4
\$\begingroup\$

I want to ask a question about the ways people have used ChatGPT to create NPCs, characters, settings, and adventures for their D&D campaign. The question in mind is something along the lines of "what is the best way to use ChatGPT to help create content for my D&D campaign" or "what ChatGPT prompts are best for D&D character creation. Are question of this nature appropriate for this stack, or should they be posted elsewhere?

\$\endgroup\$
3
  • 10
    \$\begingroup\$ "what is the best way to use chatgpt to help create content for my dnd campaign" is very likely going to be opinion-based. \$\endgroup\$
    – VLAZ
    Apr 25, 2023 at 6:39
  • 1
    \$\begingroup\$ What problem are you trying to solve? \$\endgroup\$
    – Akixkisu
    Apr 25, 2023 at 14:20
  • 2
    \$\begingroup\$ An open ended "How do people use this tool?" will probably be closed very quickly. "What is the best way to make content/a character?" are better, but probably still too subjective. The narrower you can focus those questions, the more likely they are to stay open. \$\endgroup\$
    – Novak
    Apr 25, 2023 at 19:45

2 Answers 2

7
\$\begingroup\$

These kind of questions should probably be asked on an AI stack

Reviewing the discussion here, I think my questions are better fit for Artificial Intelligence Stack Exchange I've realized that my questions/problems to solve with this tool more closely relate to Prompt Engineering. For example:

  • How can I use ChatGPT to produce NPC and Monster stat blocks that are consistent with WoTC guidelines?
  • How can I use chatGPT to generate a campaign setting?
  • What prompts are effective to make ChatGPT generate Random encounters?

Ignoring the fact that these questions may be opinion based or have too wide scope, these are really questions about how to prompt ChatGPT, not about how to create RPG game content.

This doesn't mean that all ChatGPT questions should be asked elsewhere

For example, a question like:

  • Is this AI generated encounter appropriate for 3rd level PCs?

Could be perfectly reasonable depending on the actually body of the question

How do we determine when questions relating to chatGPT are appropriate?

When we determine if a question relating to ChatGPT is appropriate, we should consider:

  • is the question asking about how ChatGPT works?
  • is the question asking about how to use content from ChatGPT?

In the first situation, I think in general the question should be asked on AI stack exchange, and in the second I think it's more appropriate here.

What I'm missing

The best challenge to what I have laid out here is from Matthieu's answer, where they ask (paraphrased): what separates this from other tool related questions? Thomas Markov answers this with:

...it is not on par with other tool related questions. Other tool related questions are framed as “how can I use X tool to solve Y” where Y is a specific problem to solve, and “generate context for a campaign” is not a specific problem to solve. It’s an open ended discussion prompt for generating ideas, which we don’t do here

Thomas points out another excellent way to look at ChatGPT related questions. The problem is that it's very difficult to give conclusive answers even to question formatted like this. Even for a question like "how can I use ChatGPT to solve (specific problem)", there are so many different approaches you can take with ChatGPT that quantifying which answer is the best an becomes nearly impossible.


\$\endgroup\$
3
  • 1
    \$\begingroup\$ The thing that makes me the most reluctant to bringing this to an AI-related stack or forum is that those places are lacking the TTRPG expertise needed to qualify the quality of the generated content. ChatGPT is known for being able to generate content that seems valid to the general eye, but that may be wrong when subject to an expert's view. And it seems unviable to me to make one question per generated output from the tool. Qualifying the general viability of the generating prompt (and the tool itself) seems better in the long run. \$\endgroup\$
    – Matthieu
    Apr 27, 2023 at 8:58
  • \$\begingroup\$ You otherwise make a solid argument. Great answer. \$\endgroup\$
    – Matthieu
    Apr 27, 2023 at 9:01
  • \$\begingroup\$ @Matthieu this is exactly why I say we should talk about content they bring. It's difficult both ways, because (in general) there won't be many people in this stack that are good enough t prompt engineering to answer those questions, while in an ai stack many people as you pointed out don't know enough about ttrpgs to give a tailored response. It's unfortunate but it might just be that these kinds of questions are better off of stack exchange in general \$\endgroup\$
    – Jasmine
    Apr 27, 2023 at 17:42
-1
\$\begingroup\$

This seems on par with other tool-related questions

There are plenty of questions about how to use specific tools for TTRPG content. The most obvious that comes to mind is Anydice. When you take them apart, most Anydice questions are nothing more than questions about how to use the tool, or about a mathematical problem for the tool.

