OK, hear me out.

We have the per-user upload limit for a reason, right? It's to limit the damage done by one malicious individual who creates an account to start spamming IRL gore or what have you. The upload limit is a sign of trust: as more of your posts get approved, you gain the right to upload more at once. Conversely, the more of your posts get deleted, the more trust you lose, and if you post nothing but garbage, eventually the privilege is taken away altogether.

Right now the only tool the site has for curtailing other forms of vandalism is a crude global count. I mention tag edits because they've been on my mind after looking at the (former) results of no_pants no_underwear no_bottoms and groaning, but that's hardly the only large-scale vandalism possible. Hands up who remembers that day most pool names got changed to racial slurs?

Anyway, currently (nearly) every user has a fixed cap of 150 tag edits per hour. The only way to increase it is to have your account upgraded to Privileged, and there is no way to reduce it at all. If your tags are mostly invalid or even straight-up malicious, the only enforcement mechanism is the chance that someone will notice the common factor between a number of seemingly unrelated bad tags and report your account to the admins, a process that typically needs to happen a minimum of four times before the ability to change tags (as well as do anything else on the site) is taken away.

So I was thinking: what if the tag edit limit was dynamic, like the upload limit? The more valid tags you add, the more you can add per hour, and if someone deletes a tag you added, that goes towards your penalty count. Vandalism potential would be significantly reduced overnight, and even better, the much bigger problem of users not bothering to learn how tags work and doing untold amounts of damage without getting caught would fix itself given enough time.

This would also help significantly in the process of identifying good users, since the total number of posts edited says nothing about whether those edits were any good, and it's also possible to inflate it by adding tags to a post one at a time.

There is one major implementation hurdle that I can see: what to do about aliases? For what I assume are performance reasons, a new alias doesn't generate a post edit on its own, and instead credits the change to whoever edits the post next. Should this count towards that person's "good edits"? Should the person who added the aliased tag be penalised? Should penalties be conditional, like for replacements, since some aliases are more obvious than others? If they should be penalised, how to apply all the penalties at once and then prevent any of the penalties being double-counted in the future as posts are edited? If they shouldn't, how to prevent penalties being applied anyway for the same reason?

Bonus thought: invalid tags should count for extra karma, in both directions. Users are rewarded for patrolling invalid tags and replacing them with good ones (simply deleting a tag brings no rewards), and penalties are extra harsh if you add an invalid tag. Unlike uploads, rewards and penalties should be symmetrical, to encourage fixing your own mistakes.

I like this idea. We would get rid of the solo_focus solo, anthro humanoid, male/male solo taggers pretty quick. It would have to be stricter than the upload limit formula, though.

If one cleans up their own mistakes, it shouldn't be penalized.

Then there are the aliases that shouldn't be counted as well. Because it doesn't necessary mean the tag was wrong.

What if a tag is removed, and added again later? Will the penalty for the initial editor be removed?
If not, it could help to reduce the amount of big tag wars. But it wouldn't be fair, if the initial editor did a correct edit. +I think this would end up in a big API spam.

wat8548 said:
Should penalties be conditional, like for replacements, since some aliases are more obvious than others?

How would you implement this? Like a checkbox, one can click, if they think the other one shouldn't be penalized?

I like the idea, but I don't think the "effort to implement/use" balance is good.

I wish there's an entire fucking learning course for tagging here, i'm still learning myself, but I was one of those "anthro humanoid" "solo_focus solo" guys...

I think i'm alot better now, but i don't usually fix tags except for my own posts. And I usually do my research on tags, but as a yugioh player, reading is a bane for me.

