what you've done, even if that person is quite like you. It's pretty easy to imagine excluding anyone who replied to that same thread from meta moderating (or maybe doing something more sophisticated than that). It's not really about the trolls upvoting ASCII swastikas but the people regularly downvoting otherwise OK posts because they dislike someone's argument or their name or what they said in a completely different conversation. Or maybe it's about people upvoting low effort content (but for that to work, you would have different people metamoderating like you say).
It's hard to judge based on outcomes but I always felt Slashdot made you think about voting in a slightly different way. The system as it was there wouldn't work with Reddit for a variety of reasons but it's still useful to think about.
As such you could get +20 flamebait, for a popular or welcome flamebait post,
or -10 funny, and so on - the tags where not inherently good or bad.