Hacker News Reader

But two comments with equal points given at different times on the same thread are likely not equally informative; the newer one had fewer impressions but receiving the same points has a higher points to views ratio. Part of the problem with points is that people try to use them like you are -- to gauge agreement for correctness, or who 'won' a discussion or whatever -- but points are a really crappy indicator of that. Points should be normed by impressions, and then smoothed (or given credibility intervals or something).

> Points should be normed by impressions, and then smoothed (or given credibility intervals or something).

I agree that this might work better for determining correctness of informative comments than just plain point scores, however point scores work better than nothing at all.

If someone states something that's incorrect, reply to the comment and explain why! ;)

But often there are multiple replies with opposing opinions. Without the points we don't know which opinion HN agrees with.

HN isn't an entity - and you can't rely on aggregation to form an opinion. You still need to think and form your own opinions.

>> Points should be normed by impressions, and then smoothed (or given credibility intervals or something).

While technically you're right, that seems like taking a really simple idea and making it really complicated.

Comment scores act as a "trend" of what people are thinking - it's not like a football game where the highest score wins. If I see a discussion it's useful to see a) how many people care enough to bother to vote either way and b) what sort of proportions are involved.

I think most people here can accept that not every comment in a thread gets the same number of eyeballs, and therefore the "scores" are not necessarily directly comparable.

And yet clearly people do compare scores.

The problem is that scores aren't like trends; they're like integrating a 'popularity' measure over time. As a consequence, you get a strong bias towards the stuff with the most points being comments from near the beginning of the discussion, not what's most correct or interesting or insightful.

Since I can't see your score, and you can't see mine, I honestly have no idea at this point whether more people agree with you than with me. Or indeed if no-one cares at all.

When this thread is referenced to a week from now there is no way to see what sort of agreement occurred with either one of us. There are a bunch of "for" comments, and a bunch of "against" comments, but ultimately no resolution (that either of us can see.) So a plethora of "me too" postings occur to fill the void.

Of course the agreement in, and of itself, means nothing - as noted elsewhere popularity does not equal correctness, however it's a data point which exists, and in some cases is useful.

> Since I can't see your score, and you can't see mine, I honestly have no idea at this point whether more people agree with you than with me... When this thread is referenced to a week from now there is no way to see what sort of agreement occurred with either one of us.

It's almost as though the future reader would need to form an opinion... for themselves.

But that doesn't negate the value in seeing which side of the debate was most widely agreed upon at the time.

Edit: of course, this argument would be satisfied by no more than making comment scores visible on discussions older than n days.

The readers here tend toward the more sophisticated end of the spectrum, and of course the presence of scores doesn't preclude future readings in depth. We all understand that scores are blunt indicators, but why not have access to that information? Knowing how the community "graded" a particular comment is an interesting piece of meta-information, even if you believe the cognitive anchoring effect diminishes its value as an indicator of the relative quality of the comment. But given the quality of the audience, high scores are likely to be positively correlated with quality comments, don't you think?

> high scores are likely to be positively correlated with quality comments, don't you think?

"likely ... positive", granted, but strong correlation, no.

Keep in mind the change was an experiment to mitigate effects of HN growth and the corresponding trend towards average of audience. The "tend" you mention had been unquestionably moving back to the middle.

That said, I don't have a problem with points switching on after the comments page goes to archive mode. I can ignore them then, while those who want to see them then can neither vote based on bias nor vote to correct perceived bias.

Discussion and reading should be thoughtful instead based on clicks

One's own comments can be self-appraised by seeing one's own score in real time, influencing one's own future thoughtfulness.

There are strong opinions and good arguments on both sides. Perhaps we should leave scores hidden, but provide a user preference to turn them on -- after reaching a certain karma level. If one argues that the rapid growth of the site has led to an exaggerated score influence, maybe a "norming" period makes sense. Downvoting privileges have to be "earned," so why not score visibility?

>"When this thread is referenced to a week from now there is no way to see what sort of agreement occurred with either one of us."

That just means that people will have to read the comments and decide for themselves based on the merits of the comments.

One might argue that that is better than deciding based on points because some people interpret points as assent to the beliefs expressed in the comment while others interpret points as a indicating the technical quality of the comment, e.g relevant supporting rationales and examples assembled logically.

But it's not necessarily about judging a comment for an upvote/downvote.

The number can be used to gauge whether you should stop and read something carefully. Maybe an expert comes along and makes an extremely informative comment and maybe I just scan it. Seeing a bunch of upvotes indicates that I should probably slow down and try to learn something from it.

Or maybe a well-reasoned debate has a large voting discrepancy. I value the collective judgement of the HN community and this information helps me learn from the arguments.

The value of this exceeds the more fair voting that we might have now.

>"The value of this exceeds the more fair voting that we might have now."

In your argument, you are using yourself as the measure of value, rather than using the community, i.e. saving you as an individual time by helping you ignore the less popular side of an argument doesn't demonstrably improve the quality of comments of the community in general (or even your comments in particular - though it may help you choose which side to comment on in order to maximize the upvotes your comments receive).

