[Q/D] AI-generated WA proposals as plagiarism

[The Ice States]

Thread summary: would Moderation consider it plagiarism to submit WA proposals whose text is generated entirely or primarily by Chatgpt or another AI language model? Should it be considered plagiarism?
-----
Over the last weeks, we have seen a very significant proportion of WA proposals being substantially, if not entirely generated by Chatgpt or other AI language models. Just upon a quick review of the last 30 days, I can find not one, not two, not three, not four, not five, but six proposals written entirely or primarily by an AI language model. That is not even counting AI-generated drafts posted to the forum, eg [1][2][3][4][5].

The problem with such proposals is two-fold. Firstly, people writing "their" proposals simply by copying text generated by an AI, rather than authors writing the text themselves, fundamentally degrades the fun and engaging nature of the WA, as players no longer have to do any real work to write their proposals. However, more importantly, it represents persons taking credit for work which is not theirs. Players should not get badges or the clout from authorship for work they did not do. Like submitting text from eg a Wikipedia article is (rightfully) considered plagiarism, I strongly believe that submitting AI-generated proposals ought to be considered plagarism, and thus illegal.

One counterargument against considering AI-generated WA proposals plagiarism is that the submitting author, presumably, was the one who entered the query to the AI. This logic is flawed for two reasons. Firstly, there is no real way to police who entered the query into the AI. It is often impossible to know whether the person who submitted the proposal in fact entered the query themselves, or asked a friend to enter a query into Chatgpt and then submitted it without their permission. However, more importantly, this does not apply to eg asking a friend to write a proposal and submitting without their permission, which is absolutely considered plagiarism. Why is it that for AI models whoever asked for the proposal is entitled to full credit for the proposal, but in all other cases they are not?

Another counterargument is that it is effectively impossible to accurately know whether an AI generated a proposal. This is a selective argument which does not apply to other rules which are likely harder to enforce, and also uncontroversial. There are probably infinite ways to violate the GA Branding rule; naming a committee the name of the author backwards, including an acrostic naming the author in the first letters of the clauses, including such an acrostic in the last letters, and so on. Yet the rule remains enforced. Likewise, if a proposal is plagiarised from some obscure internet website (CF: Wikipedia etc) or even some obscure physical book, that is still plagarism, even if it is very difficult to know if this happened. However, it is still not hard to run a proposal which looks AI-generated prima facie through a website such as Zerogpt, and see if it appears to be more than eg 50% AI-generated.

There has been no formal ruling on this matter by Moderation, and to that end I would like to request an explicit ruling on whether it is considered plagiarism to submit a WA proposal generated entirely or primarily by Chatgpt or another AI language model.

Thanks.

[United Calanworie]

The current stance of the Moderation team is that the usage of AI to generate WA/SC proposals does not qualify as an inherent plagiarism violation for the sole reason of using AI. It is entirely possible for an AI proposal to violate the plagiarism rule through the way AI works, or for it to violate any number of other rules.

With that said, it's probably worth hearing more from the community.

[The Ice States]

The current stance of the Moderation team is that the usage of AI to generate WA/SC proposals does not qualify as an inherent plagiarism violation for the sole reason of using AI.

Could Moderation clarify the reasoning for this stance, if it is possible for that to be shared?

[United Calanworie]

The current stance of the Moderation team is that the usage of AI to generate WA/SC proposals does not qualify as an inherent plagiarism violation for the sole reason of using AI.

Could Moderation clarify the reasoning for this stance, if it is possible for that to be shared?

We've looked into what it would take to be able to determine whether or not a proposal was AI generated or not and came to the conclusion that there would be a lot of cases where it's "is this just a really shitty draft, or is it AI-written?"

So if the AI-written draft somehow miraculously doesn't plagiarize any other laws/papers/resolutions/whatever-have-you, but it's just real shit, we don't want to be in the business of determining whether or not it's shit that came out of a human's brain or shit that came out of an AI.

