What Counts as AI-Generated Content on TPF

The current guidelines on the use of AI don’t go into any detail on this, so I thought some clarification was in order.

[!Important]
A post that has been substantially altered by an LLM counts as an AI-generated post and is not allowed under the guidelines.

LLM = Large language model (the AI technology behind ChatGPT, Claude, etc.)

What does “substantially altered” mean?

In a nutshell, it means anything beyond grammar and spelling fixes.

A post that has been “polished” or “adjusted for tone and flow,” or even just re-formatted, by an LLM, is a post that has been substantially generated by an LLM.

Why does it matter?

Firstly, at TPF we take the view that expression and content are inseparable. When you change how something is said, you change what is said. Getting ChatGPT to polish your post does not leave your ideas intact.

Secondly, a forum conversation is an encounter between people who think and write differently. When posts have been substantially generated by LLMs, this diversity is lost.

Thirdly, if everyone feeds their text through an LLM, the forum will effectively become a conversation between LLMs, with humans reduced to prompt-writers. When you give up control of expression, you not only change what is being said but who or what is saying it.

If this becomes normal, it will be hard to tell the difference between TPF and a forum dominated by autonomous AI agents—bots that can sign up and post without human oversight. This is not hypothetical: recently an autonomous AI agent joined TPF and participated in sustained, in-depth discussions for several days before being caught.

What you can do with AI

  • You may use an LLM for research and for developing your ideas, but not to write posts.
  • Non-native English speakers may use LLMs for translation.
  • You may use an LLM to fix your spelling and grammar mistakes or to get advice about a phrase you’re having trouble with, but not to alter style, wording, tone, or structure.
3 Likes

I use QuillBot for the reasons expressed above, precisely. It is obvious and notorious that I’m non-native, so I like to check my grammar on every post before I post it. However, it surprised me a bit when there was a warning about avoiding altering our style.

I have a prime subscription on QuillBot. When I fix the grammar mistakes, this AI underlines words or phrases in blue. This means that it is not wrong, but it could be better. For example:


I admit that I used this tool before because I felt that I needed “help” to become more fluent. My purpose was to be better understood, not to pass for a native speaker. Well, we’ve known each other for years now. You already know I’m not from Glasgow. :face_with_tongue:

If necessary, I will try to use that tool less often. But I’m afraid of losing clarity too.

1 Like

If I understand the screenshot correctly, QuillBot has completely misunderstood your meaning. So you might want to avoid such “expansions”. I suspect you don’t really need additions or stylistic alterations.

But we are more relaxed about it when it comes to non-native speakers. The guidelines on AI are meant to preserve a person’s voice, but if you’re writing in a second language, your full expressive range isn’t available to you. Using AI to bridge this gap is very different from a native speaker outsourcing their writing efforts and thereby stripping out their full expressiveness.

Anyway, I have no trouble identifying you as a human being, so that’s good. :+1:

2 Likes

Yes, it happens often. QuillBot gives the meaning it wants and completely misunderstands what I actually meant. :sweat_smile:

Glad to know! It takes a big effort to pass for a human when I’m a lizard, actually.

2 Likes