In the case where a "well-reasoned" debate has a large discrepancy then using points as a measure of value indicates an interpretation of votes as agreement with ideas rather than an objective indication of a comment's quality.

>> That just means that people will have to read the comments and decide for themselves based on the merits of the comments.

You imply that the sole benefit of the score would be to allow me to form my own opinion. Whereas I hypothosize that I already know my own opinion - the score allows me to measure the opinion of others. Those who have no opinion of their own are easily swayed, with or without a score, however for those with an opinion of their own, it is useful to know the relative popularity of that opinion.

More data is almost always better - especially for those who respect the value, and limitations, of raw data.

>"I hypothosize that I already know my own opinion"

I think part of the reason for hiding the points is to affect the process of forming opinions and better facilitate changing opinions already formed - I doubt that the use of scores to confirm one's own opinions is seen as beneficial to the community. For people who believe that scores should reflect the quality of a comment rather than the popularity of the sentiment it expresses your argument tends to support the hiding of scores.

Comment sparklines?

I agree. I actually often find myself only reading the top one or two comments on a thread now, whereas before I would scan down the list of comments for high point values. This is especially true of replies-in-replies, where the sort order is less indicative of agreement/value.

I miss the community-provided context of comment points.

You can see how agreeable a comment is by how far up the page it is.

I think having the scores hidden is a good decision. It means that we concentrate on content rather than a 'karma game'.

""that is not how mysql works" with 2 Points And "that is not how mysql works" with 102 points Are not the same piece of info."

In my opinion, both of those comments are bad comments unless they further qualify the statement. I don't think karma should be used to infer whether a statement is correct.

Does not work that way if a reply has 100 upvotes, does it?

Also I love the discussion in this post since, when with Georgify (http://news.ycombinator.com/item?id=2410195) I had tried first to seperate the upvotes with points of a comment, everyone (and yes a blanket term) had said it was a wrong thing to the extent that I had to make changes.

Of course it is not my site to make decisions for the users of the site, but still almost a dual reaction when the same decision was taken on HN by two different people in a very nearby time span, one being me (for Georgify) and the other being PG himself (for HN).

His opinion is that such a bad comment would never reach +100 upvotes since, the post being devoid of additional information, people like him might downvote if they saw the numbers getting "too high".

But a really good reply to the bad comment could get 100 upvotes yet be down, right? This is my interpretation of the rules here, not sure if they are correct.

lol... ok.. then it is agreed.. we should bring back comment score because the top comment is ----- "bring back comment score"... =)

Just expand the example: pretend it was "that is not the way mysql works blah blah explanation". I agree that I don't want to read this if it has 2 points but I do if it has 100.

Since we don't see points anymore, instead of upvotting you because I agree with you, I just have to add a comment saying "I agree". Is that what people want?

See kristofferR's comment above. He got downvoted for saying "I agree." This may have worked if implemented in the beginning of HN, but now the community has pretty strong norms against simple "I agree" or "I disagree" comments. Either eliminate voting completely or show the results, this in-between situation is foolish, bordering on stupid.

So many of the discussions around these points here center around the "TL;DR" point of view.

I wonder how these people ever managed to make it through a newspaper without scores telling them which articles are worth it.

Learn to survey, skim, and dive in. All your endeavors will be the better for it.

PS. Your method (don't read if don't like score) also ensures you won't be soiled by accidentally reading an unpopular but truthful point of view.

You need to reword this as a top level comment and repost. Some will say that having to make such a request is proof that scores need to come back, but I'm just concerned that everyone on the "pro" comment score side will never see it.

A better example would be choosing which newspapers to read based on your friend's recommendations, versus choosing by yourself. All articles from the same newspaper share a source and it already ensures some quality.

> All articles from the same newspaper share a source and it already ensures some quality

You make an excellent point, applicable here: newspaper brand = web brand.

Articles not flagged [dead] on news.ycombinator.com already ensure some quality, and both similarly share a "front page".

Wait, are you talking about articles on HN? Because only comments have their score hidden, you can still see how many points each article has.

My point was comparing newspapers brands/articles to HN comments, not articles, since you don't have much guidance anymore over what comments are worthwhile or not.

> Learn to survey, skim, and dive in. All your endeavors will be the better for it.

Then why sort comments at all? just order them in date order (or the reverse) and be done with it. Let the reader decide what is useful and what isn't.

Seems like having comments sorted in an implicit order of 'goodness' (even if there is no score) would handicap people in the development of their ability to 'survey, skim and dive in'.

To your point about not beings soiled by unpopular views, the truth is that unpopular views already have that font color applied to them which goes a lot further in making sure I don't ready it (with my eyes I often have to copy paste it to a text editor before I can read it comfortably). I would much prefer all comments be the same font color and being able to see the score so I can make my own decision as to whether to read it or not.

You can see how agreeable a comment is by how far up the page it is.

This comment is near the top of the page. Does that make it agreeable?

I don't think karma should be used to infer whether a statement is correct.