But obviously, if it does plagiarize from other laws/papers/resolutions/whatever, it's plagiarism, it doesn't matter that it was written by AI.

[Lareine Alpine]

Thread summary: would Moderation consider it plagiarism to submit WA proposals whose text is generated entirely or primarily by Chatgpt or another AI language model?
-----
Over the last weeks, we have seen a very significant proportion of WA proposals being substantially, if not entirely generated by Chatgpt or other AI language models. Just upon a quick review of the last 30 days, I can find not one, not two, not three, not four, not five, but six proposals written entirely or primarily by an AI language model. That is not even counting AI-generated drafts posted to the forum, eg [1][2][3][4][5].

The problem with such proposals is two-fold. Firstly, people writing "their" proposals simply by copying text generated by an AI, rather than authors writing the text themselves, fundamentally degrades the fun and engaging nature of the WA, as players no longer have to do any real work to write their proposals. However, more importantly, it represents persons taking credit for work which is not theirs. Players should not get badges or the clout from authorship for work they did not do. Like submitting text from eg a Wikipedia article is (rightfully) considered plagiarism, I strongly believe that submitting AI-generated proposals ought to be considered plagarism, and thus illegal.

One counterargument against considering AI-generated WA proposals plagiarism is that the submitting author, presumably, was the one who entered the query to the AI. This logic is flawed for two reasons. Firstly, there is no real way to police who entered the query into the AI. It is often impossible to know whether the person who submitted the proposal in fact entered the query themselves, or asked a friend to enter a query into Chatgpt and then submitted it without their permission. However, more importantly, this does not apply to eg asking a friend to write a proposal and submitting without their permission, which is absolutely considered plagiarism. Why is it that for AI models whoever asked for the proposal is entitled to full credit for the proposal, but in all other cases they are not?

Another counterargument is that it is effectively impossible to accurately know whether an AI generated a proposal. This is a selective argument which does not apply to other rules which are likely harder to enforce, and also uncontroversial. There are probably infinite ways to violate the GA Branding rule; naming a committee the name of the author backwards, including an acrostic naming the author in the first letters of the clauses, including such an acrostic in the last letters, and so on. Yet the rule remains enforced. Likewise, if a proposal is plagiarised from some obscure internet website (CF: Wikipedia etc) or even some obscure physical book, that is still plagarism, even if it is very difficult to know if this happened. However, it is still not hard to run a proposal which looks AI-generated prima facie through a website such as Zerogpt, and see if it appears to be more than eg 50% AI-generated.

There has been no formal ruling on this matter by Moderation, and to that end I would like to request an explicit ruling on whether it is considered plagiarism to submit a WA proposal generated entirely or primarily by Chatgpt or another AI language model.

Thanks.

While the idea behind limiting or banning or repealing AI generated proposals is very much welcome and that I would agree with, I suspect it cannot constitute as plagiarism.
Plagiarism is copying of existing contents or ideas while as far as I know language models like GPT create unique human-written like text content...

[The Ice States]

Could Moderation clarify the reasoning for this stance, if it is possible for that to be shared?

We've looked into what it would take to be able to determine whether or not a proposal was AI generated or not and came to the conclusion that there would be a lot of cases where it's "is this just a really shitty draft, or is it AI-written?"

So if the AI-written draft somehow miraculously doesn't plagiarize any other laws/papers/resolutions/whatever-have-you, but it's just real shit, we don't want to be in the business of determining whether or not it's shit that came out of a human's brain or shit that came out of an AI.

But obviously, if it does plagiarize from other laws/papers/resolutions/whatever, it's plagiarism, it doesn't matter that it was written by AI.

I think the best way to address this issue, then, would be to give deference to the author where it is uncertain whether a proposal is written by an AI, and/or apply a higher standard of likelihood to punish someone than to discard/pull a proposal. In cases such as this or this there is no way that they are human-written; and if it is deemed that a proposal is likely, yet not entirely certain, to be AI-written, the proposal can be pulled without additional action against the author. This would largely prevent plagarism from AI without resulting in authors being unjustly punished for ostensibly AI-written proposals which are in fact human-written.

