/tg/Station 13

I'm seeing this pattern pop up more and more, and in admin channels we have had discussions about it, but it's nigh time I just come out and talk about it formally getting added to the FNR rules. The pattern is that people are making lazier appeals where they don't respond to the banning admin and just paste some text generated by ChatGPT to tell the admin "what they want to hear". So far, it's mostly gotten people's appeals killed for trying it, but I'd rather clear any confusion about AI's purpose in an appeal/complaint

So anyways, here's my suggestion: If you use AI to write anything in a ban appeal or admin complaint, you have to make it clear what parts of the appeal/complaint were AI generated.

>why not ban AI entirely, if it's being used in bad faith?
Because it can help some people who are not fluent in english appeal.

Wholeheartedly agree. This is coming up enough that we should probably codify a rule about whether or not it's allowed and I absolutely agree with Arm's opinion here.

I don't think it's necessary. This is a non-problem. If it's obvious, it's obvious and the headmins can do whatever they like with it. If it's not obvious, then it doesn't matter. If it's borderline, then it still doesn't matter.

I believe our admin team should dedicate less headspace to appeal microaggressions, and more headspace to resolving the appeals based on the merits of what is said.

I think it's a simple step to make clear that it's being used in good faith and not just to give automated responses to admins. Furthermore, I'd rather we be perfectly clear on what is AI because admins are using online "AI checkers". These checkers have a hilarious false positive and false negative rate. Removing from the equation having to gauge how real an AI post is or not by just making AI posts clearly marked is again, harmless

I think the fact we're almost prompting players that they can use AI to help them with their ban appeals will do more harm in encouraging AI use than staying silent on the matter would do.

I'm also against exhaustively listing every single minor inconvenience that may ever grace the admin team into places nobody is going to read. When it comes to rules, brevity is the soul of wit. The less we write, the more likely they will be read and the less excuse people have for not reading them. And we know players don't read the rules because almost every single note appeal someone has to link the note appeal rules.

I put forward that this just isn't necessary.

I just don't think it's necessary because I don't think we need a rule governing something I think is ostensibly irrelevant. I, personally, don't really care how the appeal was written. Whether it was written by an AI, ghostwritten by BONERMASTER or partially written by a friend. What I care about is the substance of the appeal.

Does it have merit? Does it seem genuine? Is the player like to fuck up again?

How to handle appeals is highly discretionary, unique to each admin, based on all their experiences and applying a bit of classic common sense.

We can genuinely spend all our time legislating against every last thing anyone could ever think of, or we can simply apply some common sense with weird edge cases and keep the appeal rules only as long as is strictly necessary to ensure they actually get read.

Eh, fair enough. I'm okay with spirit of rules stuff. I'll wait and see more AI stuff dribble in as the crazy wow new thing grows, but I have my own suspicions I'll find myself returning here thinking "Yeah, it went how I thought it would"

I was going to lock this, I'll leave it open so people can leave their opinions

Timberpoes wrote: ↑Wed Apr 19, 2023 9:58 pm What I care about is the substance of the appeal.

Does it have merit? Does it seem genuine? Is the player like to fuck up again?

But can we really call a player genuine if they're having an AI write their appeal for them?

I don't see anything less genuine than "I can't be assed to write this appeal so I'm going to have someone else do it for me".

Is there really a difference between asking ChatGPT to write you a formal apology letter and submitting it with your appeal and writing a formal apology letter by hand and submitting it with your appeal?

They're both meaningless but not having a formal apology with your appeal makes the admin mad anyways.

This is the future, if you think its good enough or not, and soon there will be chrome addons that compete entire sentences at a time that will become entirely normalized. We are only seeing its fledgling form. This isn’t something we can stop. Trying to ban it is pointless. Trying to police it is perhaps more realistic, but ultimately, what will be achieved?

iamgoofball wrote: ↑Wed Apr 19, 2023 11:19 pm Is there really a difference between asking ChatGPT to write you a formal apology letter and submitting it with your appeal and writing a formal apology letter by hand and submitting it with your appeal?

