Hacker News new | comments | show | ask | jobs | submit login

Ok, comments over a threshold (currently 5) now have a red dot prepended to them. Is that better?

It should probably be scaled to the number of comments in the thread, or to the average score in the thread. Any static threshold will be easier to hit in the threads with higher traffic, and too many threads with a red dot adds to the noise.

The other side of that is that threads with very little traffic don't need as much meta-information, but there's an in-between.

> Is that better?

No, it's about the same.

It's not a problem with hiding/showing scores/hints, it is a fundamental problem with the voting model. As it stands, the process of voting is meant to re-balance the discussion tree to make it more interesting to an average reader.

IMO what makes more sense is to use votes to determine the preferences of an individual specific reader. If I keep voting up posts by a user X, then in all likelihood I will be more interesting in X's posts in the future. The same goes for downvotes. Additionally if X votes up Y's posts, then I may also be interested in Y's posts as well.

Lastly, the up/down voting score for a specific post is going to be equivalent to the score under existing model, and that can be used to sort posts for anonymous or uninitiated users.

PS. This is not a trivial change and it is a lot of work. I realize that.

Reinforcing the average makes good sense if your goal is to keep the community aligned, rather than simply please the widest array of readers.

Perhaps this is just me, but doesn't red seem counter-intuitive for a positive score? I think we're conditioned to think red is bad and green is good.

Could this dot be changed to green? Also, could you put an alt or label on that dot with something like "over 5 votes" or some other descriptive text? A bit more accessible.

I'd stay away from red and green and go with orange, to match the header color. These dots shouldn't map to any specific connotation; they're just visual indicators.

It's not every day you get a genuine "what color to paint the bike-shed" discussion :)

It's more common than you might think. It comes up every time you design a website, brochure, and many other products created everyday.

The dot adds visual noise (just as the redundant "by").

It might be worthwhile to try and map the color code onto the arrows. Red up arrow instead of red dot could serve both as highlight and as "stop, don't click me".

I really don't think that coloring the arrows is going to be a good idea. That will carry entirely too much meaning. People will think the red has something to do with the button and will be confused.

I also suspect that you haven't thought this through. Try mocking up a page full of red arrows instead of red dots. Even without doing so myself, I predict that because the arrows have twice the area of the dots you'll find that it looks far more cluttered.

Meanwhile, my bikeshedding alarm is flashing. ;)

I really, really like the basis of this idea.

But the more I ponder what you're trying to solve, the more I think you should provide scores weighted within the discussion. Perhaps, the quintile a comment is in. Use a bar that adds segments and shifts from one non-judgmental color to another. I wouldn't bother showing anything below a threshold of ten comments on a single link.

Otherwise, your piling-on problem returns, just with a little less fidelity.

Well, twelve hours ago or so I suggested the quintile idea. So I feel qualified to vehemently disagree with myself. ;)

On second thought, the problem with the quintile idea is that it presumes that every thread has an equal distribution of valuable contributions, middling contributions, and lousy contributions. But we don't want to rule out (e.g.) the possibility that an entire thread might be great and that every post deserves a dot.

Moreover, the quintile idea retains aspects of the piling-on problem: A sufficient quantity of super-upvoted comments will push other worthy comments out of the top quintiles.

Finally, quintiles are just way too complicated.

So I'm prepared to believe that I was wrong twelve hours ago. I like the threshold idea. Especially because the threshold can be tuned. If 5 isn't enough it can be turned up to 11 by a well-meaning moderator, acting at the appropriate time.

I'm surprised nobody suggested adding a mouse over to the orange dots. The first thing I did was mouse-over to look for a tooltip to explain the dot and got nothing.

So, now we know if a comment is at 0 or lower, and if a comment is at 5 or higher... Wouldn't it be easier and more consistent at that point to just show the score?

Although I guess I never saw the karma thing as a huge problem to begin with, so I suppose I'm a bad person to opine about this.

It's distracting. Can't the experiment not have any threshold indicators, run it like that for a couple of weeks, get some idea about how the comments are, and then add/subtract features?

People can check Lists, for interesting stories/comments. http://news.ycombinator.com/lists

The dot actually kind of hurts the eyes. . . among all that gray, those little dots are almost distracting. Something a little less loud perhaps that could blend into the overall visual scheme? Smaller dot; gray or silver?

How about none?

No, unfortunately it's really bad. I appreciate it as an example of how far afield you are willing to search for a solution, but it's visually distracting and difficult to interpret.

Leave the numbers all visible, or only show me the number on my own comment, but otherwise, stop messing around. This is distracting. I know there's a mob-voting mentality, but I actually read the comments, not the votes. Stop clubbing me over the head about the votes.

