About a year ago, Danbooru rolled out a system to obscure the uploader of a post, replacing it with an idea called Top Tagger. The top tagger is whoever adds the most tags to a post. This was an attempt to discourage upload sniping, since credit would be rewarded not to the person who uploads the post first, but whoever supplies the most information. At the time of release it was estimated that about 5% of posts would change attribution.
Running the analytics now, it's closer to a 2% change of attribution.
Perhaps uploaders are more incentivized to provide more tags on the initial upload. There's definitely been a trend towards that, but it's not clear there's any correlation to the top tagger implementation versus just a general increase in tag usage.
So from my perspective, the experiment was a failure. I plan on rolling back most of the changes to simplify the system.
I wanted to create this topic as a forum for discussing thoughts, ideas, and feedback.
When you say "most changes", which are the ones you won't roll back?
Also, I seriously think the only good way to stop upload sniping is to either enforce a minimum tag amount on upload (this is hard to make a good number) or remove uploader name completely (for builder and lower) and/or score from posts.
I don't think the idea to do this for discouraging sniping was effective. I do contend however that the system could have other useful applications. For instance, users without the unrestricted uploading permission could be granted an extra upload slot for each post they become the top tagger of, encouraging new users to become more thorough taggers while reducing under-tagging. I think that competitive pressure of many users looking to tag new uploads would result in an increase of the volume of initial tags and promptness by which posts are tagged by their original uploaders.
I don't have numbers, but I think most of my top tagger gets came incidentally while gardening either old posts, or posts by newer users. (Old posts don't usually have the "color_object" tags, and many new users have a limited tag vocabulary and miss a bunch of important tags.) I only remember taking top tagger from a sniper a handful of times, and at least one of those involved a tagging war to see who could add the largest number of filler tags to the post (not exactly beneficial).
The top tagger system caused me to take longer to tag posts before uploading them, instead of uploading with "sufficient" tags and continuing to add more after it was done. Sometimes this meant losing an upload to someone who was still using the "sufficient" method. In those cases, I wouldn't be able to take top tagger from them because they hadn't missed quite enough tags for me to numerically outdo them, even if I'd used more meaningful tags. (For example, bra_lift or crop_top are more meaningful than collarbone or eyebrows_visible_through_hair.) issue #3130 could've also helped, by adding my tags to the post right away before the other user could add more themselves -- but as long as the post actually gets tagged, it probably doesn't really matter if I was going to do it a millisecond before they did.
IMO the kind of sniping that really hurts the site (aside from hurt feelings, which absolutely hurt the site, but do so in ways I can't really speak on) is when people upload with minimum tags and don't keep tagging afterwards. And since I don't upload kancolle or fate or touhou, I haven't really experienced that kind of sniping (if it still exists), so I don't know how top tagger affected it.
[…] This was an attempt to discourage upload sniping, […]
Perhaps uploaders are more incentivized to provide more tags on the initial upload. There's definitely been a trend towards that, but it's not clear there's any correlation to the top tagger implementation versus just a general increase in tag usage.
Running the analytics now, it's closer to a 2% change of attribution.
So from my perspective, the experiment was a failure.
I’m not sure if it’s statistically valid to try to deduct upload sniping from attribution changes. Attribution changes would show upload sniping with undertagging, but they won’t show upload sniping when the uploader finishes up tagging after upload but before someone else does it. Users, like fossilnix above, take more time to tag properly, so if the goal had been to reduce undertagging, it would’ve been a success, IMO.
Now since we can’t really measure upload sniping, what about complaints about upload sniping? I don’t remember seeing any, at least on the forum, so I think that’s a win?
Introducing top taggers looks like pretty successful change to me.
Now since we can’t really measure upload sniping, what about complaints about upload sniping? I don’t remember seeing any, at least on the forum, so I think that’s a win?
I know for a fact that there is still a couple of users uploading with minimal (<10) tags even with Top Tagger active so it hasn't completely eliminated it, maybe just less people complaining since there is a way to turn that against them now.
IMO the kind of sniping that really hurts the site (aside from hurt feelings, which absolutely hurt the site, but do so in ways I can't really speak on) is when people upload with minimum tags and don't keep tagging afterwards.
If the goal is to get rid of people uploading images with an minimal amount of tags just to be the FIIRST!1! one and then never touch those posts again, wouldn't it make sense to keep track of an user's tagcount history and use that number to lower their Upload Limit accordingly? Or change the recovery speed of the Upload Limit with some kind of multiplier to slow down their upload rate?
You could even go as far as lowering it to 0 and effectively blocking that user from uploading for X amount of time. If someone triggers that level, you could make DanbooruBot send an automated message telling that user to improve their tagging.
For example: Keep track of an user's last 10 posts. If 5 of them have less than 10 tags, you could half the Upload Limit and send the warning message.
Something like this should stimulate people who unwittingly use not enough tags to improve, should pretty much block the repeating offenders from uploading at all, and should thereby give some breathing room to the more dedicated taggers to upload well tagged posts.
That might cause users to add nonsense tags to their uploads, just to meet the minimum quota, which will be a lot harder to fix for other users than posts with few tags.
Wasn’t there a colored indicator somewhere that told you how well you’re doing on the tagging? If you want to teach the ignorant users, that should be enough. Anyone who still doesn’t tag properly simply doesn’t care.
