Hacker News Reader

The way it works on Perlmonks(.org) is that you don't see points on a comment/submission until after you've voted. If you want to see the score you can do a "null vote". I definitely think this is a better way to handle it. I really can't think of any reason why you'd want to show score before voting.

Yeah, that would work too. Didn't think of that, but that sorta gets the best of one world and a good part of the other. You still get to see the points eventually, but it doesn't influence the voting.

The reason that you would want to show the score before voting is that HN isn't all about voting and karma. If someone were to just want to read the best answers to a question, they might just scan by score for the highest ones.

Maybe there could be a manual-override for one page. But it would make you be conscious of the fact that you're going to bias yourself by doing it. It would be for people wanting to read but not really vote.

>I really can't think of any reason why you'd want to show score before voting.

It really depends on whether you see yourself as a Hacker News reader, or as a Hacker News editor. It's just like Amazon: If I'm a customer, I want to hear what everyone else thinks about something. But if I were hiring someone to be a professional reviewer for Amazon, I'd hide the other ratings to make them as objective as possible.

So here are two ways ratings make HN a more pleasant experience:

- If I have limited time to go through HN, I want to be able to zoom in on the comments others judged to be of the highest quality.

- If the comment is on a topic I have little familiarity with, it's often hard to judge its accuracy/validity on my own. A decent number of votes gives me confidence that the person is right.

So in these respects, it's just like Google rankings or Amazon ratings.

Having points probably does cause some amount of groupthink. I bet hiding votes would scrunch the distribution of votes around the mean.

By the way, if HN's database contains the exact sequence of upmods/downmods for a particular story, e.g. 1, 1, -1, -1, -1, 1, ..., it would be possible to test how much groupthink is involved by modeling the sequence of votes as a random walk and testing how much the probability of an upmod/downmod is conditional on the current number of votes. Or, if HN starts hiding votes, you could analyze how this affects voting patterns. Sounds like an interesting college stats/psychology project for someone :)

few things wrong with that: a) HN presorts comments more or less by karma rating, you pretty much know where everything will wind up before voting. b) it also changes the color when you start getting negative points.

So sure you may not know that something is at 1 or 500 points, but you you sure can guess

>> I really can't think of any reason why you'd want to show score before voting.

When I vote (here), I'm pushing a comment towards an idealized score. To make the decisions "do I think this is rated too highly" and "do I think this is rated too lowly" requires knowledge of the current score.

This closely matches the way I mod comments as well.

I'd much rather deal with the ramifications of "groupthink" than wade through dozens of banal comments.

Vote counts give me more information. While that may affect my understanding of an idea, it's for the better.

Points may have an inherent bias towards older posts or well known authors, but I value what they represent.

But part of what groupthink does is make the points mean less. The opinions of a few can start a groupthink snowball burying good comments and raising mediocre comments to high scores.

So as of now, that information you're using is unreliable because it is affected by groupthink.

Which is exactly what happened on Reddit, where you have to wade through all the up-modded pun threads to get to the real comments.

Completely agree, and said as much not long ago:

http://news.ycombinator.com/item?id=447563

I tend to find myself influenced by the karma count more than anything else. I have to log myself out after commenting to increase the cost of checking the change to my net karma.

I'm a strong advocate of hiding the karma counts entirely (for users, submissions, and comments). That greatly reduces the conditioning effect, not to mention karma-related drama and groupthink (as Surowiecki argues, crowd wisdom is best procured when evaluations are performed independently). Submissions and comments can still be ordered by net karma over time, as they are now.

Whatever you choose to do, I think there's a strong argument in favor of experimentalism. Especially since the site isn't a commercial undertaking and can deal with the occasional disruption.

(Um, and yes, that user is me. I locked myself out of my account to keep myself from compulsively commenting and visiting the site. Apparently I still wanted back in.)

You are not good at locking yourself out of your account. While I appreciate the struggle with compulsion, I'm not sure you're fit to dictate guidelines.

Why do you think he is not good at locking himself out of his account? (He is now posting from a new account)

In what way do you think he is "dictating" guidelines?

Sometimes I'll just upvote a submission on the 'new' page with 0 points to trigger group think because I might think the article will spark interesting debate, or to help prevent an 'ask hn' or 'rate my startup' submission from falling, which keeps other users here from seeking advice from the community.

Seeing the scores doesn't affect my opinion! I'm impartial!

... except that a few times, I've observed my attitude towards a comment change when its score changed. What is a little frightening is I have also witnessed a struggle within me between what seems to be true to me, and fitting in with the group's opinion.

This psychological phenomenon has been confirmed experimentally (though I have met one extremely detached person (borderline autistic?) who I would bet would be above it). http://en.wikipedia.org/wiki/Asch_conformity_experiments

A struggle for perception of reality is a struggle for reality.

I would think that people should be able to handle the truth (either way) - painful as it may be. I mean if the Internet stands for one thing, it's transparency. Hiding stuff is not something we should strive for.

