Site rule addendum: restrictions on AI use - an open discussion

[DaveC426913]

I offer the suggestion that, in the interests of improving the quality of SciFo content, a new rule be introduced that forbids wholesale copy-pasting of chatbot responses.

The rationale:

• The idea is not to outright ban any invocation of Chatbots merely on principle, but to encourage the practice of actual, registered SciFo members making actual arguments in their own words. After all, no chatbots are registered members here (as far as we know), so why should they get to directly engage in discussions, verbatim, as if they are?

• If I wanted to argue with a chatbot, I'd do so directly; I wouldn't engage a member here who simply acts as a mouthpiece for a third party who isn't here to speak for themselves. This is, after all, a discussion forum. And that requires participants being present, not at arm's-length.

• It will encourage members to think about their responses critically (a primary goal of this site, if I understand James R's ethos), not just hit Ctrl-C Ctrl-V.

• It will encourage brevity and therefore clarity. Users will think twice before mindlessly typing out reams and reams of monologue that are so long they routinely exceed the limits of the posting editor (not picking on Tor exclusively, just the most recent example).

• It would encapsulate, in a single place, the explanation of why chatbots are only appropriate for certain specific tasks. They do not think; they do not understand; they do not create solutions; and they are very bad at getting objective things correct. It gets tiring explaining this every time a user vomits forth some new idea they asked a chatbot about.


I ask that this thread stay on-topic. It's specifically about the pros and cons and details of such a rule being written and invoked. it is not for anecdotes or examples of good or bad chatbot behaviour, (unless such examples specifically address a particular point being made).


Ideally, of course, the first and most important member to weigh in should be James R, since, potentially a simple "I'm already on it!" or "not gonna happen!" would neatly obviate further discussion.

 

[exchemist]

I offer the suggestion that, in the interests of improving the quality of SciFo content, a new rule be introduced that forbids wholesale copy-pasting of chatbot responses.

The rationale:

• The idea is not to outright ban any invocation of Chatbots merely on principle, but to encourage the practice of actual, registered SciFo members making actual arguments in their own words. After all, no chatbots are registered members here (as far as we know), so why should they get to directly engage in discussions?

• If I wanted to argue with a chatbot, I'd do so directly; I wouldn't engage a member here who simply acts as a mouthpiece for a third party who isn't here to speak for themselves. This is, after all, a discussion forum. And that requires participants being present, not at arm's-length.

• It will encourage members to think about their responses critically (a primary goal of this site, if I understand James R's ethos), not just hit Ctrl-C Ctrl-V.

• It would encourage brevity and therefore clarity. Users will think twice before mindlessly typing out reams and reams of monologue that are so long they routinely exceed the limits of the posting editor.

• It would encapsulate, in a single place, the explanation why chatbots are only appropriate for certain specific tasks. They do not think, they do not understand, they do not create solutions and they are very bad at getting objective things correct. It gets tiring explaining this every time a user vomits forth some new idea they asked a chatbot for.


I ask that this thread stay on-topic. It's specifically about the pros and cons and details of such a rule being written and invoked. it is not for anecdotes or examples of good or bad chatbot behaviour, (unless such examples specifically address a particular point being made).


Ideally, of course, the first and most important member to weigh in should be James R, since, potentially a simple "I'm already on it!" or "not gonna happen!" would neatly obviate further discussion.

The .net site has a rule like this. However the problem is enforcement. It is often hard to confirm unambiguously that a piece of text is chatbot output. That forum also has a rule that any voicing of suspicions of someone being, or using, a chatbot is ad-hom, off-topic and thus subject to -ve moderation. One can report a suspicion, but it is hard for mods to act without firm evidence, which is hard to come by.

It seems to me one would need to have a rule that if a mod thinks it looks too much like chatbot output, the post can be deleted at mod discretion, without having to confirm it conclusively. That would put the onus on participants to make sure their posts do not look like chatbot text.

I think I would support such a rule. Chatbots are becoming more and more pervasive and, as we’ve discussed, while they have their uses, their actual output is verbose, ingratiating and frequently false. It is lazy - and therefore discourteous- to expect readers to wade through chatbot output.

 

[C C]

The .net site has a rule like this. However the problem is enforcement. It is often hard to confirm unambiguously that a piece of text is chatbot output. That forum also has a rule that any voicing of suspicions of someone being, or using, a chatbot is ad-hom, off-topic and thus subject to -ve moderation. One can report a suspicion, but it is hard for mods to act without firm evidence, which is hard to come by. [...]

