Need a way to Avoid report system being abused
complete
r
rerorerorero
I am not mean to attack sb.
But look at the "Guardians Leaderboard", the first one has "1.5k+" shots while "800+" misses, which is about 7 times of the second, and that means lots of models were reported wrongly.
We all know that it is important to avoid child porn but isn't someone too sensitive? Here, https://civitai.canny.io/feature-requests/p/report-users-warningban-system，two images are nothing to do with child porn but still get reported.
Report abusing will not only make moderaters tired and feel their time wasted, but also hurt creators' passion and enthusiasm. Creators should not be worried about things like "is the character in my image looks mature enough?" "I should add huge breasts so that it won't be reported and get unpublished".
This post was marked as
complete
n
natewilliams
I also see removal of uploads that are actually great facial expressions, most recent one had a bad title something about b**** crying in pain or something like that. The expressions it produced based on sample images were a very realistic crying face, not necessarily one in pain, and even so, highly applicable to non-NSFW situations. Grief from the death of a loved one, great sadness, etc.
Why couldn't this just have been a forced title change? The LoRA was otherwise really good.
N
Nekos4Lyfe
natewilliams: Yes. My guess the name choice caused a stir among some users. Hopefully the poster will have the chance to upload the LoRa under a new name.
I did manage to grab it though. Its saved under a new name here: https://huggingface.co/Nekos4Lyfe/beach_breaker_v1/tree/main
Overall , I think Civitai needs better filtering and possibly some better TOS guidelines regarding LoRa names.
Do you think it would be a good idea to add two checkboxes in addition to "Mature", maybe?
The first one could be called "Controversial" and will cover models/posts that could be interpreted as political, chauvinistic , insulting or otherwise divisive.
Second checkbox could be called "Extreme" and will cover unpalatable models/posts that a majority of users would not want to see in their feed.
Both "Extreme" and "Controversial" content will be hidden from the user when starting a new account. The user will have to manually turn off these filters in their settings tab, in order to see such content.
That ought to solve a lot of the mass reporting, I think. But I'm open to suggestions.
n
natewilliams
Nekos4Lyfe: More filtering options seems like a good first step. The only thing I can personally think of that should be a definite removal is actual sexualizing of minors. A character looking young is otherwise applicable to so many scenarios that are perfectly harmless and SFW.
Take a scene with a bunch of kids/teens playing pirates, one image has them dressed up as pirates using clothes that are clearly their parents clothes, oversized jackets, a scarf wrapped around the head, holding a spatula, etc., then you generate a second scene with the same kids/teens in full fantasy pirate setting, on a pirate ship with fitting pirate clothes.
This kind of content is harder to do when they remove anything that they consider "loli" that aren't even sexual in nature. It's stifling creativity just like Adobe and other corporate solutions for AI generative art overly censors. A character LoRA can be very useful in keeping consistency between the two images. The same scene isn't as adorable with grownups, although thinking about it, adding two more images to the series with the same kids/teens as grown up adults still playing pirates could be a fun idea.
Anyways, if it's straight up or borderline CP, then yes, remove the content and the user. If it's just a young looking character, there's absolutely nothing wrong with that. And if the model creator has uploaded questionable images, e.g., sexualized poses, clothing options, remove those images and warn them.
I think mostly everything else goes. In example, I don't see why choking poses is necessary to censor. Again, can be used outside of NSFW situations, even in comedy-style imagery (think Homer and Bart Simpson "Why you little!").
w
warrensky
Very agree with this, I mean.. come on.... there's some one who makes 1300 report and almost 800 of them are missed report, if that's not an abuse then I don't know what it's called
l
lotte_peisch764
In order for this to be possible, CivitAI is going to have to define what is allowed and what is not allowed A LOT more clearly. Some of y'all are REALLY confused as to what a child looks like.
c
ccd355_
can not agree more. balance between creators and model raters is important especially for websites like civitai. but a distributed rating system does not work if users are not forced to vote/judge, which unfortunately a common pratice in many source sharing websites like civitai. so a social committe is needed to address such dilemma.
r
rerorerorero
For Example:
if someone's reports have missed 3 or more times continuously, he or she will be warned and can not report in a period of time, that is to say, "Cool Down" or "CD".
If someone's “success rate” is keeping low, his or her report will be regarded as "half report", which means only when there are more people reporting the same image will his or her report become a valid one.
And the Guardians Leadboard also needs to be improved.
T
TheOnlyHolyMoly
rerorerorero: I think this is an absolutely amazing idea. Keeping discipline to "think before reporting" makes absolute sense. Someone that is constantly right has no negative downside of such mechanism, only the abusive reporter type. Also it's bringing more clarity why certain reports are not followed-up too intensely if the reporter has a record of false-positve reports.