How about deleting some Touhou?

Actually, all of your skipped touhou posts were uploaded by contributors, so the approvers had nothing to do with it. Perhaps you should demote people who you feel are unable to objectively judge the quality of Touhou art to privileged status instead. (If this includes me, I have no objection; I generally don't post Touhou anymore though, unless it's from an artist I follow.)

One of your skipped touhou posts was uploaded by albert, so perhaps you should talk to him about having his posts require moderator approval as well.

Suiseiseki said:
Slightly off topic, anybody that thinks that anything Touhou gets accepted, please search 東方 on pixiv and upload everything you see there, and watch how much gets deleted. There is a TON of bad Touhou art on pixiv that will never see danbooru.

Or better yet, throw touhou status:deleted at them. Plenty of examples right there.

LaC said: My results seem to support 葉月's observation.

For your opinion only, though. I disagree with some of your skips, including 3 definites that are touhou, which would bring it to ~6 making it statistically insignificant for the very small sample size we'd be left with. What is so deletable about post #456222, post #456202, or post #456157? One or two others are probably debatable but I don't feel strongly enough to bring them up.

[Edit: It's also worth noting that 4 of the 5 or 6 I agree are sub par were by the same person. *Who* is uploading in the small sample size has a big impact too. Though before it sounds like I'm picking on mr_gt, the majority of his uploads are perfectly fine.]

Also, why was this run on safebooru? I can see some benefit in that it would reduce bias for/against explicit images in general, but there's a huge methodological problem. As I'm sure you know, a far higher percentage of Touhou posts are rating:safe than is the norm across all images.

Quick calculation shows something like 59% of all images are safe, but for Touhou it's 86% safe. Therefore, any given "bad" Touhou image is much more likely to be found when restricted to safe images only than any given "bad" image over all.

I appreciate that you tried to quantify it but your numbers don't support the case very strongly. And at it's core it's still highly individual and subjective, as disagreement over those three images shows.

The point remains the same. You deal with this by demoting the users who make questionable uploads, not by a thread like this.

jxh2154 said:
For your opinion only, though. I disagree with some of your skips, including 3 definites that are touhou, which would bring it to ~6 making it statistically insignificant for the very small sample size we'd be left with. What is so deletable about post #456222, post #456202, or post #456157?

Deletable is not the right word: I almost never delete posts. I do find those posts mediocre, and think they'd probably not have been approved if they didn't feature Touhou characters.
post #456222 and its child are retarded. We have enough unfunny comics, but this one doesn't even make sense.
post #456202, once de-tohoed, would be just a girl with elephantiasis.
post #456157 is just ugly, her mouth and nose are misaligned from the axis of her head (or perhaps the right side of her skull is stretched out, I don't know).

But yes, it's a subjective test. It has to be: clearly, for every picture that has been approved, there is at least one person who thinks it's fine.

Also, why was this run on safebooru? I can see some benefit in that it would reduce bias for/against explicit images in general, but there's a huge methodological problem. As I'm sure you know, a far higher percentage of Touhou posts are rating:safe than is the norm across all images.

Quick calculation shows something like 59% of all images are safe, but for Touhou it's 86% safe. Therefore, any given "bad" Touhou image is much more likely to be found when restricted to safe images only than any given "bad" image over all.

Yes, but it's not like people set out to post a bad image and think "let's see, for idolmaster, let's pick a bad explicit one... ok, next up is touhou, let's go with a bad safe one".
What we're looking for are hidden mechanisms that make people post or approve bad images unwittingly. We are already aware of a big one: when it comes to porn, people tend to think with their little head, and approve subpar content. Therefore, any tag that has little explicit or questionable content is going to be underrepresented in the bad posts, simply because it doesn't trigger porn mode. But there might still be other mechanisms that affect the approval of non-erotic content; if we want to find them, we have to correct for the effects of porn mode by excluding erotic posts. And that's why I used safebooru. (Also, I was on safebooru when I read this thread.)