This goes back to the difficulty that learning institutions have had in discerning discrete instances in their epidemic of AI cheating. Especially since students have reciprocally responded by manually modifying the content, revising their LLM requests, and other strategies. Still, the most indolent or least talented at deception get punished, I suppose -- that's the case with life abroad, too.

• https://nymag.com/intelligencer/art...eating-education-college-students-school.html
  
  [...] while professors may think they are good at detecting AI-generated writing, studies have found they’re actually not. [...] There are, of course, plenty of simple ways to fool both professors and detectors. After using AI to produce an essay, students can always rewrite it in their own voice or add typos. Or they can ask AI to do that for them: One student on TikTok said her preferred prompt is “Write it as a college freshman who is a li’l dumb.”
  
  Students can also launder AI-generated paragraphs through other AIs, some of which advertise the “authenticity” of their outputs or allow students to upload their past essays to train the AI in their voice. “They’re really good at manipulating the systems. You put a prompt in ChatGPT, then put the output into another AI system, then put it into another AI system. At that point, if you put it into an AI-detection system, it decreases the percentage of AI used every time,” said Eric, a sophomore at Stanford.

 

[exchemist]

This goes back to the difficulty that learning institutions have had in discerning discrete instances in their epidemic of AI cheating. Especially since students have reciprocally responded by manually modifying the content, revising their LLM requests, and other strategies. Still, the most indolent or least talented at deception get punished, I suppose -- that's the case with life abroad, too.

• https://nymag.com/intelligencer/art...eating-education-college-students-school.html
  
  [...] while professors may think they are good at detecting AI-generated writing, studies have found they’re actually not. [...] There are, of course, plenty of simple ways to fool both professors and detectors. After using AI to produce an essay, students can always rewrite it in their own voice or add typos. Or they can ask AI to do that for them: One student on TikTok said her preferred prompt is “Write it as a college freshman who is a li’l dumb.”
  
  Students can also launder AI-generated paragraphs through other AIs, some of which advertise the “authenticity” of their outputs or allow students to upload their past essays to train the AI in their voice. “They’re really good at manipulating the systems. You put a prompt in ChatGPT, then put the output into another AI system, then put it into another AI system. At that point, if you put it into an AI-detection system, it decreases the percentage of AI used every time,” said Eric, a sophomore at Stanford.

OK but most posters on a semi-moribund science discussion forum are hardly likely to go to those lengths. And as the object here is not to prevent academic fraud but just to have a good discussion, if it reads like genuine human input, why would we complain? Any falsehoods can be addressed in discussion.

 

[TheVat]

A broader rule on the order of "please provide attribution on all quoted material" seems simpler. Enforcement isn't possible, but members can create some social pressure by asking posters to use their own words and always provide citations for quoted material. Another good informal rule is to ask for a brief abstract, if someone is presenting a theory or conjecture. That promotes the idea that a reader is owed certain courtesies before they are asked to wade into pages and pages of content.

 

[DaveC426913]

The .net site has a rule like this. However the problem is enforcement.

True, but it doesn't have to be a hard and fast enforcement.

We don't have to enforce that they don't use chatbots at all foe their research - we only have to make sure they take full personal accountability for every word they post. (which should go without saying!)

It will
- force users to actually read and comprehend the whole text, which will encourage critical thinking before posting
- force users to edit it so it doesn't look like chatbot response, which could be an onerous task, and might make them think twice about how much they want to write and pare down
- remove the pat defense: "WelI I asked my chatbot, and it knows more than you do", etc.

Ultimately, it forces the user to take personal ownership of every word they post, rather than fobbing off responsibility onto a third party who is not here.

It doesn't actually matter if they got some of their words from a chatbot. That's OK. I got some of my words from my high school science teacher. But I internalized the wisdom, so now it is my own. I don't come here and say "My science teacher told me 2+2=5, so if you've got a problem with that, take it up with him." If I think 2+2=5 I have to own that myself and I am responsible for defending it - as well as posting it in the first place.

If anybody want to look like a fool by posting dumb false stuff, let them. But have them take direct personal responsibility for every word of that dumb false stuff. No excuses.

 

[exchemist]

True, but it doesn't have to be a hard and fast enforcement.

We don't have to enforce that they don't use chatbots at all foe their research - we only have to make sure they take full personal accountability for every word they post. (which should go without saying!)

It will
- force users to actually read and comprehend the whole text, which will encourage critical thinking before posting
- force users to edit it so it doesn't look like chatbot response, which could be an onerous task, and might make them think twice about how much they want to write and pare down
- remove the pat defense: "WelI I asked my chatbot, and it knows more than you do", etc.