They're both meaningless but not having a formal apology with your appeal makes the admin mad anyways.

There's a substantial difference both in the amount of effort put in and the believability of the appeal.
If someone writes a post for the forum, there's a chance that it's trustworthy.
That chance plummets when they use AI for that post. They didn't write those words. They wrote the prompt that made a program come up with the exact words the admins would want to hear. It's the equivalent of typing up a fake apology that you don't mean, but you want to play funny spaceman game, so you just say it so you can get unbanned.
I don't think there's value in allowing it in FNR, even for the purpose Armhulenn outlines in the OP.

This isn't even me hating AI. It's me not liking the idea of people posting empty garbage when they should be either accepting that they were wrong in the eyes of the rules and responding appropriately to that, or disputing their ban and having an open discussion with the banning admin about it. There's no place for being disingenuous here.

iamgoofball wrote: ↑Wed Apr 19, 2023 11:19 pm Is there really a difference between asking ChatGPT to write you a formal apology letter and submitting it with your appeal and writing a formal apology letter by hand and submitting it with your appeal?

They're both meaningless but not having a formal apology with your appeal makes the admin mad anyways.

I'd say making it clear AI was used is just a show that AI wasn't an attempt to dodge a discussion in bad faith. Admins already see AI and assume bad faith because it's being used by a lot of people in bad faith

datorangebottle wrote: ↑Thu Apr 20, 2023 1:58 am I don't think there's value in allowing it in FNR, even for the purpose Armhulenn outlines in the OP.

I have a big soft spot for people who aren't great at english and could really use AI to help, as long as it's clear. What can I say?
Also, the point of a ban appeal for me is either:
1. point out i made a mistake (we all make em)
2. if there was no mistake, just convince me you won't do it again (hence, something was learned) and i'll lift anything early.

AI responses shake my confidence you're being genuine if you're not explicit about it, hence shaking my ability to trust people enough to unban them in an appeal.

Everyone posting that using AI to write your appeal somehow makes it less genuine confuses me.
It's like arguing whether your apology letter was written with a ball point pen or some ink and a quill. It doesn't matter. The AI is just a writing tool. There's nothing stopping somebody from writing an appeal full of empty promises about how they're going to be better or improve without AI. When someone posts an appeal, they are the ones with the final say of the contents of the appeal and the ones pressing that submit button. Whether they actually mean any of it has absolutely nothing to do with the AI. The AI just improves their writing.

SkeletalElite wrote: ↑Thu Apr 20, 2023 3:32 am Everyone posting that using AI to write your appeal somehow makes it less genuine confuses me.
It's like arguing whether your apology letter was written with a ball point pen or some ink and a quill. It doesn't matter. The AI is just a writing tool. There's nothing stopping somebody from writing an appeal full of empty promises about how they're going to be better or improve without AI. When someone posts an appeal, they are the ones with the final say of the contents of the appeal and the ones pressing that submit button. Whether they actually mean any of it has absolutely nothing to do with the AI. The AI just improves their writing.

This is like saying that AI art is exactly the same as regular art despite one taking actual effort and time and the other requiring you to tell someone else how to do it.

Definately not written by chatGPT double post time:

When it comes to the use of AI writing tools, such as ChatGPT, in writing appeals for bans, the question of whether or not the tool influences how genuine the appealer is can be a complex one.

On the one hand, it's understandable that some might be concerned that using an AI tool to write an appeal could make the appeal seem less genuine or heartfelt. After all, there's a perception that using a tool to generate language removes some of the human touch from the writing process, which could make the appeal less emotionally resonant.

However, it's important to remember that ultimately, the content of the appeal is what matters most. Whether the writing is generated by a human or an AI tool, the content should be clear, concise, and persuasive. If an AI tool can help an appealer express themselves more effectively, then it could be a valuable tool.

Furthermore, it's worth noting that the use of AI writing tools is becoming increasingly common in many areas of life, including business, academia, and even creative writing. As these tools become more advanced, it's likely that we'll see them used more frequently in a variety of contexts.

Ultimately, the decision of whether or not to use an AI writing tool when writing an appeal for a ban should be up to the individual. Some may find that using a tool helps them express themselves more effectively, while others may prefer to write their appeals by hand. What's most important is that the appeal is honest, sincere, and persuasive, regardless of how it was written.

If goof's headmin campaign worked out we wouldn't have to consider the 'sorre for poor englesh' crowd cause we would have FNR attorneys to help write appeals.
In any case, FNR rules should be kept concise and most will try typing the appeal themselves if ai writing doesn't become the norm.

I assume there won't be a problem if I just auto-reject anyone who uses an AI boilerplate for not taking it seriously

If you use an AI to write an apology, doesn't that defeat the entire purpose?

I'm for this if its obvious.

I want to see people put the effort into doing their appeal themselves. If you can't be assed to make your own apology and highlight the things that went wrong (or right) and highlight all the little details you are missing the point. I don't even want an apology in appeals, saying sorry doesn't mean anything to me.

I'm not interested in reading through disingenuous garbage. Every AI assisted appeal has been the exact same for me, and you can smell them a mile away. Its full of filler garbage and misses the points it needs to be hitting nine times out of ten, and instead just looks like overbearing self-pity and inhuman levels of placating others.

In a way these shitty appeals are doing people a disservice from the very start because they have no actual substance. My eyes instantly glaze over and I immediately just don't care. If I ever got one I'd tell the person to piss off and be direct and to the point with me.

Having a giant AI textwall is far less meaningful than briefly saying:

"I fucked up by doing x and it was wrong because of y, I'll avoid doing so again by doing z".

It even takes LESS effort to do that than it does to figure out to get an AI to be a bullshit generator for what you want it to say. That is what pisses me off. We aren't asking or demanding essays, in fact...

NO ONE ACTUALLY WANTS TO READ A BULLSHIT AI GENERATED ESSAY ITS A WASTE OF TIME SAY IT WITH YOUR CHEST, MAN UP, OWN UP.

Farquaar wrote: ↑Thu Apr 20, 2023 10:24 am If you use an AI to write an apology, doesn't that defeat the entire purpose?

If you're using it to communicate better, no. If you're using it to auto generate generically nice responses so you can get back to playing quicker, yes.

People who are so pathologically incapable of formulating a sentence containing the words "I'm sorry" that they need to use a chatbot to do it for them frighten me.

Timberpoes wrote: ↑Wed Apr 19, 2023 9:58 pm What I care about is the substance of the appeal.

Does it have merit? Does it seem genuine? Is the player like to fuck up again?

I think if a player uses chat GP and doesn't disclose it and you find out it definitely affects the merit of an appeal and how genuine it is if the appeals merits are based on a personal statement.

Certain appeals that aren't based on fact and more based on coming to terms with wrongdoing and showing resolve to reform and improve can't be viewed in good faith if they aren't coming from the person whom you actually want to hear from.

Having a Byzantine rule structure does suck but I'm not sure if ruling on A.I generated and when denying an application would be applicable is simply "rule bloat."

Especially when players will use them if they think it gives them a better chance of being unbanned.

It wouldn't be too hard to just write in the appeal template if you had help writing it or not, for fact based appeals it wouldn't be relevant anyway.

A simple checkbox "did you have help writing this appeal yes/no"
If the player checks the box they are covered. This way the admin can request personal statements specifically without writing aids when relevant and deny appeals on the basis of bad faith writing aids when undisclosed at their discretion.

Lacran makes good points, I agree with them

Armhulen wrote: ↑Thu Apr 20, 2023 4:34 pm Lacran makes good points, I agree with them

I'm also a fan of what they brought up.

It seems like adding a "gotcha" to appeals to make it easier to deny them.

The appeals that take up by far the most headmin time are ones that are denied. And I really don't care to have people AI checking every appeal. It's obvious or it isn't. Either way the outcome is the same; the player posted a set of words they want you to take into account.

I'd rather the admin team address the appeals themselves and not dedicate all their headspace to the mode or method they're made.

If the appeal was obviously AI written or assisted, it can be addressed on that basis. If the admin doesn't care, they'll just ignore that fact it was AI written/assisted.

And if it wasn't obviously AI written or assisted, who cares? Players telling us what they think we want to hear is a tale as old as time. I think AI-assisted appeals are more prone to failure than success, but we honestly waste so much time making players jump through all the appropriate hoops when that time is best spent elsewhere.

We could focus our time better on turning ban appeals into an environment where players don't feel like their best chance of success is getting an AI to write their appeal, than what it feels the status quo is where we're requiring players to jump through arbitrary hoops for our amusement and stick to the established script.

As long as straying from the script results in appeal failure, players will want to generate an admin-friendly script as a skeleton key to unlock a successful appeal.

I see this as an issue of providing a service. It's like where Spotify provided a service more convenient than music piracy, so people used it and subscribe to it. Appeals shouldn't be so shitty that players feel the need to write AI appeals in the first place.

Timberpoes wrote: ↑Thu Apr 20, 2023 5:22 pm
I really don't care to have people AI checking every appeal. It's obvious or it isn't. Either way the outcome is the same; the player posted a set of words they want you to take into account.

Yes, but when those sets of words are an answer to "Explain to me why you think X is bad and why you wont do it again" you are testing sincerity and comphresension. An undisclosed A.I answering that for you is neither sincere nor establishes comphrension.

The issue here now is what does the admin do?

If the writing aid is disclosed,sincerity can be assumed but comphrension cannot be established. The admin can request a rewrite without the writing aid to establish comprehension.

If the writing aid has been deliberating undisclosed you can safely assume the personal statement is insincere and comprehension is moot.

There's no sneaky trick involved, the "gotcha" is: "did you lie in your appeal?"

Timberpoes wrote: ↑Thu Apr 20, 2023 5:26 pm We could focus our time better on turning ban appeals into an environment where players don't feel like their best chance of success is getting an AI to write their appeal, than what it feels the status quo is where we're requiring players to jump through arbitrary hoops for our amusement and stick to the established script.

As long as straying from the script results in appeal failure, players will want to generate an admin-friendly script as a skeleton key to unlock a successful appeal.

I see this as an issue of providing a service. It's like where Spotify provided a service more convenient than music piracy, so people used it and subscribe to it. Appeals shouldn't be so shitty that players feel the need to write AI appeals in the first place.

Why is this an either or situation?
A.I impacts the appeal system, you can have a ruling or policy on A.I AND you as headmin can take steps to reform how appeals are tackled by admins.

It sounds like the issues you are alluding to between staff and the appeal system is a very broad topic and would have slow pathway to reform anyway.

Lacran wrote: ↑Thu Apr 20, 2023 5:48 pm Yes, but when those sets of words are an answer to "Explain to me why you think X is bad and why you wont do it again" you are testing sincerity and comphresension.

This is the crux of what I'm saying. As an admin, when it's not about facts but rather about my confidence that someone learned something from a ban and will change their behavior, I NEED THESE. Most importantly, I need sincerity, and that's what being clear about using AI is.

So Arm, your problem is that there's a potential that someone appealing a ban:
1. Uses an AI
2. That you don't detect
3. This tricks you into unbanning them
4. When you wouldn't have unbanned them if you realised they'd used an AI
5. But you would have still unbanned them if they'd admitted to using an AI with the same post?

For this entire suggested ruling to matter, the player has to not admit to using an AI and be caught using an AI when they would have otherwise been unbanned.

To me, you're suggesting we codify that using an AI in this way is perfectly fine and acceptable in all circumstances as long as you admit it. You've changed the test from "are their words sincere?" to "did they say they used an AI?".

If they said they used an AI, it's in the rules that it's allowed and you now can't question the sincerity of their words because admitting to using the AI in their appeal was proof their words are sincere, otherwise why include this clause at all in the appeal rules?

