Hi everyone! Apologies for the delay on a response - I've finally had time to organize and sort things out, so without further ado: an update!
Summary
Since this post itself is a bit long, below is a very brief summary of the main points:
Forum-wide ban on GenAI images
This is a change in our initial stance, as we've received more feedback and found ways to address our initial concerns with a forum-wide ban. The below were the main points of concern, both raised in this thread and for our initial decision (in no particular order):
For (1), we will be creating a thread dedicated to helping with avatars and signatures where all members can contribute and assist! At present, The Signature Guide is the best existing thread for this purpose, but a more dedicated thread with a clear purpose for members and staff to help each other will be made in the near future for this purpose*! We're hoping this will help minimize any frustrations that might arise when trying to find an avatar/signature that fits within guidelines. This thread will also hold a list of alternative image resources (see the bottom of this post) which members can look to for assets!
* Likely to be made within a week or two, when Mistreil gets a chance!
Moderation workload and ambiguous cases
For (2) and (3): we have decided to operate largely on a report-basis. While staff will try to keep an eye out as well, many of the staff are not currently confident in their ability to identify GenAI images and thus were initially hesitant to make calls in situations that may be more ambiguous. However, operating on a report-basis alleviates most of these concerns.
Additionally, members are expected to (or will be asked to) explain why they believe an image is AI generated. While this does unfortunately put the burden on the reporting members, we're hoping that this will both help less experienced moderators learn to identify GenAI images faster (and thus act faster as opposed to waiting for another staff member to be available to help), as well as reducing false reports.
There were many concerns on the potential of artists being reported, as incorrectly removing an artist's work is particularly harmful. In regards to this: we will be looking at ambiguous cases more carefully and erring on the side of caution or asking for further information before acting. This means we will be taking a member's history into consideration as well: if they are known to be an artist, we will likely be slower to act as we try to reach out to them to clarify.
As a result: some reported cases may take longer to handle than others, or our judgement may err on the side of caution and not remove a suspected image if we cannot make a clear call. While this might not seem ideal, we hope this is understandable and acceptable.
Level of punishment and behaviour around reports
Initially, we felt that warnings for what would be likely accidental use of GenAI images would be too punishing and particularly frustrating for members that didn't have the skills or knowledge on how to avoid or navigate the growing amount of GenAI images. However, within this thread, it seems like a consensus that treating these similarly to signature violations seems fair. Coupling this with having resources to help members with avatars and signatures, this no longer feels too concerning.
With regards to posts containing GenAI images: treating these similarly to low-quality posts seems like the most reasonable level of moderation. We would like to discourage the creation of GenAI images, and as they likely will not contribute much to a discussion and would likely act to derail discussions, this seems fair.
We would like to note that we may make exceptions on what is permissible in terms of GenAI, though we have no firm stance yet. In particular, for games that use GenAI concept art, we aren't quite sure whether to disallow discussions on this or not. Largely: we would like discussions to be allowed for the purposes of bringing attention to the use of GenAI as this is educational and will help members make an informed decision on whether to support a game or not. In lieu of making a pre-emptive decision, we'll be keeping an eye on this if it comes up in the future, and we may handle such things on a case-by-case basis.
In addition, as there are concerns about the behaviour of members making reports: we would like members to simply make their report and avoid confrontations/judgements on members that post genAI images or have genAI avatars/signatures. Most of the time, GenAI cases are accidental or done by members without an understanding of the problems surrounding GenAI images, and so we hope members will give each other the benefit of the doubt. While we don't believe this will be too large of a problem, we would like to set this expectation explicitly!
Alternative image resources
While Google images has been the main source of images for many members, there is an increasing number of AI results which are becoming harder to navigate. As such, we will be building a list of resources that may be helpful in creating avatars or signatures that do not use GenAI.
For Google images, there are various search terms that can be added in order to avoid certain websites or terms. For example, adding
If you are okay with results that are older,
There are also additional resources at the bottom of Is This AI? - A Comprehensive Guide to Spotting AI Art - though these are more tailored towards art references, they are still helpful resources!
I will be updating the opening post of this thread in the future to better reflect these changes - but I don't quite have the time for a formal revision! So if the first post seems outdated for a bit... please be patient and I'll get to it eventually!
Summary
Since this post itself is a bit long, below is a very brief summary of the main points:
- We have revised our stance: GenAI images are not allowed on the forum (in posts, avatars, or signatures.)
- GenAI images will be handled on a report-basis, with the members making the reports responsible for explaining why they believe it is made with GenAI.
- GenAI images in avatars/signatures will be treated like signature violations.
- GenAI images in posts will be treated as low quality posts*
- In the near future, we will be making a general purpose thread for helping with avatars and signatures: this is where members can ask if an image they want to use is GenAI or not, or for help finding an avatar/signature that they could use!
- This thread will also be where our list of alternative image resources for avatars and signatures will be more permanently housed. For now, this list is at the bottom of this post!