Another way to look at it is this: if there were more Touhou porn, it wouldn't replace the bad safe posts we have now; there would simply be a bunch of bad explicit posts in addition to them, and the percent of bad Touhou posts overall would be higher. But the situation with the safe posts would be unchanged, and that's what I was concentrating on.

The point remains the same. You deal with this by demoting the users who make questionable uploads, not by a thread like this.

Indeed, while I agree with 葉月's assessment of the problem, I don't agree with his proposal to address it. I'm afraid there's nothing we can do.

LaC said: post #456222 and its child are retarded. We have enough unfunny comics, but this one doesn't even make sense.

I don't like the child, but hadn't even looked at it before and am biased against explicit things anyway. No, the parent doesn't make much sense (may be referencing a Touhou joke I'm not familiar with), but I like the oekaki style. At the last panel and into the child, not so much. I would have approved the parent and not the child.

post #456202, once de-tohoed, would be just a girl with elephantiasis.

The legs look like a POV thing, not like they're intended to be unnaturally enormous in proportion to her body. I like the style of the rest of it too.

post #456157 is just ugly, her mouth and nose are misaligned from the axis of her head (or perhaps the right side of her skull is stretched out, I don't know).

This is the weakest of the three, but if I saw it in the new posts listing I wouldn't see anything about it that would make me look twice and wonder why it was there, Touhou or otherwise.

Yes, but it's not like people set out to post a bad image and think "let's see, for idolmaster, let's pick a bad explicit one... ok, next up is touhou, let's go with a bad safe one".

My point is that it makes your raw percentages questionable. In terms of simple statistics, you used a flawed sample. You propose that the percentage of bad Touhou is higher than the percentage of bad images generally, but then use a sample that is proportionally more likely to have bad Touhou, which means it's not an apples to apples comparison.

You can offer your subjective impression (which is what my earlier comment was where I stated the opposite was), but the statistical evidence doesn't appear to offer much insight into this. I make a point of highlighting the flaws in the methodology because once people start seeing numbers, they're apt to accept them as "more trustworthy" if nobody challenges it.

Its not about how much of it there is, its about quality. Putting touhou up as a sacrifice doesn't make sense even though there is an assload of it.

If poor quality stuff gets in the system, then its the moderators/admins/contributors that need to be examined, not a copyright alone.

EDIT:

evazion said:
On the other hand, publicly singling out other people for their uploads could easily start a bunch of butthurt and e-drama, so I dunno.

This 100%. If anything the mods/janitors should silently delete posts anonymously so nobody gets butthurt at a single person. I see plenty of crap that I want to flag, but would it really help? People will just think you're a jerk if you flag their stuff.

If it's quality we're concerned about, I don't see how touhou is involved. We should purge pictures that have no merit whatsoever regardless of the touhou tag.

...on another note, are we in danger of running out of space or something? What's with this sudden push for more urgent removal of worthless content?

One more post against starting a witch hunt. Bad posts slip through for all copyrights. If anything, we should have a reversed version of the "deletion appeal" thread, but I agree with jxh2154 – it's probably better to deal with contributors and janitors individually here.

Granola said:
If poor quality stuff gets in the system, then its the moderators/admins/contributors that need to be examined, not a copyright alone.

Yeah, I've been seeing a pretty surprising amount of crap being uploaded by contributors lately. And not just DeviantArt-quality crap, but things that aren't even supposed to be on here, as per the TOS.

Well if you've seen it, post the examples.

aldeayeah said:
Well if you've seen it, post the examples.

But that would cause massive drama.

How about the moderators have their own deletion queue for already approved images? That way they can stick pictures that are low quality in it and it would be anonymous deletion. Through this we can get moderator consensus on what should be deleted instead of having a decentralized thread about it (its just not a good way to organize stuff). This way we avoid the "OH GOD SHINJIDUDE YOU DELETED MY 1500 DICKGIRL UPLOADS IM GONNA HATE YOU FOREVER" event.

I personally don't care for Touhou, but I have to echo everyone else about going through specific copyrights. To change gears slightly, I love K-on! and Mio, but some of the ones that slip through are less than desired quality.

