The Use of Artificial Intelligent Agents (AI) on Debate Sites Should Be Banned

This argument was not augmented by Perplexity Pro.

My case is based on these two claims:

1. Debate sites should ban users from presenting another person's arguments as their own in a debate.
2. Artificially intelligent agents, or at least the ones that people are likely to use to augment their arguments, are capable enough debaters that they qualify as "another person" as described in (1).

(1) should go without saying. In a one-on-one debate, presenting somebody else's arguments as your own is plagiarism if done without their permission and cooperative cheating otherwise. I would not necessarily object to AI-enhanced arguments being allowed in a separate category of debates, but this would still qualify as AI being "banned", just not totally banned, since debates which allow AI assistance would be cordoned off. This would be the same as allowing team debates which are clearly categorized as such, while still forbidding individuals in non-team debates from receiving help from others.

(2) is because there are now artificially intelligent agents, like ChatGPT or Perplexity AI or many others, which can generate entire debate submissions that are at least superficially human-like and reasonable. This fundamentally separates AI agents from other technologies that a debater might employ to help formulate their arguments, like encyclopedias or search engines, because they can, with nothing more than a prompt, write something that could be posted in an online debate like this one and, without any modifications by the human debater, be considered by voters a passable submission, although perhaps not the most persuasive piece of rhetoric ever. A person could also do this, whereas books and online databases cannot. Therefore, in the limited sense of seeking help from other people, AI agents are people.

From the debate description:

Proponents of banning AI argue that reliance on AI tools may undermine the development of genuine critical thinking, give unfair advantages, and increase the risk of misinformation or fabricated evidence. Opponents contend that AI can enhance research efficiency, improve argument quality, and support inclusiveness and transparency when used responsibly.

All the claims listed here for both sides are probably true and certainly irrelevant. What matters is whether or not using AI is equivalent to having another human write your argument for you, which it is. That is my entire case, but I might as well address some specific claims you made even though rebutting them is tangential.

In a world where online discussions often devolve into digital food fights, AI chatbots can help maintain the kind of thoughtful, evidence-based discourse that makes debate worthwhile in the first place.

Online discussions should not be devolving into digital food fights anyway, especially not on a debate site, on the understanding that "debate site" means "civilized-debate site". If discussions are becoming the cyberspace equivalent of cafeteria brawls, this should be addressed directly. Saying that debate sites should allow AI because AI can be thoughtful is like saying that boxers should be allowed to use guns because guns can have a safety mechanism, notwithstanding that AI agents do not always generate thoughtful and evidence-based discourse, and are notorious for hallucinating non-facts.

After all, if we're going to argue about pineapple on pizza until the end of time (and let's be honest, we probably will), we might as well do it with style, substance, and the occasional well-timed citation from our digital friends.

An "occasional well-timed citation" from ChatGPT, as opposed to wholesale outsourcing, sounds innocent enough, but, in the context of debates on a debate site, it is still equivalent to an occasional well-timed sentence plagiarized from another human. If stealing a car is wrong, stealing a catalytic converter is wrong too. (That said, I suppose people who pirated movies in the early 2000s might laugh at the notion that they wouldn't steal a car.)

The future of online debate isn't humans versus machines – it's humans with machines, creating richer, more thoughtful, and dare we say it, more fun discussions for everyone involved. And really, isn't that worth debating about?

Online debate is necessarily a matter of humans with machines. Otherwise it would be offline debate. The problem is that AI agents are machines human enough that using them makes online debate a matter of humans with kinda-sorta humans. Letting competitive bicyclists use a motorcycle would make races richer and more interesting, if not more thoughtful. This does not mean that the next Tour de France should be sponsored by Harley-Davidson.

Augmented by me

Before rebutting, I will restate my case. Firstly, getting another person to write your debate for you is wrong. Secondly, AI agents like ChatGPT count as a person for this purpose. Con needs to refute at least one of these claims. His Challenges might seem like they have done so. I will show why they fail to refute my case and sometimes fail to address it.

Rebuttal to Challenge 1: "Another Person" equivalence makes perfect sense in this context

Con claims that AI agents are a lower type of mind than ours, as they, "lack consciousness, intentionality, and genuine understanding", are, "sophisticated pattern-matching systems, not thinking beings", and their output, "synthesizes patterns from training data without understanding or belief". First off, I won't name names, but I can think of quite a few humans who seem to fit that description. More importantly, questions like what is going on in the AI's "mind" or whether or not the AI is self-aware are irrelevant to the key point, which is that AI-generated arguments can look human-generated. The seeming coherency of Con's argument is his own worst enemy here. By posting an argument better than what a lot of humans could come up with, he has proven my point that getting help from Perplexity Pro is like getting help from a person.

Somewhat tangentially, Con's calculator analogy reveals a destitute view of mathematicians. Research-level mathematics is the investigation of logical structures. Buying a soldering iron does not make you a welder.