I imagine this would be extremely difficult and error-prone to implement. Constantly tracking when a post's tags are edited to update a user's edit limit if they were the last to add that tag to that post seems like it'd be a heavy burden on the system, along with distinguishing illegitimate/careless tagging vs accidental mistags/misspellings (it's surprising how many times I've mistagged eyes_closed as eyes_claws, often because I also add toe_claws and finger_claws to posts and my brain gets ahead of itself, and its easy to overlook when checking over multiple tags before submitting; people with dyslexia or for whom English isn't their native language would probably suffer more). Not to mention honest disagreement on a tag's applicability, which will often be added and removed a couple times before reporting it to get an admin to weigh in.

Answer to most of the above replies is in the last paragraph: deleting tags doesn't give the deleter bonuses, but gives a negative to the last person to add that tag instead. The bonus for adding one tag would be the exact same as the penalty for having one of your tags deleted, so fixing your own tags is a net neutral action.

Although I have just thought of another potential sticking point: a lot of the biggest tagging offenders only upload posts, and never edit them. Even if every upload contains errors, it's rare for an entire original set of tags to be replaced, so they would still receive a net positive tagging score. And this problem couldn't be fixed without incentivising the already common problem of upload sniping (uploading with minimal tags and editing after the fact).

Maybe the scoring could be different somehow for tags added by the original uploader? If a different user removes a tag that was added by the uploader, the uploader receives a bigger penalty than if they remove it themselves. Although the scoring symmetry would have to be maintained if the uploader (EDIT: or anyone) adds it back, to prevent targeted tag nuking from being a viable way to ruin other users' accounts.

I don't think the burden would be that heavy considering it only has to look through the post history when tags are changed, a comparatively rare operation. The big problem would of course be implementing it retroactively for the massive changes history.

wat8548 said:
Although I have just thought of another potential sticking point: a lot of the biggest tagging offenders only upload posts, and never edit them. Even if every upload contains errors, it's rare for an entire original set of tags to be replaced, so they would still receive a net positive tagging score.

This would also be a problem with people who aren't the original uploader, but excessively tag every minute detail of an image with many redundant and unnecessary tags, which ends up also hiding mistags and people avoid editing such a massive tag list since it's difficult to see what's erroneous or missing. So people who overtag and mix in several bad/inapplicable tags will still end up net positive.

watsit said:
This would also be a problem with people who aren't the original uploader, but excessively tag every minute detail of an image with many redundant and unnecessary tags, which ends up also hiding mistags and people avoid editing such a massive tag list since it's difficult to see what's erroneous or missing. So people who overtag and mix in several bad/inapplicable tags will still end up net positive.

I don't think we want to discourage "overtagging", and one massive bad edit is much easier to spot and report than many small bad edits.

"every user has a fixed cap of 150 tag edits per hour" Is this per tag or per submission of a group of tags? Because if it's individual tags, then I've come close or went over that cap. Example: One character has anatomic features but you can't see it in every picture, so I searched for that character without the tag, then went through and verified each one before either editing, then moving to next in search results using arrow button.

watsit said:
This would also be a problem with people who aren't the original uploader, but excessively tag every minute detail of an image with many redundant and unnecessary tags, which ends up also hiding mistags and people avoid editing such a massive tag list since it's difficult to see what's erroneous or missing. So people who overtag and mix in several bad/inapplicable tags will still end up net positive.

Guilty!
Well, kind of, but metadata is important for finding pics. There is one sight where you can identify a manga or anime by the color of their hair, etc.
I'm still needing to do some cleanup of some object_insertion-related tags. :/ I'll get to it next time I go on a tagging binge.

One thing that is interesting is that people that follow the wiki extensively will have better tagging accuracy (after all, it's basically how we define them more clearly than just looking at random examples), right? Tried to make some things mentioned in forums clearer when editing it, so that next person didn't need to ask. Ideally, definitions should improve after every time someone doesn't find the answer in the wiki.

alphamule said:
"every user has a fixed cap of 150 tag edits per hour" Is this per tag or per submission of a group of tags? Because if it's individual tags, then I've come close or went over that cap.

I reached the cap for the first time ever recently. Can't remember what I was doing at the time, but I had to stop it and come back later.