Completely agree. In my limited tenure here at HN, I've learned that karma scores have never been about being correct, factually or otherwise. Rather, they should be given to comments that help increase and move the discussion around the given topic even if that commenters opinion or assumption on something isn't factually accurate.

I'd be curious to see how this approach compares to the opposite, which strikes me as being more effective at reducing karma gaming. I.e., your own karma is invisible but you can see everyone else's.

You can see how agreeable a comment is by how far up the page it is.

Except if it's nested.

Sometimes the most useful comments are replies to less-interesting top-level comments. In this case the lower-scored item is positioned higher up the page than the higher-scored item.

Exactly! Right now it's impossible to tell if more people agree with you, or the person you are replying to. Frustrating.

Agree

Why aren't they the same piece of info? Scores just reflect popularity, not expert analysis. I find scoring systems like at reddit or slashdot just promote, usually wrong, group think (ZOMG OBAMA IS A COMMIE KENYAN!!) or whatever group values the majority accepts (GAYS SHOULDNT GET MARRIED!!).

There's a lot of faith in the wisdom of crowds with these systems and that's not a faith I find particularly convincing. The lowest common denominator shouldn't be encouraged. These systems are good for getting rid of trolls and other cruft, but not magical quality or truth machines.

Surely that's no different to what we have now: the most "popular" comments / submissions are, with the application of some secret sauce, the most visible and promoted to the top of the thread. The fact that we can't see the underlying numbers makes no difference.

I don't think you've been a member of the community long enough, then. On YC, I certainly find a correlation between the points for a certain comment and the thoughtfulness put into that comment. This isn't true on other systems (reddit, etc) where clever noise is preferred to depth, but I generally find the most helpful comments getting voted up here.

I agree. I've almost forgotten how disappointed I am that sites like HN, Reddit, and so forth don't use my up votes as signals for a collaborative filter and suggest which comments and stories I might find interesting. (And by "interesting," I mean interesting, not something I agree with, a distinction too many people on HN don't seem to understand.)

That's a great idea. I could see a simple implementation that'll could work out. If I like comment 4954 and you like comment 4954 and later I like comment 5004 then it should recommend it to you.

All comment scores do is reveal the inner workings of the site. HN already has an algorithm to turn vote count and time since posting into relevance, that's how we order posts now. Perhaps that could be a bit more fine grained, so that you can know more easily to skip barely upvoted leaf nodes, but all seeing the vote score does is let you replace that algorithm with your own internal one.

You are absolutely right though about source links on Google. They give us context for Google's ordering. While Google has done the work to return the best ordering of results possible, which might be the best results in the average case (though maybe not) site have earned a personal reputation for us. Personally, I know that I don't want to check the expertsexchange.com sites, so I skip over them.

I think that what HN would benefit greatly from is a more codified personal reputation system. Right now, the only way we have that is by learning to recognize the names of HN posters which, for me, has only happened with a precious few of the most prolific and respected. If HN would only give me some indication that I had up (or down) voted this commenter n number of times in the past, that would be far more useful than displaying the current vote count. It would allow me to discover whose voices I like, and who I might ignore.

I recognize that not having this is part of the design of HN. Instead we are encouraged to consider every comment fresh, without any context. It's a meritocracy of comments. I don't have much answer to that, other than to say that I would really like to see people be able to earn a good reputation. It would make the community feel a bit smaller and closer.

I'd venture in the opposite direction as an experiment. Make all posts temporarily anonymous until the thread is closed or less popular. Allow the discussion to exist ONLY according to the merits of the content. Then maybe after a thread is closed or enough time has passed, show the usernames and maybe the points as well.

Start with the assumption all comments are wrong; after all there's no absolute truth. The scores are a measure of popularity, which is often right and sometimes wrong. Plus everyone is entitled to an opinion. I don't think scores help anything, it's a measure of someones self-righteousness that i 'd rather not have. It's enough that unpopular opinions float to the bottom, where they belong.

If anything i'd suggest that comment scores should also have minimal impact in karma, except for questions, where an answer is a valuable contribution. Otherwise, everyone's entitled to an opinion.

Starting with an assumption that all comments are wrong is a prudent approach, which also considerably lowers the utility of HN for certain types of usage.

I read HN primarily for the informative comments, not the opinion pieces. A high comment score on an informative comment that contains a factual statement is a relatively good way how to easily tell that the information has been validated by the community and is likely to be correct.

>> It's enough that unpopular opinions float to the bottom, where they belong.

Unpopular opinions are exactly what make the conversation interesting. While not every unpopular opinion is necessarily valid, it's fair to say that on many occasions the "unpopular" viewpoint is both valid, and often correct.

But regardless of the correctness of the opinion, it is the unpopular viewpoint which makes us stop, think about, and then articulate our own viewpoint.

Now wouldn't it be ironical if this very comment is deemed unpopular?

Unpopular opinions more often than not are valid and uncomfortable truths, so they're likely to attract trolls and personal attacks. It's best to keep them down there, for those who are interested enough to browse all the threads.