Ultimately, it forces the user to take personal ownership of every word they post, rather than fobbing off responsibility onto a third party who is not here.

It doesn't actually matter if they got some of their words from a chatbot. That's OK. I got some of my words from my high school science teacher. But I internalized the wisdom, so now it is my own. I don't come here and say "My science teacher told me 2+2=5, so if you've got a problem with that, take it up with him." If I think 2+2=5 I have to own that myself and I am responsible for defending it - as well as posting it in the first place.

If anybody want to look like a fool by posting dumb false stuff, let them. But have them take direct personal responsibility for every word of that dumb false stuff. No excuses.

What does that mean? How does one enforce “taking responsibility”?

 

[DaveC426913]

This goes back to the difficulty that learning institutions have had in discerning discrete instances in their epidemic of AI cheating.

Let me reiterate: It doesn't actually matter if they got some of their words from a chatbot, or if we think so. That's OK.

I got some of my words from my high school science teacher. I don't deny that. But I internalized the wisdom, so now it is my own.

If I think 2+2=5 I have to own that myself and I am responsible for defending it - as well as posting it in the first place. I can't hide behind my teacher and say "My science teacher told me 2+2=5, so if you've got a problem with that, take it up with him." or "he's smarter than you".

Ultimately, it forces the user to take personal ownership of every word they post, rather than fobbing off responsibility onto a third party who is not here.

 

[DaveC426913]

What does that mean?

Currently, users are telling us they got their answer from a chatbot. If we challenge their answer, they
1. often can't defend it without going back to the chatbot and
2. just say "well the chatbot said it so it must be true" and "it's smarter than you."

By implementing a rule saying "no chatbot responses", we are removing their ability to fob off responsibility to a third, absent party. Everything they say will be theirs to defend.

It simply forces them to take personal responsibility for every word they post (whether they actually came up with every one of those words themselves is not our concern).


To be clear: we are not trying to prevent them using chatbot as a tool to form their ideas (even form illogical ideas) - practically, we can't stop that, nor do we need to, or want to.

What it does prevent is members freely using "this is what chatbot thinks" as a thesis and prevent members from using"chatbot said so" as a defense.


"Our response is: you posted your words, here. Defend your claim or it is invalid."

 

[exchemist]

Currently, users are telling us they got their answer from a chatbot. If we challenge their answer, they
1. often can't defend it without going back to the chatbot and
2. just say "well the chatbot said it so it must be true" and "it's smarter than you."

By implementing a rule saying "no chatbot responses", we are removing their ability to fob off responsibility to a third, absent party. Anytihng they say will be theirs to defend.

It simply forces them to take personal responsibility for every word they post (whether they came up with those words or not).


To be clear: we are not trying to prevent them using chatbot as a tool to form their ideas (even form illogical ideas) - practically, we can't stop that, nor do we need to, or want to.

What it does prevent is members freely using "this is what chatbot thinks" as a thesis and prevnet members from using"chatbot said so" as a defense.


"Our response is: you posted your words, here. Defend your claim or it is invalid."

OK I understand now. So they can quote chatbot text so long as they don’t say “this is what the chatbot says” as an attempted disclaimer, i.e. they must own what they post, and not try to distance themselves from it. That’s a more modest aim but clearly more easily achievable.

However it still leaves us with the problem of long screeds of turgid stuff, actually from a chatbot but which the poster is implying is his own words.

 

[DaveC426913]

OK I understand now. So they can quote chatbot text so long as they don’t say “this is what the chatbot says” as an attempted disclaimer, i.e. they must own what they post,

If they quote their chatbot we can tell them that's not allowed.

They can obfuscate the text by rewriting it, but look what happens: That will encourage comprehension, critical analysis, paring down, etc. It will encourage them to write a proper post that has had some thought put into it. As well as making them personally responsible for anything they post.

All, important principles of SciFo's mandate.

However it still leaves us with the problem of long screeds of turgid stuff, actually from a chatbot but which the poster is implying is his own words.

We can challenge them directly, because they are now implicitly claiming it's their own words.

 

[C C]

A broader rule on the order of "please provide attribution on all quoted material" seems simpler. [...]

Providing the original human sources at least takes us back to the pre-LLM era. As it stands, the chatbot as a mediator of information poses itself as a barrier to that. (An exception might be those putative "AI answers" that search engines now sport at the top of results, that seem to provide a link to what they are quoting.)
_

 

[(Q)]

The one good thing about using AI here would be having those who normally just post word salads sound somewhat intelligent.

