Hacker News Reader

Comment by ssp | Hacker News Reader

A variant of this is to publicly display the logarithm of the score rounded up to an integer, maybe even as a number of stars. If the base-3 logarithm was used, the displayed scores would be:

    -infinity to 0:    bad comment, should be dimmed
    1:                 
    2-3:               ★
    4-9:               ★★
    9-27:              ★★★
    27-81:             ★★★★
    81-243:            ★★★★★

etc. Though I think scores should be computed in a more sophisticated way than they are currently.

View on HN

There's a lot of good thinking here, but all of these suggestions share the same failing as integer scores: They still don't preclude bandwagoning. A polarizing and emotional rant with a five-star/bright-green/100%-bar rating is barely less susceptible to mob voting than a polarizing and emotional rant with 55 points.

That said, I don't think hiding scores is the only solution to this problem. I'd like to suggest an escape valve of sorts: If we accept that, from time to time, a bandwagon-bait comment will inevitably get voted up, we can consider a mechanism to deflate it. There is a simple way, to designate moderators who can demote these comments; then there is a more interesting way:

I hypothesize that there is an observable pattern to the votes over time of bait comments. It would be very interesting to see a comparison of highly-rated comments through a histogram: Is there a particular curve that insightful comments follow? Do bait comments have a signature spike at their tipping point? Can these be modeled in a straightforward way? The system could then be set up to slow, halt, or even reverse upvoting when a comment fits the model.

Is mob voting an observed problem with the old way? I seem to recall PG mentioning[1] that overall top comment scores are now /higher/ than before, suggesting that upvoters didn't mob in the old system, but based their votes on the perceived value of the comment vs its current score. Thus, instead of the prevalence of vote mobbing, a more important question is: Do you think a score should be based on a ternary choice between {good comment, indifferent, bad comment} summed over a large (sometimes not-so-large) group of votes, or based on the average of the community's perception of the comment's value on a sliding scale?

[1]http://news.ycombinator.com/item?id=2465271

I think it would be a good idea to display the scores after you vote. That way you would know that you are upvoting a comment for it's merits, and it's not a case of mob mentality.

I think that would be more polarizing rather than less: You would have an incentive to vote one way or the other, even when you feel the comment deserves neither.

We could add a "+0" vote in addition of +1 and -1 votes to solve this problem.

A neutral vote ("I don't want to vote") should definitely be present, since many users don't have the ability to downvote (e.g. myself).

SHOUTING is a good way to get karma under the new system. It's rare for people to downvote, so comments that scream LOOK AT ME will get lots of upvotes.

Just as long as you don't overdo it, or say anything too controversial.

> A polarizing and emotional rant with a five-star/bright-green/100%-bar rating is barely less susceptible to mob voting than a polarizing and emotional rant with 55 points.

IIRC, the comments I made that got the highest scores were accidental emotional rants that, more than embody some clever original insight, were in agreement with a large number of people. Usually the well reasoned ones get around 5. To get above that, I'd really have to play for the crowd, something I am not inclined to do. I value this place.

Then maybe a comment should get a single star if it gets a moderate share (5%) of the total points, or at least 5 pts, whichever is higher. Zero starts would be the default.

If we're using stars, why not base stars on comment percentile? Every night at midnight, take all the comment scores for the last week, and determine the 90th, 70th, 50th, and 30th percentiles. So every comment in the top 10% gets 5 stars, every comment in the top 30% but not in the top 10% has 4 stars, etc.

Also, I think we should switch it up so that comments are dimmed at more like -1 or -2. Otherwise, if the first vote is a downvote, people might overlook an otherwise decent (or controversial) comment.

> Also, I think we should switch it up so that comments are dimmed at more like -1 or -2. Otherwise, if the first vote is a downvote, people might overlook an otherwise decent (or controversial) comment.

This is needed. I'm constantly upvoting valid, neutral things even if I don't really have an opinion on them simply because their first vote was down, likely by a disgruntled parent post.

>likely by a disgruntled parent post.

You've never been able to downvote replies to your own comments.

Oh wow 3 years and I've never noticed. Haha tells you've I've never tried to then. ;)

Unless they have a friend or a second account who can do it for them ;)

The karma threshold for downvoting comments is 500, iirc. That would make socking a fairly laborious prospect. As for meatpuppetry, if you're going to go through that much trouble just to downvote someone, then I'd suggest you have deeper issues.

I'm often drawn to greyed things, as they're different...not sure how prevalent that reaction is, but its at least a counterweight to overlook.

Comment's should only reflect the current thread (well, everything underneath a news story). Not all stories receive the same traffic. A quality comment in one story might receive far less votes than a less quality comment in another story that is far more popular.

However, that is merely the method, and the idea of identifying quality stories is still sound.

I use something much like this on my forum and it seems to work well.

People only need a general idea of the score for it to be informative of value.

Reply on news.ycombinator.com