I think most of the Top Tagger switches were not due to image sniping but because there are or used to be users that are adding only a few tags to their uploads. It's really easy to get Top Tagger spot from a post with only 20 gentags if you are willing to tag an image up.
What seems actually better are the emotes that show up when you tag an image. It's frowning until you have 25 tags or more. It would be nice to see what this change caused to undertagging.
I certainly wouldn't mind being able to see both the top tagger and uploader for posts. Both are useful/interesting tidbits of information in their own right, perhaps even more so when right next to each other. For me at least, it would be nice to be able to easily see both of these on uploads.
That might cause users to add nonsense tags to their uploads, just to meet the minimum quota, which will be a lot harder to fix for other users than posts with few tags.
Wasn’t there a colored indicator somewhere that told you how well you’re doing on the tagging? If you want to teach the ignorant users, that should be enough. Anyone who still doesn’t tag properly simply doesn’t care.
If a user is willing to go through all that trouble to bypass the tag limit, then they either have the energy to properly tag it or they will be dealt with since its putting the effort into something disruptive and without good-faith.
To me though the whole upload sniping issue is more to do with the emotional desire to be first. In that case, I think anything that inconveniences the user could be used to have them follow the path of least resistance in a way that's beneficial. I don't know if anything more abstract will solve this particular issue.
I never did see the concern for upload sniping beyond pettiness and spite. I also never noticed anyone leaving Danbooru because they got sniped one too many times. I'm fine with the system either way.
I also never noticed anyone leaving Danbooru because they got sniped one too many times.
I've seen and dealt with a few cases, generally in the comments of a random post, hence why they aren't so noticeable. There's been a few angry outbursts on Discord too but I don't think any resulted in quitting.
I think it'd be entertaining to implement a delay. Warn uploaders that once they post, they won't be able to edit the tags on it for X number of minutes and that another user may come along and steal the limelight. If you want to get really fancy, queue any edits they make to be submitted only after X number of minutes. Whether that complexity is worth it or just funny to contemplate I can't say by myself.
I think it'd be entertaining to implement a delay. Warn uploaders that once they post, they won't be able to edit the tags on it for X number of minutes and that another user may come along and steal the limelight. If you want to get really fancy, queue any edits they make to be submitted only after X number of minutes. Whether that complexity is worth it or just funny to contemplate I can't say by myself.
This has been brought up before and it would be a very annoying thing for people like me who tend to always remember a few more tags after submitting, even if it's not a fresh Pixiv upload, same reason non-merging tag changes was annoying.
+1 to removing top tagger. I don't see much evidence that it's made a difference in uploader behavior.
It's hard to accurately measure the impact it's had, especially due to various issues that led to the top tagger being determined incorrectly in certain cases. For this reason, I don't think it would be a good idea to rely on it for other purposes.
Frankly, I think the whole issue of sniping is overblown. Sniping, in the sense of people uploading with minimal tags, isn't really why people get frustrated. People get frustrated because another user beat them to a post, period. It doesn't really matter whether the post is well tagged or not. It's still frustrating to be beaten, either way.
If mintagging didn't exist, and all uploads were magically fully tagged, uploaders would still get frustrated at being sniped. They would just place the blame on other things. Instead of blaming others for mintagging, they would blame others for 1upping, or for using tools and scripts to gain an unfair advantage, or just generally accusing others of being no-life NEETs who sit on Pixiv all day.
I think there's not much that can be done here, because what people ultimately want - which is, the ability to upload from popular artists without competition from others - is not something they can have.
...I think it'd be entertaining to implement a delay. Warn uploaders that once they post, they won't be able to edit the tags on it for X number of minutes and that another user may come along and steal the limelight. If you want to get really fancy, queue any edits they make to be submitted only after X number of minutes. Whether that complexity is worth it or just funny to contemplate I can't say by myself.
On top of being annoying for all the reasons mentioned by other users, this will also get in the way of translators who translate immediately after uploading. (or encourage 'false' tagging like preemptively tagging a post with commentary or translated to get around the issue.)
(On the uploader side of things, personally I often run into situations where I see something taggable but don't quite remember the name of the tag [usually this is a pose/expression or a character-specific clothing/hairstyle], so I tag as best as I can when uploading, before consulting the wiki and filling in the missing tags. Sometimes I also have to fix typos or misclicked autocompletes, or I notice new tags as I'm translating a post [like a pun or reference that I didn't get when just skimming]. Having an arbitrary delay post-upload will be very frustrating for me.)
Alright, I don't think we need a fifth testimony that uploaders are prone to mistakes and desire extra time to garden their uploads. The lucky few who have to deal with the occasional hothead getting upset that their uploads keep getting stolen is outweighed by the inconveniencing of the masses.
Speaking just for myself, I've sometimes held back on fully tagging posts because I didn't want to steal credit, because I like it when people upload my favorite copyrights and want to encourage it, which includes letting them have the credit.
Additionally, there are posts where I don't want to steal credit because they're not the kind of posts I would upload, and I don't really want my name attributed to them in such a way.
Both of the above sometimes require a bit of finagling to find that sweet spot where I can add a lot of tags bot not take credit. Instead, I'd rather not have to worry about such and just tag as I normally do.