 

[wegs]

I checked this site's posting rules, and this could somewhat fall under ''spamming,'' albeit loosely? I don't consider spam to only be someone posting links to their business / advertisements. It can simply be lengthy, unwanted rambling posts, with parts or all coming from chatbot info.

 

[(Q)]

I checked this site's posting rules, and this could somewhat fall under ''spamming,'' albeit loosely? I don't consider spam to only be someone posting links to their business / advertisements. It can simply be lengthy, unwanted rambling posts, with parts or all coming from chatbot info.

Then, what do we do about Tiassa? Just joking T.

I would suspect that the worst thing it could be is plagiarism. It is literally the copy/paste of someone else work.

 

[C C]

I checked this site's posting rules, and this could somewhat fall under ''spamming,'' albeit loosely? I don't consider spam to only be someone posting links to their business / advertisements. It can simply be lengthy, unwanted rambling posts, with parts or all coming from chatbot info.

Especially fitting since there are so many online advertisements these days (videos, images, text) that are inexpensively generated by AI. No need to hire actors, models, etc anymore, or interview pre-paid real human shills on the street.

It will surely be encountered occasionally: The latest member to join SF just hawking either their or somebody else's "theory" slash wild proposal, and employing LLM for outputting quick promotional spam.

Goodness knows how they ultimately profit from such, since they're usually too unqualified to even be accepted in the corrupted part of the publish or perish industry or the paper mill sector. But some nook is turning a buck. Christopher Langan didn't become an interview magnet for all the crackpot spotlights overnight. It required lots of sloganeering the old-fashioned way for that widespread theater to become aware of him.

Of course, there are also those like maybe "Jammer" who are avatars of the trickster or court jester archetypes. No incremental rise to public fame sought there... Just fulfillment of some performance art hobby, like a mime in the park who has a real night job.
_

 

[wegs]

Especially fitting since there are so many online advertisements these days (videos, images, text) that are inexpensively generated by AI. No need to hire actors, models, etc anymore, or interview pre-paid real human shills on the street.

It will surely be encountered occasionally: The latest member to join SF just hawking either their or somebody else's "theory" slash wild proposal, and employing LLM for outputting quick promotional spam.

Goodness knows how they ultimately profit from such, since they're usually too unqualified to even be accepted in the corrupted part of the publish or perish industry or the paper mill sector. But some nook is turning a buck. Christopher Langan didn't become an interview magnet for all the crackpot spotlights overnight. It required lots of sloganeering the old-fashioned way for that widespread theater to become aware of him.

Of course, there are also those like maybe "Jammer" who are avatars of the trickster or court jester archetypes. No incremental rise to public fame sought there... Just fulfillment of some performance art hobby, like a mime in the park who has a real night job.
_

I can’t help but wonder if some of the actual spammers who post links here might also be hackers; clicking those links could potentially lead to serious issues, whether a virus infects your laptop or something else. Legitimate businesses don’t usually need to advertise in forums to find customers, so I’m happy that SF permanently bans them right away, before members here click on something potentially harmful.

DaveC426913 - why the sad 'reaction' to my post above? I think having a spamming rule already in place, your idea can springboard from it. Like all rules though, they have to be enforced.

 

[DaveC426913]

DaveC426913 - why the sad 'reaction' to my post above? I think having a spamming rule already in place, your idea can springboard from it. Like all rules though, they have to be enforced.

Because it's nothing like spamming. At all.

 

[wegs]

Because it's nothing like spamming. At all.

Well, it's not spam in the strictest of definitions, true. It's not the same as posting bulk ads or something, but it's just as distracting and can create a lack of genuine engagement. That said, I think AI content can be useful, if members disclose it as part of their research on a topic, for example. It seems like you just want some basic parameters on how it can be used on here in a way that doesn't take away from 'organic' discussions and interactions. I agree with you, and I'm also interested to see what James R thinks about it.

Like all rules though, someone has to enforce them. If someone is copying/pasting verbatim, over and over, AI generated content with or without disclaimers, there should be an infraction 'tier' like any other forum violation, and after so many of these infraction points...banned forever.

 

[Sarkus]

If you don't like what someone posts, either it's content or it's source (e.g. LLM) then, here's an idea: don't respond to it. Say "thanks very much, but I'd like to have a discussion with a human and not an LLM", and leave it at that. If the person almost exclusively uses LLMs in lieu of thinking for themselves, then put them on ignore.

I can't see that having a rule specifically to disallow it is worthwhile, especially since this site barely moderates against the existing rules.