Meta: What do you think of a karma vote checklist?

I'm not sure it will do much good, but here is the post, and this is a permanent link to the discussion as it stands now. This was a goodbye post, to AD, one of the seemingly saner members of the RationalWiki community, an elected moderator. There is a link in my goodbye post back to AD's comment in a discussion that included history, but that's a lot more than I expect people here to be interested in. Suffice it to say that the user has a history of being exactly what he says he is, a highly effective troll. He says "professional."

(To understand some of the discussion, "promote" on RatWiki means "remove sysop privileges" or sometimes "block.")

Thanks for writing the followup.

No.

EDIT: Okay, still no, but Slashdot has a very different culture than over here. It's all news postings, there's a huge rush, most posts are shit and spam, the fact that some good stuff manages to rise to the top is a minor miracle. The truth is, we have a culture that encourages more thought per posts, a smaller number of active posters, and if you do write something crappy or unpopular, most people'll just reveal it anyway. On Slashdot, doing through all the comments on one post would take hours for a typical post, so you write something one mod downvoted cause he had some free ones to spare or he doesn't like you very much, no one sees it, and your Karma drops to below zero, so no one will ever see your post ever again.

The proper way of dealing with trolls is to ignore them altogether. Even downvoting them gives them the attention they don't deserve.

it's clear to me that I can get upvoted reliably if I clearly articulate and intelligently apply the LW-consensus analyses. I will get downvoted severely if I--stupidly or intelligently--articulate an original position that contradicts the LW consensus. --common_law.

Let me substantiate. Here's a post where I recently articulated and applied the LW "line."

On the other hand, a post in this thread, where I advance an idea that should be important to to people who build a rationalist community:

The main factor (but not the only factor) determining a vote is whether the rater agrees with the poster. Perhaps that's as it should be, since without it, LW could never evolve a party line (so to speak).

Not to be immodest, but this insight is the product of years of watching and experimenting with LW, and I only reached it recently. If it's true, it's important because LW is obstructed in constructing a rational community when it ignores the primary function of its "institutions" and substitutes idealistic thinking ('an upvote means you want more of the same') for a functional analysis.

Once you see the actual role of karma, you might realize that it couldn't be otherwise. A a massive intellectual community must find a way to evolve a dynamic consensus. It requires objective incentives to coordinate on a singe outlook (or on a narrow spectrum of outlooks).

In point of fact, this how leading LWers sometimes speak. For example, LukeProg argued that the community isn't a cult around E.Y. because early on, his posts were upvoted in some cases when E.Y.'s were downvoted. LukeProg's comment implies LW took an ideological or practical direction through the karma mechanism.

In this comment I'm not evaluating the karma mechanism but pointing out that it is L.W.'s soul.

I experiment with these things. Based on a period of years, during which I occupied different personae, it's clear to me that I can get upvoted reliably if I clearly articulate and intelligently apply the LW-consensus analyses. I will get downvoted severely if I--stupidly or intelligently--articulate an original position that contradicts the LW consensus. As I say, I don't see how it can be otherwise, if LW is to function as a community with consensus views on eclectic matters.

I wonder what you mean by "in theory." If that means "according to the standard LW rhetoric," I agree.

We need separate buttons and votes need various changes to be effective at encouraging improvement. Negative karma cannot provide new users with useful feedback, deter trolls and inform the popularity sorting system all at the same time. Here is why:

Unexplained downvotes don't tell users what to improve, so they're as good as ignored

If we feel a need to have a way to mark the low quality comments, and provide feedback to discourage annoyance, we should definitely ask that person to give a few words of feedback. As a new person who occasionally gets vote downs, I have absolutely no clue whether there is any pattern to them. They're not encouraging me to acculturate at all. It just looks random. I've theorized that maybe somebody doesn't like me, or that maybe I made a spelling error I didn't catch, or maybe a troll is harassing me. I have no clue. This speculation isn't getting me anywhere. I need specific feedback.

What would help me (and other new users) improve

If the vote downs provided at least a couple of words of feedback, and didn't censor me, I'd be happy with that. Getting "Not new information" or "joke was not funny" would be much preferred to random anonymous discouragement. A category would be better than nothing, but a couple specific words would be best.

Reserving feedback for older members will reduce endless September risk and increase feedback value

The way this would reduce endless September risk is by highlighting the interactions with older users, which will amplify their ability to maintain culture if the ratio of new to old people is thrown off. Also, as a new user, I want to know that the feedback I'm getting is from somebody helpful, not a troll, or a newbie. This doesn't guarantee that old member's feedback is useful but it would prevent me from brushing them off thinking I might be getting harassed by trolls. If there's a possibility trolls are giving the feedback, I really can't take it seriously.

Attaching a punishment to a downvote sacrifices honest feedback in favor of not punishing the user

Consider this: You dislike somebody's comment, but you don't see them as a troll or a bad person. They've got two downvotes. Do you downvote them further and hide their comment? Probably not. That's what a lot of people are thinking, because most of the time when I see downvotes, it's just one or two. You know what I think when one or two people didn't like my comment? "Can't make all the people happy all the time. Oh well."

Now if TEN people vote me down, I know something's going on. THEN I can be completely sure it's not a personal vendetta or someone having a bad day. We cannot take vote downs as a serious source of feedback if we only get one or two, and people won't give you more than one or two if you're punished at three.

Merging upvotes and downvotes into one figure promotes black and white thinking about popularity

Also, seeing the number of up votes versus down votes would be useful. If I say something that gets 40 up votes and 43 down votes, all I see in the current system is down votes. I may think most people are ignoring me or don't like what I said. It could simply be that I said something controversial - and that can be a good thing. To avoid discouraging controversy, or to put it another way, to prevent conformity, showing both sides of the story is important. It may be the practice of showing only one side of the story that promotes conformity. Showing both might be a good way to prevent conformity.

How the separate ban button could work for trolls:

Giving a large portion of users the power to ban trolls would expedite the process of removing them. This can't be a small number of people, or it would invite politically motivated "assassinations". This can't be new users because if new users flood the forum they may ban old users due to cultural misunderstandings who were contributing to the culture. If, say 5% of older active users press the ban button, it would disable the troll's account. Currently, that would mean around 25 to 50 people if Vladimir is correct. They can't be people with too little activity - this would encourage trolls and malicious users to create 50 dummy accounts, wait a few months, and then ban people they don't like. The ban button can be totally anonymous and invisible, preventing trolls from getting any feedback at all, because it is completely separate from the voting system.

A "public vote system" has been used for centuries in standard deliberative process. You go to a Town Meeting and think that a question should not be considered, and you so move, and that is subject to immediate and very public vote. Private voting systems have been used and often have an abusive effect. Such systems, in standard process, when allowed, generally require a supermajority. Elections are an exception, where secret ballots are standard.

Much comment here seems to assume yes/no on "private." It's possible to collect data on "impressions" that is private, and it is not necessarily abusive. It can become abusive when this is used in a fixed decision-making system.

The karma system is quite popular, and the way it works should not, ideally, be damaged by "improvements." Improvements may address the ways that it does not work, and there are a number.. There are many good ideas in this thread. Some of them, implemented raw, could do harm. Hence the need for discussion and the development of informed consensus, which can be very different from raw, knee-jerk consensus. Such raw consensus can be used to develop starting points, and is worthy of respect, but not worship.

Otherwise a community is vulnerable to cascades and to confirmation bias.

Standard deliberative process uses committee systems for topics not ready for full consideration and vote. The conversations take place in small groups, where brainstorming may be more open and less harmful, and, ideally, all significant points of view are represented in those groups. Distributed communication is essential for sound and efficient social process.

Hey look it seems like replying to a comment that is a reply to a comment that was negative doesn't cost anything. So as long as one person takes the 5 karma hit, the disincentive no longer applies. Good to know!