While the idea behind limiting or banning or repealing AI generated proposals is very much welcome and that I would agree with, I suspect it cannot constitute as plagiarism.
Plagiarism is copying of existing contents or ideas while as far as I know language models like GPT create unique human-written like text content...

Copying AI content is copying the "existing contents or ideas" the AI creates. A university or school would, in the vast majority of cases, consider copying text from an AI for an assignment to be plagiarism; practical issues with enforcement are the more relevant argument against declaring copying AI text plagiarism.

[The United Penguin Commonwealth]

One counterargument against considering AI-generated WA proposals plagiarism is that the submitting author, presumably, was the one who entered the query to the AI. This logic is flawed for two reasons. Firstly, there is no real way to police who entered the query into the AI. It is often impossible to know whether the person who submitted the proposal in fact entered the query themselves, or asked a friend to enter a query into Chatgpt and then submitted it without their permission.

wouldn’t this be true of normal plagiarism too?

However, more importantly, this does not apply to eg asking a friend to write a proposal and submitting without their permission, which is absolutely considered plagiarism. Why is it that for AI models whoever asked for the proposal is entitled to full credit for the proposal, but in all other cases they are not?

does it make sense to credit a program for a creative work? there are generative synths that produce music without intervention. should they be credited as the creators of the music? this is a pretty deep and philosophical question.

sorry if I’m not supposed to post here.

[Lareine Alpine]

Copying AI content is copying the "existing contents or ideas" the AI creates. A university or school would, in the vast majority of cases, consider copying text from an AI for an assignment to be plagiarism; practical issues with enforcement are the more relevant argument against declaring copying AI text plagiarism.

What if AI is considered as a complice ? in other words, the ideas didn't exist until the person asked to. It can be considered as ghost writing instead

In the case of assignments for school, the issue is more about lack of original thinking or content than it is about plagiarism or rather it has to be that way.

[Barfleur]

I'm worried about how (at least for now) it is not always easy to tell when something has been generated by AI. Aside from some obvious cases where it is clear that a person did not do the work, how would the mod team even enforce a prohibition against AI-generated proposals? It's not like a school assignment where you can pull the student aside one-on-one and test whether they actually have knowledge of the material. Unless something drastic changes, I think an outright ban on such proposals would just result in a wave of GHRs about AI, some of which would have zero basis and others which would not be clear one way or the other, resulting in possibly legal proposals being deleted and their authors punished for plagiarism.

[Terminus Station]

i think AI generated WA proposals are a boon instead of a detriment. We could use it to improve current and future proposals or come up with ones that haven't been proposed yet. Unless you ask the AI to copy an existing proposal verbatim I dont see how it would fall under "plagiarism". As for copypasting a proposal written by the AI without modifications I dont see how thats an issue if the main objective of WA Proposals is to come up with new legislature, instead of say, an academic paper meant for submission.

[Wrapper]

Just upon a quick review of the last 30 days, I can find not one, not two, not three, not four, not five, but six proposals written entirely or primarily by an AI language model.

How can you even tell with near-100% certainty that these are AI written?

[The Ice States]

i think AI generated WA proposals are a boon instead of a detriment. We could use it to improve current and future proposals or come up with ones that haven't been proposed yet. Unless you ask the AI to copy an existing proposal verbatim I dont see how it would fall under "plagiarism". As for copypasting a proposal written by the AI without modifications I dont see how thats an issue if the main objective of WA Proposals is to come up with new legislature, instead of say, an academic paper meant for submission.

Community engagement is also an important aspect of the WA. If people just get robots to write "their" proposals for them, that aspect is lost. Why bother learning about and crafting policy yourself when you can just get Chatgpt to do it for you?

Just upon a quick review of the last 30 days, I can find not one, not two, not three, not four, not five, but six proposals written entirely or primarily by an AI language model.

How can you even tell with near-100% certainty that these are AI written?

AI written proposals have a very characteristic, repetivite writing style which is obvious in all of these proposals. Putting any of those through an AI text detector, eg [1] would show that they are AI-generated, besides obviously having the characteristic writing style AI texts have.

[Terminus Station]

i think AI generated WA proposals are a boon instead of a detriment. We could use it to improve current and future proposals or come up with ones that haven't been proposed yet. Unless you ask the AI to copy an existing proposal verbatim I dont see how it would fall under "plagiarism". As for copypasting a proposal written by the AI without modifications I dont see how thats an issue if the main objective of WA Proposals is to come up with new legislature, instead of say, an academic paper meant for submission.

Community engagement is also an important aspect of the WA. If people just get robots to write "their" proposals for them, that aspect is lost. Why bother learning about and crafting policy yourself when you can just get Chatgpt to do it for you?

How can you even tell with near-100% certainty that these are AI written?

AI written proposals have a very characteristic, repetivite writing style which is obvious in all of these proposals. Putting any of those through an AI text detector, eg [1] would show that they are AI-generated, besides obviously having the characteristic writing style AI texts have.

You have no idea what you're talking about do you?

[The Ice States]

Community engagement is also an important aspect of the WA. If people just get robots to write "their" proposals for them, that aspect is lost. Why bother learning about and crafting policy yourself when you can just get Chatgpt to do it for you?


AI written proposals have a very characteristic, repetivite writing style which is obvious in all of these proposals. Putting any of those through an AI text detector, eg [1] would show that they are AI-generated, besides obviously having the characteristic writing style AI texts have.

You have no idea what you're talking about do you?

I actually do know how the WA works, as a 23x (soon to be 24x) author with over 2000 posts in the WA forums (across various nations) versus someone with a grand sum of two posts in the WA forum. Community engagement is indeed an important aspect of the WA, and that aspect is harmed by people just getting AI to write "their" proposals for them.

[Makko Oko]

How can you even tell with near-100% certainty that these are AI written?

AI written proposals have a very characteristic, repetivite writing style which is obvious in all of these proposals. Putting any of those through an AI text detector, eg [1] would show that they are AI-generated, besides obviously having the characteristic writing style AI texts have.

You realize 90% of the time those "detectors" put out false positives right? Have you read the news recently? Professors, teachers, many people have used them and have gotten false positives and gotten people in trouble. That's why people now say to do your work on like Google Docs where version history exists.

[Astrobolt]

Completely agree with Ice States that AI generated proposals have no place in the WA.

Unfortunately the problem is its hard to get proof that an AI tool was used. You might be able to make a good guess, but given that even AI detection tools result in false positives, I'm not confident moderation will be able to discern which proposals used something like ChatGPT and which did not.

[Free Papua Republic]

The overwhelming consensus in the AI community is that universally """"AI detectors"""' are unscientific trash shit whose results is worse than useless. I sympathize with the concern regarding generated proposals, but cut out any and all mention and usage of """"AI detectors"""".

[Kenmoria]

I have been able to recognise the ChatGPTed proposals on the GA as being from ChatGPT, but this wasn’t a reliable, firm knowledge. I don’t think that I would be happy with a similar test of recognition being imposed as a basis for a violation of one of the most important rules for proposals. Looking back through the GA, a lot of proposals from very new nations, from younger players, strike me retrospectively as being akin to ChatGPT, though that is plainly impossible. I think that this is a matter where it is better to leave it to delegates having the common sense not to approve, and big regions having the common sense not to vote in favour of, AI-generated proposals.

Besides, most proposals written by ChatGPT break some rule or another anyway, because ChatGPT has no real knowledge of the GA. It frequently contradicts extant resolutions, does something beyond the scope of the GA’s power, or doesn’t write proposals with operative clauses. In my testing, it has also begun justifying its choice of clauses in the text of the proposal itself, arguably straying beyond the confines of a law. There doesn’t seem to me to be a need for a separate rule targeting ChatGPT.

[Hannasea]

Players should not get badges or the clout from authorship for work they did not do.

Just get rid of badges and remove nation names from proposals altogether, problem solved.

I have been able to recognise the ChatGPTed proposals on the GA as being from ChatGPT, but this wasn’t a reliable, firm knowledge. I don’t think that I would be happy with a similar test of recognition being imposed as a basis for a violation of one of the most important rules for proposals. Looking back through the GA, a lot of proposals from very new nations, from younger players, strike me retrospectively as being akin to ChatGPT, though that is plainly impossible. I think that this is a matter where it is better to leave it to delegates having the common sense not to approve, and big regions having the common sense not to vote in favour of, AI-generated proposals.

Besides, most proposals written by ChatGPT break some rule or another anyway, because ChatGPT has no real knowledge of the GA. It frequently contradicts extant resolutions, does something beyond the scope of the GA’s power, or doesn’t write proposals with operative clauses. In my testing, it has also begun justifying its choice of clauses in the text of the proposal itself, arguably straying beyond the confines of a law. There doesn’t seem to me to be a need for a separate rule targeting ChatGPT.

Agreed with this. Weirdly though this post felt like it was written by ChatGPT itself.

[Flanderlion]

I don't see a practical way of enforcement, so I think that it should remain legal. If it's blatantly ripping something off like now, sure, it should be smacked like it already is today. It's on players to determine whether AI written stuff should pass a vote/detect whether it is, not moderation.

[Simone Republic]

*Snip*

I am inclined to agree with Kenmoria's view above. An experienced author (not to say a Gensec, not to say a Moderator) can tell from a quick glance that the text would fail. It might increase Gensec's workload though.

Also GA you need a lot of "community engagement" ("socializing" in corporate-speak) and I don't see that happening with a ChatGPT written proposal. SC there are already some well established boilerplate text from some well established authors for liberations anyway, but the recent plagiarism issue over there already speaks for itself.

So based on the above I also tend to agree with Moderaion's view (as UC mentioned) that if a proposal is a piece of crap, it'd be a piece of crap whether it's AI generated crap or human generated crap.

Just my $0.02.

[Wrapper]

AI written proposals have a very characteristic, repetivite writing style which is obvious in all of these proposals. Putting any of those through an AI text detector, eg [1] would show that they are AI-generated, besides obviously having the characteristic writing style AI texts have.

Mmm hmm. Apparently my first resolution was about 50% written by AI.

[Wallenburg]

The current stance of the Moderation team is that the usage of AI to generate WA/SC proposals does not qualify as an inherent plagiarism violation for the sole reason of using AI. It is entirely possible for an AI proposal to violate the plagiarism rule through the way AI works, or for it to violate any number of other rules.

With that said, it's probably worth hearing more from the community.

That's interesting, because all of GenSec were given the impression that we were giving the mods the chance to provide input to GenSec's stance that AI-generated content is plagiarism.

[Querria]

So what if we make a nation ran by ChatGPT and we post proposals it makes?

I feel like this is a thread made due to the fallible fear from TV and movies aka the terminator franchise, & pushed by people like Elon Musk.. and NOBODY should listen to a person such as Elon Musk.

[Saor Alba]

So what if we make a nation ran by ChatGPT and we post proposals it makes?

I feel like this is a thread made due to the fallible fear from TV and movies aka the terminator franchise, & pushed by people like Elon Musk.. and NOBODY should listen to a person such as Elon Musk.

(Image)

Elon Musk was not mentioned once in this thread and he is not the only person to raise concerns about AI. Regardless, given that AI models are trained on human-created content there is good reason to have concerns over whether or not their output constitutes plagiarism.