Every time this comes up, I like it less. I now think it is indeed a bad idea. It gives people meaningless information in the guise of useful information.

The purpose of voting is for individuals to say "this node is of helpful to the community" and "this node is harmful to the community". You can't even count on that, in all cases.

For a node's reputation to have any meaning as a number, you'd need several other statistics: the number of people who voted null on it, the number of people who viewed it, the number of people who had votes that could be spent on it, the number of positive votes, and the number of negative votes.

You'd also likely need people to provide some sort of reason why they voted one way or another. We've had that discussion before, several times.

I'd really like to see an example where knowing a node's reputation would have been useful. I apologize in advance for not counting "'cuz I'm curious!" as a useful reason.

(As a side note, one of the biggest mistakes I've ever seen in reputation systems was on Slashdot, by making karma public. My rule is, "Any moderation system that causes conversations to degenerate into discussions about the moderation system is broken." Omphaloskepsis is the death rattle of a community.)

It gives people meaningless information in the guise of useful information.

I don't agree that a node's reputation is completely meaningless. In the context of someone who has asked a question near the edge of their abilities and who now must sort through N disparate suggestions, the ability to discover that N/4 (or even 1) of those suggestions has been deemed to be wrong is very helpful. Ordering by reputation (but without seeing reputation) doesn't give you that. You can't distinguish between the end of the list still being a good suggestion, and it being deemed nonsense.

Again, a node's reputation has, at best, a tangential relationship to its correctness.

Again, I don't see how knowing that one reply has a reputation of 12 and the other has a reputation of 2 means anything. Too many other factors affect node reputation: time of day, day of week, tone of post, logged-in users, whether a node is on the front page, whether a node is in Newest Nodes, along with features that are even harder to measure.

This proposal is trying to put weight on a foundation that won't hold.

I've replied to dozens of incorrect posts that had positive reputations. The best way to mark a node as incorrect is to reply with a correction, not to downvote it and hope that the original poster sees the reply before attempting to take the advice of the incorrect node, that the original poster casts enough null votes to rank the reputations of these nodes in relationship to each other in a meaningful way, or that the original poster has reputation ordering enabled.

The purpose of voting is for individuals to say "this node is of helpful to the community" and "this node is harmful to the community".

I don't agree. I think that a reply like "read this documentation" or "I have an article on that" might be helpful to the community but I often downvote them, especially if a better answer occurs in the thread. In fact, my normal strategy for voting on the replies in a thread¹ is to try to order them best to worst (IMHO, of course) when viewed by reputation. I do that by first upvoting the reply which I think is best and sometimes by downvoting nodes which have a higher reputation than it does. (I order by rep so I know the relative order.)

For a node's reputation to have any meaning as a number, you'd need several other statistics

Other statistics would be nice. They aren't necessary.

Do you vote with any strategy other than a random one? Because, if you do, then you don't act as if node reputation is meaningless. It's ridiculous to cast a vote with intention and then argue that node rep is meaningless. I wholeheartedly agree that the meaning of a node's reputation is hard to pin down. It exists only in context. It's entirely relative. But that doesn't mean that it doesn't exist. I vote as if it does exist; therefore, it does.

My rule is, "Any moderation system that causes conversations to degenerate into discussions about the moderation system is broken." Omphaloskepsis is the death rattle of a community.

I think that contemplating the moderation system is a bit more like pondering one's tactile nervous system than his belly button.

1. This is the strategy I usually apply to replies to technical questions. On matters of opinion, like this thread for instance, I usually just upvote nodes I agree with. I upvote meditations and replies to them if I find them interesting.

Edit: Changed "A think" to "I think".

-sauoq
"My two cents aren't worth a dime.";

Do you vote with any strategy other than a random one? Because, if you do, then you don't act as if node reputation is meaningless.

Before trotting out the tired old accusation of hypocrisy, please review my argument. The reputation of an individual node is almost completely meaningless.

On those rare occasions when I vote (perhaps one vote a week), my strategy is very much like yours. I upvote posts, such as the one by chunlou in this very discussion, that appear near the bottom of the discussion and, as such, are undervalued.

My desire is not to say "This node deserves a 3, while this node should never be more than 1". It's to make a single suggestion as to the relationships between nodes within a discussion for people who have reputation ordering enabled. That's it.

My vote counts for very little in a discussion. That's fine. No one's ever promised that the absolute best, clearest, and most accurate answer will always float to the top, without fail. I distrust any system that tries to make that kind of promise — it's not a promise that can be fulfilled, and I'd much rather say "We have a working system that gives you a very rough guide that may be useful. Take it as a suggestion, but continue to use your brain."