It's just an experiment. We should encourage pg to mess around all he wants -- it's his playground and we stand to benefit from experiments (and the downside isn't disastrous).

The purpose of upvotes here is what? Is it to decide where in a thread which comments will go? If that's the case, I see no reason for any visual cue as to number of upvotes. Their position means I will see them first. (Comments about weighted averages of comment threads used to determine position also work well.) If it's a tool to reward people for their insightfulness, why not let them show it off? Perhaps the problem is that the comment system is designed to do both, and one function is conflicting with the other.

Personally, I like seeing my karma go up when I submit useful comments or stories, so they work as a reward for me (and possibly everyone else here). However, I think the more important function of up/downvotes is bringing the 'best' discussions to the top and burying the garbage.

My dictatorial solution would be to hide comment count (and all visual indicators other than thread position), from everyone but the user they belong to. I would still get my 'reward', the mob-voting influence would be diminished, and comment quality would go up.

Is that better? No and please let us not get two dots for tens lest we start re-discovering variants of Chinese counting rods.

Seems like a lot of people have their own specific ideas to how they may want comments sorted / listed.

How hard would it be to allow people to implement their own version of the ranking function and then have the comments passed through that. Store the function in each person's profile, make them visible, and more importantly, make them share-able (like an app store?) so you can see which functions turn out to be most popular...

There's a giant can of worms there safety-wise (people running their own code on your server ?!?!) and unknown loading effects (nothing to stop people writing infinitely recursive loops...), although that said, it didn't stop you with viaweb did it?

Also haven't looked at the news.arc code recently so not entirely sure which part could be customised...

How hard would it be to allow people to implement their own version of the ranking function and then have the comments passed through that?

Harder than you think, I expect. You're proposing to do a custom query, of arbitrary, user-chosen complexity, for every single page view.

Moreover, the result will be of dubious merit. Individualized filter functions screw up conversation. That's what's wrong with conversation on the rest of the web -- people all read blogs, but nobody reads the same blogs, and people encounter the blogs in arbitrary order, and some blog entries are more out of date than others, and the result is like trying to carry on a conversation with a bright but slightly deaf time traveler who is probably a troll.

We have a distributed, personalized version of HN comments: They're called blogs and RSS feeds. The quality tends to be higher. Your personal filtering options are much more numerous and powerful. And they're just not the same thing. Otherwise I, for one, would be writing for my blog instead of posting on HN. Lord knows it would be better for my career.

To clarify my "how hard" sentiment - I had in mind that the normal sort function becomes a "function selector" that grabs the user's chosen sort function from their profile, then applies that to the comments.

The difficulty in all this would be policing the kind of code that can be written - perhaps a small DSL that compiles into arc code, or possibly direct arc code that gets vetted by a/some moderators.

The actual sort function seems to be (from arc3.tar, but may well have been changed here... nothing is safe from a REPL!) EDIT: indeed! see http://news.ycombinator.com/item?id=845932

  (def display-subcomments (c user whence (o indent 0))
    (each k (sort (compare > frontpage-rank:item) c!kids)
      (display-comment-tree (item k) user whence indent)))
so that in itself should not be too hard to modify, and provided it is not that much more complex, I can't see it putting that much more strain on the server. The sort function gets called for each page view, the user's profile (username / logout / threads) gets accessed for each view, so the only thing that needs to be resolved is the specific sort function.

The conversation threading could indeed become a bit fragmented, but then again, each time you look at a story on HN the comments will be in a different order due to the weighting and voting.

It seems that the dots were implemented rather quickly after removing scores due to user feedback.

I am wondering was there actually conclusive data (other than negative feedback, which might have been enough?) during when the score was removed and before the dot was implemented. As in, without a way to see the high scoring comments, did the quality of child comments go down? And if so what? Voting behavior changed?

Or was it an issue of being able to easily see popular comments? (And if it was the later, then maybe there are other solutions of visualizing highly ranked comments although likely more complicated, like a sort)

Better but in no way perfect. The problem is that the dots don't convey enough information to accurately skim a thread for good comments. In a thread with more than 200 comments a binary value simply isn't enough.

The problem could of course be solved by showing the points...

What do the grey dots mean? <Later>Hm. Grey dots vanished. This is better. I also saw a black dot. What about those?</Later>

And I really don't like the red dot. It doesn't fit into the flow of the layout and it distracts me from the content.

Tossing the vote count is an interesting experiment. The orange dot, hard to say. It was confusing to me and I'm not really sure it adds value. It may skew data out of the experiment of eliminating vote count.

I don't think it's clear what the red dot means. The issue with removing scores is I can't speed read only the "good" comments. Can you make the top n% of comments stand out somehow I wonder?

Why is it other people's job to decide what is and isn't good?

What do the grey dots mean?

The gray dots were something to turn red.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact