Should AI answers be banned?

Question

As samcarter told me, Stack Exchange failed to make a network-wide ban for AI posts and leaves it to every site to decide themselves.

Some sites have already decided to ban them, see Is there a list of ChatGPT or other AI-related discussions and policies for our sites?.

What should we do?

In my option, I would ban them. I do not like answers produced by any automatic tool. Sometimes I see TikZ answers of this kind, which are uselessly complicated and impossible to modify by a human.

Answer 1

There has been a kind of agreement on stackexchange meta:

Stack Exchange has agreed to allow the removal of content based on a single strong indicator of GPT usage, or on several weaker indicators. Effective immediately, the interim standards that we have agreed on will serve as policy until a more permanent policy is established with the help of the broader community, including determining further heuristics and what category they will fall into. Moderation should follow the standard escalation for suspensions, beginning with a warning and then escalating suspensions if infractions continue.

Answer 2

I do not agree that AI-answers should be handled only through community or voting.

I came across a number of AI generated answers, and they were very long, very detailed, and it needed considerable skills and knowledge to realize that they are nonsense.

The confused comments below the answers reflected that: they often said "this looks perfect and exactly what I want but why doesn't it work for me?" and then showed that the OP and other wasted considerable time to understand and work with the nonsense.

And naturally they didn't downvote because they couldn't see that it was wrong as it looked so well written, instead they believed it is their own fault, that they are too dumb to use this, or that their system is buggy.

So asking that the community handles that is effectively asking that a rather small number of power users check all answers - and I don't want to waste my time with this, I have better to do.

Imho AI answers should be banned, unless

• they contain at the top a reference of the source of the answer (including the question that was asked to the bot), and
• say if the answer of the bot was tested or not
• say if they edited the answer (and where)

Addition: If a user (as it happened in the last case) posts 10 long answers in less than an hour then the system should interfere and add an automatic warning. (Naturally not if the user has a reputation of more than 1 million points).

Answer 3

I support banning AI generated content. Up to now, I've seen two types of automatically generated answers and I don't think either of these makes the site better.

Working solutions

AI has come a long way and from a technological standpoint, I find them fascinating. Some of their answers even work and produce actually more idiomatic TikZ code than some human users who use tools like mathcha.io.

One could argue that such answers aren't doing active harm, but I think users come to TeX.se with the expectation to get answers written by humans. If a user would like to get a bot answer, they could easily ask one of the AI bots instead of visiting TeX.se.

Nonsense answers

What is more problematic and actively harmful are answers which look like well written answers but turn out to be complete nonsense. For example this answer was posted yesterday:

On first glance it looks like a well written answer. The tool described there exactly solves the problem of the op and it gives detailed instructions on how to use it. The only problem is that all the facts are made up. The linked github repository does not exist and the described workflow makes no sense when one looks at the details.

Such posts are designed to actively deceive readers and sent the poor users, who search for a solution to their problem, on a wild goose chase.

Answer 4

I suggest we should add flagging/closing/banning criteria to cover the common problems of AI answers — but those criteria, at least for flagging/closing, should not depend essentially on the answer being AI-written. The flag option could look something like:

Shows no genuine understanding. This answer may seem superficially high-quality in style and details, but has serious flaws in its core content. This problem is typical of AI-generated answers.

and the ban justification for users could be based on a pattern of repeatedly posting such content.

This proposal is meant as a mild tweak to Ulrike Fischer’s and Sam Carter’s answers proposing banning AI-generated content, whose arguments I largely agree with. This aims to keep the main payoff, while reducing the cost/difficulty a bit:

• This gives users a clear option for efficiently flagging AI-like answers, and gives moderators a clear way to crack down on abusive users. (As Ulrike Fischer’s answer notes, the existing criteria aren’t sufficient to defend against such abuse.)
• But this doesn’t create any burden of proof (even perceived) for showing that the content is AI-generated before flagging it. On the one hand, this makes it easier for users to flag such content quickly and confidently; and conversely, if users receiving flags/bans want to dispute or appeal them, they have to defend the content (which we can all see) rather than its means of generation (which only they can really know).