Forum-wide ban on GenAI images
This is a change in our initial stance, as we've received more feedback and found ways to address our initial concerns with a forum-wide ban. The below were the main points of concern, both raised in this thread and for our initial decision (in no particular order):
- Member frustration in finding suitable avatars/signatures
- Moderation workload and difficulties in terms of identification of GenAI images
- Handling of ambiguous cases, especially artists being reported for GenAI images
- Expected level of punishment for GenAI image use, especially in scenarios wherein its use is potentially permissible such as reposting concept art for games that use GenAI
- Potential behaviour around reporting members using GenAI images
For (1), we will be creating a thread dedicated to helping with avatars and signatures where all members can contribute and assist! At present, The Signature Guide is the best existing thread for this purpose, but a more dedicated thread with a clear purpose for members and staff to help each other will be made in the near future for this purpose*! We're hoping this will help minimize any frustrations that might arise when trying to find an avatar/signature that fits within guidelines. This thread will also hold a list of alternative image resources (see the bottom of this post) which members can look to for assets!
* Likely to be made within a week or two, when Mistreil gets a chance!
Moderation workload and ambiguous cases
For (2) and (3): we have decided to operate largely on a report-basis. While staff will try to keep an eye out as well, many of the staff are not currently confident in their ability to identify GenAI images and thus were initially hesitant to make calls in situations that may be more ambiguous. However, operating on a report-basis alleviates most of these concerns.
Additionally, members are expected to (or will be asked to) explain why they believe an image is AI generated. While this does unfortunately put the burden on the reporting members, we're hoping that this will both help less experienced moderators learn to identify GenAI images faster (and thus act faster as opposed to waiting for another staff member to be available to help), as well as reducing false reports.
There were many concerns on the potential of artists being reported, as incorrectly removing an artist's work is particularly harmful. In regards to this: we will be looking at ambiguous cases more carefully and erring on the side of caution or asking for further information before acting. This means we will be taking a member's history into consideration as well: if they are known to be an artist, we will likely be slower to act as we try to reach out to them to clarify.
As a result: some reported cases may take longer to handle than others, or our judgement may err on the side of caution and not remove a suspected image if we cannot make a clear call. While this might not seem ideal, we hope this is understandable and acceptable.
Level of punishment and behaviour around reports
Initially, we felt that warnings for what would be likely accidental use of GenAI images would be too punishing and particularly frustrating for members that didn't have the skills or knowledge on how to avoid or navigate the growing amount of GenAI images. However, within this thread, it seems like a consensus that treating these similarly to signature violations seems fair. Coupling this with having resources to help members with avatars and signatures, this no longer feels too concerning.
With regards to posts containing GenAI images: treating these similarly to low-quality posts seems like the most reasonable level of moderation. We would like to discourage the creation of GenAI images, and as they likely will not contribute much to a discussion and would likely act to derail discussions, this seems fair.
We would like to note that we may make exceptions on what is permissible in terms of GenAI, though we have no firm stance yet. In particular, for games that use GenAI concept art, we aren't quite sure whether to disallow discussions on this or not. Largely: we would like discussions to be allowed for the purposes of bringing attention to the use of GenAI as this is educational and will help members make an informed decision on whether to support a game or not. In lieu of making a pre-emptive decision, we'll be keeping an eye on this if it comes up in the future, and we may handle such things on a case-by-case basis.
In addition, as there are concerns about the behaviour of members making reports: we would like members to simply make their report and avoid confrontations/judgements on members that post genAI images or have genAI avatars/signatures. Most of the time, GenAI cases are accidental or done by members without an understanding of the problems surrounding GenAI images, and so we hope members will give each other the benefit of the doubt. While we don't believe this will be too large of a problem, we would like to set this expectation explicitly!
Alternative image resources
While Google images has been the main source of images for many members, there is an increasing number of AI results which are becoming harder to navigate. As such, we will be building a list of resources that may be helpful in creating avatars or signatures that do not use GenAI.
- The Spriters Resource: A collection of assets from various games.
- Picrew: A collection of image makers made by various individual artists where you can create avatars, characters, etc.
- Different picrews will have different scopes of use - please be mindful of these and give credit (e.g. linking back to the picrew used)!
- Pexels: Royalty-free stock photos, images, and videos. As these are posted by the creators themselves, no GenAI is allowed.
- Wikimedia Commons: While not completely devoid of GenAI images, many photos have sources and details and GenAI images are in their own categories.
For Google images, there are various search terms that can be added in order to avoid certain websites or terms. For example, adding
-"openai" to your search will remove all search results that include OpenAI, though it may not filter out everything. Other terms may include -"ai" -"craiyon" and the likes.If you are okay with results that are older,
before:2021 will display only results before 2021 (and thus before GenAI images began flooding the results.)There are also additional resources at the bottom of Is This AI? - A Comprehensive Guide to Spotting AI Art - though these are more tailored towards art references, they are still helpful resources!
I will be updating the opening post of this thread in the future to better reflect these changes - but I don't quite have the time for a formal revision! So if the first post seems outdated for a bit... please be patient and I'll get to it eventually!