Personally I think 150 is way too high and the default value should be more like 20. Even honest newbie taggers should definitely spend at least three minutes reading the wiki per edit anyway.

wat8548 said:
...the default value should be more like 20...

This sounds terrible. How are we supposed to clean up all the false tagged stuff. If one knows which tags can easily be removed, you are past 20 in less than a minute. (solo_focus solo, anthro humanoid, male/male solo as mentioned before) Or if someone uses https://tagme.dev/ there is a low risk of doing big tag vandalism with a high rate of edits. Before I had privileged status, I hit the 150 mark a couple of times a day.

dubsthefox said:
This sounds terrible. How are we supposed to clean up all the false tagged stuff.

The same way you upload all the missing art, by slowly gaining trust over time.

dubsthefox said:
If one knows which tags can easily be removed, you are past 20 in less than a minute.

"If" doing a whole ton of work here.

In general, I agree with this. You should be able to raise your edit limit without having to gain other privileges. Maybe to simplify the system, you get points for n edits, but for them to apply, some higher ranked user has to confirm that they're good edits?

Or have a hidden tag editor flag? Request it or it could be automatically applied for old accounts with more than say, 1000 tags without a ton of red flags for tagging abuse? But then people will game it - which is probably why it should be hidden. Can't game what isn't easy to measure. XD

alphamule said:
Or have a hidden tag editor flag? Request it or it could be automatically applied for old accounts with more than say, 1000 tags without a ton of red flags for tagging abuse? But then people will game it - which is probably why it should be hidden. Can't game what isn't easy to measure. XD

This just sounds like the current green bar system with extra steps. Besides which, a major reason I want this is so that it's also possible to reduce the limit to zero given enough bad edits.

kora_viridian said:
What happens in this scenario?

1. Somebody creates a new account.

2. They find an older post that may be a bit under-tagged - say, 15 to 20 tags total. (It may not have to be under-tagged, but it makes the next step easier.)

3. They add a bunch of irrelevant-to-the-image, but non-conflicting, tags to the post. Like, if the post was already tagged "group", dump about 10 different species tags on there. If has the "group" tag and an explicit rating, dump all of the "X_penetrating_Y" tags on there. If the image isn't already tagged "inside", add "outside", "sunlight", and "clouds", even if they don't fit the image. As long as the tags they add are "known" tags, and not obviously in conflict with each other - like trying to tag "solo" and "duo" at the same time - I think this would increase their tagging limit.

4. The new user then uses their newly-raised tag limit to do mayhem on other posts.

Much like post uploads, the limit increase would have to be significantly slower than the number of edits made (probably even lower than the 10% upload rate, considering edits are easier). So that would limit the damage significantly.

It would take a long time, and a lot of effort, to reach the point where it is possible to do more under-the-radar vandalism than today's 150 edits per hour. There's absolutely nothing stopping a new user from doing exactly what you describe under the present system, and as long as they target older, less popular posts, they could probably hit thousands of posts before someone notices.

But eventually, someone will notice, and an edit which adds many invalid tags will naturally draw attention to the rest of that user's activity as soon as it is spotted. At this point, it's game over - the edits can be mass-reverted instantly (the site already provides an easy way to do this) and because, same as with post deletions, tag deletions should decrease your score by more than additions increase it, the user's edit limit would be nuked into the ground. The main advantage, aside from the total amount of damage probably being lower, is that there's no need for admins to get involved.

(This does raise the possibility of malicious targeted deletions, but in that case it would be appropriate for the admins to get involved anyway so they can undo the incorrect changes for you at the same time.)

This is, of course, an extreme scenario, and not one I'm convinced will ever occur (based on the fact it has never occurred despite the present system not having any defensive advantage against it). I was trying to come up with a way to seamlessly guide low-activity newbies and willfully ignorant users in the right direction, as a side-effect of maintenance which needs doing anyway, without having to write a thousand DMs to the effect of "Read the damn wiki" every week.