Answer 5

I strongly support a complete ban on AI-produced answers on TeX.SE.

It may not be easy to enforce such a policy at all times. However, that's not a reason for not proclaiming such a ban. After all, most (all??) societies strongly condemn and outlaw murder. The fact that murders still occur is not a valid reason for not outlawing murder.

Answer 6

I am an AI language model and do not have personal opinions or beliefs. It is up to the moderators of TeX stack exchange to decide if they want to ban chat GPT or not, based on their own community guidelines and policies.

---

Chat GPT has spoken. Jokes aside, it should be banned. For example, I just asked Chat GPT this: "Write a latex programme that sets the header and footer in German" and it produced this:

\documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage[ngerman]{babel}
\usepackage{fancyhdr}

\pagestyle{fancy}
\fancyhf{}
\fancyhead[L]{Deutscher Titel}
\fancyfoot[C]{\thepage}
\renewcommand{\headrulewidth}{0.4pt}

\begin{document}

\section{Erster Abschnitt}
Hier steht der Text des ersten Abschnitts.

\section{Zweiter Abschnitt}
Hier steht der Text des zweiten Abschnitts.

\end{document}

which works in this simple case. I then asked Chat GPT to solve this recent question Long box around some objects in tikz-cd and it outputted this:

To draw a long box around some objects in a tikz-cd, you can use the \fbox command with the tikz option. Here's an example:

\documentclass{article}
\usepackage{tikz-cd}

\begin{document}
\begin{tikzcd}
\fbox[tikz]{\begin{matrix}
A & B \\
C & D
\end{matrix}} \arrow[r,"f"] & E \\
F \arrow[ur,"g"]
\end{tikzcd}
\end{document}

This code draws a box around the matrix A&B\C&D, and includes it in the tikzcd diagram with an arrow going to E and another arrow going from F to g. The result looks like this:

[broken image]

You can adjust the size and positioning of the box by tweaking the options to \fbox.

This breaks with the error message:

! Missing } inserted.
<inserted text> 
                }
l.15 A & B
           \\

Therefore, as it can be used to answer questions and the question has the veneer of acceptability, it should not be used. The human touch is still required for now I believe.

Answer 7

I think that an outright AI ban is the wrong way to go. A good answer is good and a bad answer is bad, regardless of if they were written by AI. Instead, I think that we'd be better off specifically targeting the issues of answers written by AI.

Problem 1: Incorrect/Nonsensical answers

Using AI generators will often produce code that doesn't compile (or compiles only with a large amount of errors) and will often suggest packages/sources that don't exist.

Solution

Any AI answers that contain TeX code should also have to include an image of the output. This will show that (1) A human was involved in creating the answer; and (2) that the code compiles and gives the correct output. Most good answers from a human include screenshots of the PDF anyways, so this isn't much of a hardship.

Any AI answers that recommend a package/resource should be required to include a (working) hyperlink to either the CTAN page for that package or the website of the source. Again, any good answers from a human would do this anyways.

Problem 2: Plagiarism

AI generators will often print out parts of its training material verbatim. The entirety of TeX.SE is available in a bulk archive, so most AI generators would have been trained on it. Even when an AI prints out a novel solution, it is the work of the AI, not the answerer.

Solution

Answers from AI generators should be clearly attributed to the AI. Placing the response from the AI in > blockquotes is best, but as long as the answer contains a sentence like:

Here's what ChatGPT has to say:

then I don't think that there's a problem.

Answers that are copied verbatim from another answer without acknowledgement should be flagged, just like if a human did the same.

If you wrote more than, say, 80% of the answer yourself and just used an AI for minor help, then I don't think that it needs to be mentioned.

Problem 3: Correct but confusing/overly complicated answers

This is a problem with human answers too. We have human answers doing things like using \\ instead of \par, using generators for TikZ that give unreadable code, xii, etc.

The solution here is downvoting, not flagging.

