AI-detector scores in FOSS game jam or devlog moderation

ethanjamescolez

Question for free/open-source game jams and project communities.

If a FOSS game project, jam, or community showcase allows written material such as design notes, devlogs, lore, tutorials, or README updates, how would you handle AI-detector scores in moderation?

The problem I see is that a detector report can look like evidence, but it is still probabilistic. In a small free-game community, a false accusation could damage trust, while ignoring obvious low-effort generated submissions can also hurt the project.

The boundary I would use:

do not treat a detector score as proof of authorship or misconduct
expect false positives and false negatives
ask for source context such as commit history, drafts, screenshots, or design rationale
avoid public accusations in project threads
do not upload someone else's private drafts or school/work files into a detector without permission
keep any review process focused on project quality, license clarity, and contributor norms
let maintainers make a manual judgment rather than outsourcing the decision to a percentage

For free game projects, would you put this in jam rules, contribution guidelines, moderator notes, or project showcase expectations?

No product link is needed. I am asking about FOSS game community moderation and contributor trust, not recommending a detector.

TPW

I don't manage no jam, but I have a little page on Itch where devs can add their projects. If I see someone who obviously uses AI, I delete the game (and sometimes the uploader) from the page.

I wouldn't bother to use any scanner - if someone can get his slob past me, so be it.

dulsi

Like TPW, I probably wouldn't use a scanner. Include in the rules that AI generated content is not allowed. If something is questionable, I might run it through an AI detector but that would be more to confirm my suspicion. How would I ask someone to prove no AI use? I don't know.

At this point I think it is probably not realistic to try to ban AI generated code. While I'm not pleased at the training from open source repos, I think the world has changed. Much of coding work is not really creative. I would prefer not to see AI generation in story writing and artwork.

(I was playing with Hobby account with Cursor. One request used all my monthly quota. Back to regular coding.)

anaseto

In a small foss community, I'd avoid introducing any kind of non-human bias in interactions and moderation. Just stating things and assuming honesty feels the best to me, because at small scale users can still do their own manual filtering. I'm not a moderator of any community, so I cannot really say, but my guess is if I were one, I wouldn't like using a "detector", as it sounds like fighting fire with fire, probabilistically biasing moderation to fight against probabilistic generation of uninteresting content.

In any growing community, there's always a small minority that behaves badly without respecting the community, and it's a pity, but producing slop is probably not the worst thing they can do anyway.

Modanung

anaseto [...] producing slop is probably not the worst thing they can do anyway.

Depending on whether the definition of FOSS within the context of the jam includes having only free culture as assets, the use of AI may matter due to rights issues.

anaseto

Modanung Yeah, sure, that's an issue. I'd say it's not the worst thing from a moderation point of view (for a jam or community forum), but it's indeed problematic when contributing to a project or depending on it, to avoid wasting time and effort on code with unclear rights.

Modanung

anaseto I'd say it's not the worst thing from a moderation point of view (for a jam or community forum)

Well, it depends on the moderators' stance in regard to media within a free software context. If they find that nonfree media conflicts with three of the four essential freedoms - modify, redistribute and distribute modifications - then use of AI is very much one of "the worst things from a moderation point of view".

TPW

Modanung Within the reality of moderating things online, there are indeed many things that are worse - illegal material, suicide threats, and toxic users are much more of a challenge, to cite just a few examples. AI stuff can be deleted on sight, and things are fine.

Modanung

TPW You forgot threats of nuclear annihilation. 🙄
And those - from a moderation perspective - might actually not be as bad, because they're blatant and rare. The things you mentioned are only worse in terms of commonly agreed seriousness, not in likelihood or inconspicuousness, not in terms of moderation difficulty. A wave of hard to notice things that requires moderation is (accumulatively) worse than a single unlikely in-your-face occurrence.

AI stuff can be deleted on sight, and things are fine.

People increasingly reach for AI because the results are getting better. It doesn't stand out as much as it used to. And as recently demonstrated on these very forums, it can sometimes be hard to get a straight answer concerning its use. "So just use an AI image detector" you say? It doesn't take more than applying a filter or two in GIMP to fool those.

You also didn't mention venomous slithering sociopathic trolls. The type that poisons the well and weaponizes the term "toxic". Those can be pretty hard to spot.

Modanung

TPW suicide threats

Also I think it's a bit sick and twisted to misconstrue openness about suicidal ideations as "threats". Life is a right, and so is bodily autonomy. When someone is suffering (i.e. a victim) to the point of seeing a brief process of dying as the more bearable alternative, you don't add insult to injury and shut/lock them up.
Putting your own grief/stress regarding someone else's (consideration of) suicide above reasons to commit such an act, is egocentric, unsympathetic and petty.

TPW

Modanung than a single unlikely in-your-face occurrence.

Unfortunately, what I said aren't examples for rare problems - and if you give a damn, they are often not easy to deal with.

I agree that living is a right, not an obligation - but I also think it is okay to run platforms that are limited to specific topics, and aren't meant to discuss individual ones. I'm also free to run a board that is not a suicide support-group. And this is not due to ill intend: Being confronted with a notice of intend of somebody who says they want to end their life is a huge amount of stress for all the people who might not be able to deal with it, especially since the room for intervention in digital spaces is marginal. The fact that you don't want to just "shut or lock" the people in question here is what makes it difficult - and the stakes are much higher than some AI assets that might end up in a game.

But ofc you might be able to do handle such things different - mind to drop an URL, especially to public threads in a space you are responsible for, so that I can learn how you deal with such topics without limiting the freedom of the affected people?

Modanung

Terms like "huge amount of stress for all the people" and "intervention" imply only "giving a damn" about keeping up appearances and offsetting responsibility, not community. It ends in "we didn't see it coming" and "we did our best" (by having the person strapped to a gurney, thrown in a cell, stuffed full of pills, subjected to forced injections, monitored and interrogated for months while expecting improvement because "experts" did it) ergo suppressing symptoms, not saving lives.

It is not a moderation issue, but rather one of social cohesion.