(Okay, there's also downvoting to trigger automatic reaping of nodes that are completely devoid of worthwhile content, but that's a separate issue.)

I don't agree. A think that a reply like "read this documentation" or "I have an article on that" might be helpful to the community but I often downvote them, especially if a better answer occurs in the thread.

I disagree. I think a reply with a reference link can be just as useful as an answer that reveals and discusses the answer. I downvote unhelpful, flippant answers, and incomplete answers that were obviously dashed off in order to make it to be first post. I don't downvote a reply that's just a link .. sometimes, in order to solve a problem, a link to the right page is all that's necessary.

--t. alex
Life is short: get busy!

I have to agree with chromatic here.

Let's say this option was available.

You view a thread.

You vote neutral on a single node and learn that its rep is 11.

What have you learned?

Nothing.

You have to see the reputations of a bunch of other nodes before you can draw any conclusion from those 11 points.

So what you want to know is the relationship of nodes to each other with regards to their reputation. That is a useful metric. And it is one that's already available..

Makeshifts last the longest.

You vote neutral on a single node and learn that its rep is 11. . . . What have you learned? . . . Nothing.

Au contraire! You've learned that the node's reputation is eleven. What that eleven actually means depends entirely on context but it isn't meaningless. Writing about a fictional node with a reputation of eleven, as you have done, is meaningless.

You have to see the reputations of a bunch of other nodes before you can draw any conclusion from those 11 points.

That might help, as would any contextual information, but it isn't strictly necessary and may not even be available (such as if it is the only reply in the thread.)

So what you want to know is the relationship of nodes to each other with regards to their reputation. That is a useful metric.

I agree that it's a useful metric and I always order nodes by reputation. But, it is nonsensical to argue that the relative reputations matter on the one hand and then deny that the degree of difference doesn't on the other. Examples:

• The two highest rep replies to a node have reps of 73 and 71 respectively. Does their order matter more than their reputation?
• The three highest rep replies to a node have reps of 11, -2, and -3 respectively. Does their order matter more than their reputation?
• The only two replies to a node that has been on the front-page for a day and half have reps of 40 and 19 respectively. The one with 40 was posted 20 minutes after the original. The one with 19 was posted 20 minutes ago. Does order matter more than reputation? (My answer: No, they are both misleading in and of themselves, however, the actual rep gives more information than the order.)
• There are a dozen replies to a node and the top four have reps over 20. The three highest rep nodes in the thread are all by well-known and prolific saints who have misintrepreted the question. The fourth highest is an anonymonk post and answers the question accurately and succinctly. When reviewing that fourth node, which matters more, rep or relative position?
• There is a single reply to a node (so order cannot matter.) It was posted 3 minutes ago and has a rep of 7. Conclusion: probably a good reply.
• There is a single reply to a node (so order cannot matter.) It was posted 3 days ago and has a rep of 2. Conclusion: An underexposed and/or difficult question, a mediocre or controversial reply. (Should the node be FP'd?)

I think providing more contextual information, including reputation, is better. I don't think it should be provided prior to voting but I think we should have the choice to refrain from voting on a node and to reveal its reputation. Note that I advocate making such a choice free of side-effects in that it would neither cost votes nor result in XP. What's the harm in allowing a monk to reveal to himself the reputation of nodes he will never vote on? I've yet to see a convincing argument that there is one. Those who oppose it all argue the untenable position that it just wouldn't do any good. Inevitably, that argument is based on the assumption that node reputation is meaningless or "practically" so, an assumption which has been handily trounced time and again.

-sauoq
"My two cents aren't worth a dime.";

So let's say you indeed have done so and now you know that that node is at 11 points. How do you interpret this? What does that mean?

Knowing whether a node has negative rep or not is somewhat valuable in itself. I actually thought of that, though I didn't mention it. Granted.

All your other points still reduce to examining relations of node reps. What you're really looking for is some kind of standard deviation metric. The 71 rep node might have 11.3 standard points deviation, while the 73 rep one has 11.4 standard points. And all other nodes are within 1.5 standard points. Or something like that.

Is that any less information than can be drawn from knowing the absolute values of the node rep? I think not.

On the other hand, just learning about one single node that it has 11.4 standard points is already very useful without looking at any of the nodes. Because it is a relative metric.

Makeshifts last the longest.