I’m not sure adherence to simplistic ideals is something we should strive for either. The goal of HN, as I understand it, is to generate thoughtful discussion. If transparency is counterproductive to achieving that goal than we should abandon it.

As far as "handling the truth" I don’t think that speaks to his point. When we use that phrase we are talking about someone controlling their reaction to a change they can perceive. Such as someone preventing themselves from emotionally falling apart when something bad happens.

But group think, and the desire for acceptance which is its root cause, is something people don’t perceive is happening to them. They therefore can’t consciously control their reaction to it.

His suggestion is that we remove the root stimuli so as to not provoke the reaction to begin with. So your point doesn’t really apply.

The points provide information about what others think about the submission, which is useful in guaging the status of the comment. It does not have to affect your opinion, unless you let it.

I think people are much less in control of what influences them than what we'd like to admit.

I can confirm this. Halfway through reading his post, I noticed that you had more points than him, stopped to read yours, then read his with skepticism. Possibly it's because you do make a legitimate point, but I'd suspect that your score disrupted the flow of my reading.

Read Influence: The Psychology of Persuasion by Robert Cialdini. Anchoring, framing, and sex are very powerful.

I think this is a very good idea, but for a slightly different reason. The whole points-for-posts thing creates a meta-game of accumulating points above the normal game of participating in discussion. This leads to those annoying meta-disucssions where people complain about being down modded and all that goes with that.

The modding system is useful for ranking comments so that when I read a massive thread the most interesting ones tend to be around the top.

It may also be useful for users as a feedback mechanism to figure out when they're doing something that's not cool (like the things we've been discussing in the guidelines thread today) but I don't think you need a precise numerical score for that. And I think it's the precise score that gets people in the karma-game mindset. It certainly does that for me.

I'm honestly not being cheeky, I suppose I'm just slow: Wouldn't it be easier to just not think about the points? I've found the points useful when looking for a summary, takeaway, or rebuttal that earned attention.

Theoretically, that's the optimal solution. They have advantages for situations like what you listed.

The problem is that pretty much everyone is in some way influenced by them just by seeing them.

Do you really approach a comment at -10 points with the same open mind as one with +30?

I think the more important question is should you? If 10 other people before have expended the mental energy to figure out that this comment isn't worth reading, why should I expend mine to determine the same thing? I like where you're going with this, but it uses our group mental capacity much less efficiently. There is probably a better solution... looking at a number is a pretty efficient filter and I don't think we should lose it.

In most cases you’re probably right. But not always.

The problem is you don’t know the moral character or intelligence level of the people who first stumbled on to a comment. If two ignorant people happen to see a quality comment first and both vote it down that comment is essentially lost to the group. I think we can all agree there are at least a few fools running around here so the idea isn’t farfetched.

So the question is do you want to trust what might be insightful comments to the randomness of timing?

But 10 people's moral character? 30?

If the comment was really sparkling, it seems unlikely that the first two ignoramuses would drown out the larger cast of HN. It's not all about timing.

But the whole premise is that once a comment gets a negative score people don't read it anymore. 2 or 3 people could easily do that.

Also, 10 people on a site that gets in the area of 200,000 uniques isn't that significant.

What if we don't assume every vote is equal. What about weighting a user's vote by some factor of their karma? And/or capping the threshold that a user can vote. i.e A user with 0 karma can vote on comments with 0-3 points. A person with 100 karma can vote on comments -3-6 points. A person with 500 karma can vote on comments -9-9 points, etc... This implies that a comment with ridiculous votes (either positive or negative) has been decided by members of the community that the community has already shown approval for.

I suspect that comments that have 100 upvotes and 100 down votes are far more interesting than comments that haven't been voted on at all, but have no way to know....

Yes, I wish points were displayed as:

upmods - downmods = score

Points have their uses in making it easier to zoom in on the important parts of a discussion (or the best contributions), it's helpful if you have a limited amount of time to spend and you want to make the most of that.

Unfortunately they tend to turn in to a goal by themselves.

Points are the root of all evil ;)

It's a classic case of influencing the system that you are measuring, same with google pagerank or a volt meter with a low internal resistance.

I'm not so sure. How do you know that groupthink is the problem? Perhaps the problem is instead that the new users fit in a different groupthink pattern than the old users. In that case, you want to encourage groupthink when it leads to the correct behaviours.

Maybe we could do an experiment with randomly upvoted/downvoted new posts and see how much more likely the posts are to be upvoted/downvoted?

And what about hiding karma points too ? Imo, the problem with karma is that people try to get more karma, per se.

It serves a purpose, though. When you see a user with high karma, you can easily tell that he has contributed a lot to the conversation.

This is important, because it helps defend the site against trolls.

But perhaps it would be better not to be confronted with your score every time you log in.

Good idea, that could be an easy way to prevent the absurd up- or down-voting phenomenon.

Ask HN: Don't show points?