Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

DragonGod comments on LW 2.0 Strategic Overview - Less Wrong

47 Post author: Habryka 15 September 2017 03:00AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (294)

You are viewing a single comment's thread.

Comment author: DragonGod 16 September 2017 05:13:39PM *  6 points [-]

On StackExchange, you lose reputation whoever you downvote a question/answer; this makes downvoting a costly signal for displeasure. I like the notion, and hope it is included in the new site. If you have to spend your hard-earned karma to cause someone to lose karma, then it may discourage karma assassination, and ensure that downvotes are only used on content people have strong negative feelings towards.

Pros

  1. Users only downvote content they feel strong displeasure towards.
  2. Karma assassination via sockpuppets becomes impossible, and targeted karma attacks through your main account because you dislike a user becomes very costly.
  3. Moderation of downvoting behaviour would be vastly reduced as users downvote less, and only on content they have strong feelings towards.

Cons

  1. There are much less downvotes.
  2. I don't think downvotes should be costly. On StackExchange mediocre content can get a high score if it relates to a popular topic.
    Given that this website has the goal of filtering content in a way that allows people who only want to read a subset to read the high quality posts downvotes of medicore content as useful information.

I think the first con is a feature and not a bug; it is not clear to me that more downvotes are intrinsically beneficial. The second point is valid criticism and I think we need to way the benefit of the downvotes against their cost.

I think you lose one reputation, per downvote, and cause the person downvoted to lose 2 - 5 reputation.

I think downvoting costing 0.33 - 0.5 the karma you deduct from the target of your downvote is a good idea, and will encourage better downvote practices and would overall be an improvement to the karma feature.

Comment author: Habryka 16 September 2017 11:30:34PM *  4 points [-]

Hmm... I feel that this disincentivizes downvoting too strongly, and just makes downvoting feel kind of shitty on an emotional level.

An alternative thing that I've been thinking about is to make it so that when you downvote something, you have to give a short explanation between 40 and 400 characters about why you think the comment was bad. Which both adds a cost to downvoting, and actually translates that cost into meaningful information for the commenter. Another alternative implementation of this could work with a set of common tags that you can choose from when downvoting a comment, maybe of the type "too aggressive", "didn't respond to original claim", "responded to strawman", etc.

Comment author: DragonGod 17 September 2017 01:09:44AM *  1 point [-]

Hmm... I feel that this incentivizes downvoting too strongly

How does this incentivise downvoting? Downvoting is costly signal of displeasure, and as downvotes cost a certain fraction of the karma you deduct, it disincentivises downvoting.

makes downvoting feel kind of shitty on an emotional level.

This is a feature not a bug; we don't want to encourage downvoting and karma assassination. The idea is that downvoting becomes costly signalling of displeasure. Mere disagreement would not cause downvoting. Downvoting should be costly signalling.

An alternative thing that I've been thinking about is to make it so that when you downvote something, you have to give a short explanation between 40 and 400 characters about why you think the comment was bad. Which both adds a cost to downvoting, and actually translates that cost into meaningful information for the commenter.

I thought of this as well, but decided that the StackExchange system of making downvotes cost karma is better for the purposes I thought of.

Another alternative implementation of this could work with a set of common tags that you can choose from when downvoting a comment, maybe of the type "too aggressive", "didn't respond to original claim", "responded to strawman", etc.

This fails to achieve "adds a cost to downvoting"; if there are custom downvoting tags, then the cost of downvoting is removed. I think making downvotes cost a fraction (<= 0.5) of the karma you deduct serves to discourage downvoting.

Comment author: Habryka 17 September 2017 02:40:51AM 0 points [-]

"How does this incentivise downvoting?"

Sorry, my bad. I wanted to write "disincentivize", but failed. I guess it's a warning against using big words.

Comment author: DragonGod 17 September 2017 09:24:07AM *  1 point [-]

Oh, okay. I still think we want to disincentivise downvoting though.