Con says, "When a human presents an AI-generated argument, they're not presenting "someone else's" argument but rather the output of a sophisticated tool, much like using advanced statistical software or translation programs." Not necessarily. What if the human lied by saying they wrote it?

Rebuttal to Challenge 2: No, my plagiarism analogy does not break down, but your grasp of AI ethics does

Con claims that using AI-generated text cannot be plagiarism because AI cannot be wronged in this way like humans can be. I am not well versed in AI law, but I will not contest that the legal frameworks around plagiarism do not apply to machine-generated content. Whether or not the ethical frameworks apply is a different question. What if it turns out that AI agents are self-aware? Wouldn't they then be ethically, if not legally, entitled to not have their output plagiarized? When I said that presenting someone else's arguments as your own is plagiarism, I meant that in the ethical sense, not necessarily the legal sense.

Many societies throughout history have practiced slavery. If a slave, a human with no legal rights, wrote my argument for me, I would still be plagiarizing from them.

Citing the precedent of collaborative tools misses my point entirely. Nobody has ever written an entire debate by asking Clippy for help.

Rebuttal to Challenge 3: Nuances schmuances

Con claims that there is a significant difference between using AI to generate an entire argument and using AI for relatively minor help. No there isn't. Everything Con said here can be applied to humans as well as AI. If you asked another human to look over your debate round before you submit it and point out spelling errors, you would still be requesting assistance against the spirit of writing your debate yourself. Just because crime A is worse than crime B does not mean crime B is okay.

Rebuttal to Challenge 4: I was talking about ethical concerns, not practical ones

This challenge is a non-contention to my thesis. AI agents probably would give many debaters, i.e., the less competent ones, an advantage, but whether or not they do is irrelevant to whether or not using AI-generated content is equivalent to plagiarism. Unskilled human debaters also lack, "nuance, context-sensitivity, and genuine insight that skilled human debaters provide", yet, if Neil deGrasse Tyson copied a paragraph from a C- freshman essay, he would still have plagiarized it.

Rebuttal to Challenge 5: Not problematic! Just misunderstood!

Con misunderstands the point I was making with each of the analogies he mentioned. My motorcycle analogy was to point out that making something more superficially interesting or fun does not make it better. My boxing-gun safety analogy was to point out that AI agents can generate false or otherwise objectionable content. I never said that using AI gives as large an advantage to a debater as a Harley to a bicyclist or a gun to a boxer.

Rebuttal to Challenge 6: Same problems we had anyway

Does my proposal lack clear boundaries? Only insofar as the boundaries were unclear anyway. "Is paraphrasing AI content acceptable?" Yes when paraphrasing human content is acceptable and no when it is unacceptable. I don't need to go into details about when paraphrasing another human's words is or is not acceptable because my point stands that human and AI content are equivalent for this purpose.

"How would debate sites actually detect AI use?" How would debate sites detect if you paid someone else to write your debate for you? They might not, yet that does not mean outsourcing your debates should not be banned. Even if a ban were practically unenforceable, which Con suggests but does not establish beyond a reasonable doubt, this would not make a de jure ban wrong.

Rebuttal to Challenge 7: The "Context" approach ignores that plagiarism is wrong

Con says, "Academic exercises, casual discussions, and competitive tournaments might reasonably have different standards for tool usage." One thing that academic exercises and competitive tournaments have in common is that plagiarism is wrong. I accept that academic exercises could generally allow for use of AI-generated text, at least if the text is disclosed as AI-generated, but that is no different than having a co-author on a research paper. As for competitive tournaments, if the tournament is mano-a-mano (or meno-a-meno, if each side has more than one human), then using AI-generated text is plagiarism. I could see a tournament that specifically allows AI assistance, but this would need to be in a separate category from normal debates, and, as I said in the previous round, I consider this cordoning off to be a form of ban.

Rebuttal to Con's "Fundamental Philosophical Objection"

Con claims my argument rests on the "naturalistic fallacy" that debate "must remain "purely human" to be valuable". My argument actually rests on the claim that debate must remain purely the work of one human or specified group of humans to not be cheating.

Con says, "The question isn't whether AI assistance is "natural" but whether it serves the underlying purposes of debate: testing ideas, improving arguments, and advancing understanding." Again, Con missed my point entirely. The question isn't whether AI assistance is better than human assistance or readily detectable or easily accessible to anyone with an Internet connection. The question is whether or not AI assistance is, in the context of debates on a debate site, morally equivalent to human assistance. I believe I have shown that it is and refuted Con's attempts to show otherwise. Con repeats this mistake in the next paragraph: "Your framework treats AI assistance as categorically different from other cognitive enhancements (education, training, research tools) without adequately justifying this special status." Con has it backwards. Con treats AI assistance as categorically different from human assistance without adequately justifying this special status.

Con closes with, "This suggests your position may be driven more by intuitive discomfort with AI than by principled reasoning about debate ethics." Ah, yes. AId hominem.