It doesn't address the issue of contributor uploads, but what if an image had to be approved by three people or something instead of instant approval from one? Dunno how easy that'd be to implement and is probably more work than it's worth, but it would at least help with biased uploads.

Granola said:
EDIT: This 100%. If anything the mods/janitors should silently delete posts anonymously so nobody gets butthurt at a single person. I see plenty of crap that I want to flag, but would it really help? People will just think you're a jerk if you flag their stuff.

Excuse me, what? How exactly are we supposed to find the bad stuff if no-one reports it? I'd like to remind you that danbooru is not my paid job, I can't put any more time in it than I already do. And anyway, whether a picture is deleted or unflagged, the information about who flagged it is erased.

PoringSmasher said:
It doesn't address the issue of contributor uploads, but what if an image had to be approved by three people or something instead of instant approval from one? Dunno how easy that'd be to implement and is probably more work than it's worth, but it would at least help with biased uploads.

This is certainly not the worst idea ever and might be worth trying, but I think it would drastically reduce the number of images that make it through the queue... That may be a desired effect, I guess.

Besides, I think a lot of questionable quality, especially on popular tags, is coming from contributors.

I'm with you on the K-on! thing all the way btw, it seems worse than either Vocaloid or Touhou.

jxh2154 said:
My point is that it makes your raw percentages questionable. In terms of simple statistics, you used a flawed sample. You propose that the percentage of bad Touhou is higher than the percentage of bad images generally, but then use a sample that is proportionally more likely to have bad Touhou, which means it's not an apples to apples comparison.

Not at all. 葉月's hypothesis was that people are more lenient with posts that are Touhou. To verify it, we have to correct for other factors. The right sample isn't necessarily the widest sample possible.

Let me give you an example using the king of similes, the automobile. Let's say we hypothesize that people like luxury cars, and want to see if that is reflected in their buying habits. A naive approach might be to look at the number of luxury cars over the total cars sold; that would show that they make up a small percent of the market, apparently disproving the hypothesis.
However, if you understand the problem space you'll soon realize that for most people luxury cars are just too expensive. That fact introduces a strong bias against sales of luxury cars, with no relation to whether people actually like them or not. Therefore, you change your sample to include only those buyers that can afford luxury cars in the first place. This time, you find that they command a much larger percentage of the market, confirming that people do like luxury cars - as long as they can afford them.

Granola said:
If poor quality stuff gets in the system, then its the moderators/admins/contributors that need to be examined

You can't remove contributor privileges from a user unless he posts one hundred pictures of shitting dicknipples in a row. The risk of massive drama is just too high. If a contributor starts pumping out mediocre uploads, he can easily get in a state where he would not be awarded contributor status if he was a regular member - but since he already has it, he'll feel entitled to keeping it.

It would be useful to have a way to throw an auto-approved post by a contributor back on the approval queue. Not on the deletion queue, mind you; even though the end effect would in fact be deletion, it takes a blatant rule violation or incredibly horrible quality for moderators to delete a flagged post, so it's not a good way for dealing with mere mediocrity.

LaC said:
It would be useful to have a way to throw an auto-approved post by a contributor back on the approval queue.

This sounds like the better idea to me.

BTW, K-on is full of crap uploads too.

LaC said:
It would be useful to have a way to throw an auto-approved post by a contributor back on the approval queue. Not on the deletion queue, mind you; even though the end effect would in fact be deletion, it takes a blatant rule violation or incredibly horrible quality for moderators to delete a flagged post, so it's not a good way for dealing with mere mediocrity.

Yeah, I've wanted a "Disapprove" button for a long time. Seems like the mechanism we really want. Just a heads up before it gets implemented: It'd need to deal with the fact that "Uploaded X ago" no longer shows how much time to live the post has, so it'd need to be shown separately.

I'll third an anonymous "flag for review" feature. It seems the safest and most judicious approach to the problem. The one problem I can see with it is a pic could then spawn an edit war if two janitor/mod/admins disagreed on a post's worthiness.

The effect of this could also swamp the mod queue if the same contentious posts got stuck in the queue time and time again being re-approved each time. It would probably be best to limit the number of times a given image could cycle through the queue.