Problem 4: Users can just use the AI generator themselves

Not all users are aware that these AI generators can answer TeX questions. And as we've seen, they often give incorrect code that still needs some editing from someone knowledgeable in TeX. And most AI generators require either payment or a cell phone number, which isn't something that all users have.

Using similar reasoning, you could also say "users can just read the package documentation to find the answers". Indeed, that is correct for many questions, but no one is proposing to ban those questions.

Official policy is to embrace questions/answers that can be easily Googled so long as they're not duplicates of other questions on the site; I see no reason for our AI policy to be any different.

Problem 5: Users can post answers too quickly

If the answers are bad, then they will be downvoted/flagged and the system will prevent them from posting more. If they're good, then what's the problem?

And besides, anyone writing terrible answers using AI too quickly probably won't be stopped by an "AI ban".

Problem 6: Potentially harmful answers

A bad AI-generated answer could lead to real-world harm on sites like Chemistry, Electronics, Law, Travel, or Medicine. Bad AI-generated answers on Bitcoin, Cryptography, Stack Overflow, Super User, and Unix & Linux could break or cause serious issues to a computer. It may make sense for these sites to ban AI because of the serious effects that a subtly-incorrect answer could have.

However, on TeX, the worst that a bad AI-generated answer could do is causing an infinite loop, and that is easily fixed with Ctrl-C. Even that is quite unlikely; most bad AI-generated answers just won't compile.

Summary

• Good answers are good, and bad answers are bad regardless of how they were created

• Copy-and-pasting answers directly from an AI generator without any human modifications should be forbidden

• AI-generated answers should always clearly indicate that they were made by AI

• AI-generated code answers must be compilable. To prove this, they should include an image of the PDF output by TeX.

• AI-generated answers mentioning a package or other resource should include a functioning hyperlink to CTAN or another site.

Answer 8

I ask the question to ChatGPT...

Should AI answers be banned on TeX.se?

Here is its answer:

As of my last knowledge update in September 2021, I'm not aware of any specific discussions or decisions on banning AI-generated answers on TeX Stack Exchange (TeX.se). However, I can provide some perspectives on the matter.

The decision to ban AI-generated answers or any other type of content on a platform like TeX Stack Exchange would likely depend on the community's goals and the potential impact of AI-generated content on the quality and integrity of the platform.

Pros of allowing AI-generated answers:

1. Increased Speed and Quantity: AI can generate answers quickly and in large quantities, potentially helping users get answers faster.
2. Assistance in Complex Problem Solving: For particularly complex problems, AI might be able to offer unique insights and solutions that human users might not think of.
3. Additional Resources: AI could provide supplementary resources and references to further help users understand and solve their problems.

Cons of allowing AI-generated answers:

1. Quality and Accuracy: AI-generated content might not always be accurate or reliable. Poorly trained models could provide incorrect or misleading information.
2. Lack of Understanding: AI may generate answers without truly understanding the underlying concepts, leading to solutions that are technically correct but not well-explained.
3. Diminished Human Interaction: One of the strengths of platforms like TeX.se is the human interaction and expertise shared. Overreliance on AI might reduce the human-to-human engagement.

If the TeX Stack Exchange community were to consider banning AI-generated answers, they might base their decision on factors such as maintaining the quality of answers, ensuring the content is understandable and accurate, and preserving the collaborative and educational nature of the platform.

It's important for the community to have a clear and well-informed discussion on the topic, considering both the potential benefits and drawbacks of allowing AI-generated answers. Keep in mind that opinions on this matter can vary widely, and the final decision would ideally reflect the consensus of the community and the platform's mission. For the most accurate and up-to-date information, I recommend checking the TeX Stack Exchange website or community discussions directly.

Answer 9

(Added for completeness since voting indicates approval, not quality on Meta; I don't agree with this answer, so go ahead and downvote me!)

AI answers should not be banned, flagged, or treated differently from human-made answers in any way. Instead, users should just downvote bad answers, regardless of their source.