Pros

  1. Users only downvote content they feel strong displeasure towards.
  2. Karma assassination via sockpuppets becomes impossible, and targeted karma attacks through your main account because you dislike a user becomes very costly.
  3. Moderation of downvoting behaviour would be vastly reduced as users downvote less, and only on content they have strong feelings towards.

Cons

  1. There are much less downvotes.
  2. I don't think downvotes should be costly. On StackExchange mediocre content can get a high score if it relates to a popular topic.
    Given that this website has the goal of filtering content in a way that allows people who only want to read a subset to read the high quality posts downvotes of medicore content as useful information.

I think the first con is a feature and not a bug; it is not clear to me that more downvotes are intrinsically beneficial. The second point is valid criticism and I think we need to way the benefit of the downvotes against their cost.

I suggest users lose 40% of the karma they deduct (since you want to give different users different weights). For example, if you downvote someone, they lose 5 karma, but you lose 2 karma.

Comment author: NancyLebovitz 17 September 2017 07:32:02PM 2 points [-]

How about the boring simplicity of having downvote limits? Maybe something around one downvote/24 hours-- not cumulative.

If you're feeling generous, maybe add a downvote/24 hours per 1000 karma, with a maximum or 5 downvotes/24 hours.

Comment author: J_Thomas_Moros 18 September 2017 05:41:56PM 1 point [-]

I'm not opposed to downvote limits, but I think they need to not be too low. There are situations where I am more likely to downvote many things just because I am more heavily moderating. For example, on comments on my own post I care more and am more likely to both upvote and downvote whereas other times I might just not care that much.

Comment author: DragonGod 17 September 2017 08:02:19PM 1 point [-]

This is a solution as well; it is not clear to me though, that it is better than the solution I proposed.

Comment author: ChristianKl 20 September 2017 01:20:29PM 2 points [-]

I don't think downvotes should be costly. On StackExchange mediocre content can get a high score if it relates to a popular topic.

Given that this website has the goal of filtering content in a way that allows people who only want to read a subset to read the high quality posts downvotes of medicore content as useful information.

Comment author: DragonGod 20 September 2017 07:32:43PM 0 points [-]

I'll add the point you raise about downvotes to the "cons" of my argument.

Comment author: Viliam 19 September 2017 10:58:14PM *  1 point [-]

So... let's imagine that one day the website will attack e.g. hundreds of crackpots... each of them posting obviously crazy stuff, dozens of comments each... but most people will hesitate to downvote them, because they would remember that doing so reduces their own karma.

Okay, this will probably not happen. But I think that downvoting is an important thing and should not be disincentivized per se. Bad stuff needs to get downvoted. Actually, other than Eugine, people usually don't downvote enough. (And for Eugine, this is not a problem at all; he will get the karma back by upvoting himself with his other sockpuppets.)

I think it is already too easy to get a lot of karma on LW just by posting a lot of mediocre quality comments, each getting 1 karma point on average. Sometimes I suspect that maybe half of my own karma is for the quality of things I wrote, and the remaining half is for spending too much time commenting here even when I have nothing especially insightful to say.

Comment author: DragonGod 20 September 2017 07:53:21AM *  0 points [-]

Okay, this will probably not happen.

Thank God you agree, and thus I think it's value as a thought experiment is nil.

But I think that downvoting is an important thing and should not be disincentivized per se.

Disincentivising downvoting discourages frivolous use of downvotes, and encourages responsible downvoting usage.

If you just disagree with someone, you're more likely to reply than downvote them if you care about your karma for example.

Actually, other than Eugine, people usually don't downvote enough. (And for Eugine, this is not a problem at all; he will get the karma back by upvoting himself with his other sockpuppets.)

On StackExchange upvotes and downvotes from accounts with less than 15 rep are recorded but don't count (presumably until the account gains more than 15 rep). LW may decide to set her bar lower (10 rep?) or higher (>= 20 rep?), but I think the core insight is very good and would be a significant improvement if applied to LW.