We'll end up with appeals having people running them through AI checkers because that's what we as a community do to see if someone can get fucked over on a technicality. And the constant headspace dedicated to borderline appeals where it's not quite obvious if the person made it by an AI or not, and you get different AI checkers coming up with different levels of certainty that it's AI written.

When genuinely I just think we don't need to care too much if Nobody McGee is secretly duping us at the game of Spacemans Ban Appeal. If an appeal to the admin's ego with some classic kowtowing, or a genuine apology, or a bit of faux introspection or something utterly minor and subjective like that is all that's standing between a player being banned and unbanned then the stakes were pretty damn low to begin with.

Armhulen wrote: ↑Thu Apr 20, 2023 2:32 pm

Farquaar wrote: ↑Thu Apr 20, 2023 10:24 am If you use an AI to write an apology, doesn't that defeat the entire purpose?

If you're using it to communicate better, no. If you're using it to auto generate generically nice responses so you can get back to playing quicker, yes.

I mean if you are incapable of writing an coherent apology (even if it’s because you’re ESL), then you’re probably not the kind of person who is capable of playing a muliayer role-playing game, no?

Farquaar wrote: ↑Thu Apr 20, 2023 9:22 pm

Armhulen wrote: ↑Thu Apr 20, 2023 2:32 pm

Farquaar wrote: ↑Thu Apr 20, 2023 10:24 am If you use an AI to write an apology, doesn't that defeat the entire purpose?

If you're using it to communicate better, no. If you're using it to auto generate generically nice responses so you can get back to playing quicker, yes.

I mean if you are incapable of writing an coherent apology (even if it’s because you’re ESL), then you’re probably not the kind of person who is capable of playing a muliayer role-playing game, no?

I could see inherent bias in incorrect speech working against someone, and I wouldn't blame people who aren't fluent yet in wanting to use a tool that's getting better by the day for translation.

Armhulen wrote: ↑Thu Apr 20, 2023 9:24 pm

Farquaar wrote: ↑Thu Apr 20, 2023 9:22 pm

Armhulen wrote: ↑Thu Apr 20, 2023 2:32 pm

Farquaar wrote: ↑Thu Apr 20, 2023 10:24 am If you use an AI to write an apology, doesn't that defeat the entire purpose?

If you're using it to communicate better, no. If you're using it to auto generate generically nice responses so you can get back to playing quicker, yes.

I mean if you are incapable of writing an coherent apology (even if it’s because you’re ESL), then you’re probably not the kind of person who is capable of playing a muliayer role-playing game, no?

I could see inherent bias in incorrect speech working against someone, and I wouldn't blame people who aren't fluent yet in wanting to use a tool that's getting better by the day for translation.

Maybe there's an inherent bias, but I reckon the more intense bias would be, for good reason, towards the person who suddenly produces a corporate public relations style statement in flawless English after writing ESL the whole time before.

ChatGPT isn't a tool that helps you write better, like spellcheck might. It's a tool that writes for you. That might be fine for some purposes, but for an apology? Write that thing yourself, man.

I think most people understand intuitively that an apology that you didn't write isn't worth anything. This has proven especially true given the caliber of ban appeals we've seen using AI-written apologies over the past few months.

Just chiming in that the people using the "doesn't using ChatGPT to apologize makes it insincere" narrative are ignoring the fact that notes/bans can be wrong. Even if they aren't wrong, the bar to appeal notes is ridiculously high, so I wouldn't blame someone to cut that work short over a 2D space man game. "Deny it for not taking it seriously." Get a fucking grip, lads.

We are going to add a ban appeals amendment about this that'll put the ball into the banning admins court for whether they want to deal with an AI generated appeal or not.

'If an admin determines that an AI generated or assisted appeal is low effort / done in bad faith they are free to close it and request an appeal written by the appellants own hand.'

This will not outright ban AI generated appeals if they are good enough to work or can barely be recognized as being AI generated, but instead ensure admins have the ability to have workable appeals to handle when faced with blatant out of touch AI usage.