You could make the argument that those would fit more on a mathematic-related or statistics-related stack, just like in our case you could point out that an AI-related stack would fit better. But for now, it seems like those questions are tolerated, or even encouraged, on the stack.

As such, I don't see what makes "How do I use Anydice to make this model?" and "How do I use ChatGPT to generate this element?" different. Banning the latter while keeping the former makes little sense to me.

Of course, this does not remove the basics of how to ask a question. Such a question should be focused enough that it has one specific problem to solve, without being too broad. For instance, "How to generate my campaign's contents with this tool?" is not a question fit for the site's format, but asking "How to generate the character sheet for my Human Cleric with this tool?" should be focused enough.

One is about finding a formula, or putting that formula into the right code for the tool. The other is finding what kind of prompt can be used to generate a target result, or a method to build said prompt according to different variables.

The only difference I can think about is that there are many controversies around ChatGPT, or AI-based chatbots in general. However, those controversies shouldn't prevent an actual problem from getting a solution.

However, can we even trust ChatGPT as a reliable tool?

Before asking "How can I use X to do Y?", a more important question might be "Can I use X to do Y reliably?". I couldn't find a question asking something akin to "Can I use ChatGPT to generate [insert TTRPG content here] reliably?".

I have not tried using ChatGPT to generate this kind of content, but I've witnessed content generation for other medias and games that looked very convincing, but was not valid content.

As such, asserting whether or not this is the right tool for the job might be needed first, and that's where experts can share their experience with the tool and determine whether or not this use of ChatGPT is valid.

However, as Doppelgreener mentioned, this question might not be fit for the stack itself. Another meta Q&A might be needed for this decision.

\$\endgroup\$
7
  • 3
    \$\begingroup\$ I’ve downvoted because you’re missing something important here. When asking “How can I use X to do Y”, the scope of Y is really important. Y = “to generate content for my campaign” is wayyyyyy too broad and nonspecific. We can’t just say this sort of question is okay without qualifying the scope of Y, and your answer fails to do that. \$\endgroup\$ Apr 25, 2023 at 22:03
  • 1
    \$\begingroup\$ So no, it is not on par with other tool related questions. Other tool related questions are framed as “how can I use X tool to solve Y” where Y is a specific problem to solve, and “generate context for a campaign” is not a specific problem to solve. It’s an open ended discussion prompt for generating ideas, which we don’t do here. \$\endgroup\$ Apr 25, 2023 at 22:06
  • 5
    \$\begingroup\$ Advising to ask “can i use X to do Y reliably?” does not sound right; we want people to ask about their actual concrete problems, and stepping back to “can I do this?” is not their actual concrete problem. If they ask how to do the thing, and they can't, we can tell them that as part of an answer. \$\endgroup\$ Apr 25, 2023 at 22:18
  • \$\begingroup\$ @ThomasMarkov I agree with your specific example. However, I still believe my resoning stands for more focused questions. For example, something like "to generate a character sheet for my Human Cleric" sounds focused enough. \$\endgroup\$
    – Matthieu
    Apr 26, 2023 at 6:06
  • \$\begingroup\$ @Matthieu Maybe, but your answer doesn’t say that, and the two examples the question gives are way too broad. \$\endgroup\$ Apr 26, 2023 at 6:10
  • \$\begingroup\$ @ThomasMarkov I'm currently working on an edit to clarify that. Thanks for the feedback. \$\endgroup\$
    – Matthieu
    Apr 26, 2023 at 6:10
  • \$\begingroup\$ @doppelgreener that's a fair point. I do still believe the question needs to be asked, but it would probably fit better on meta. Of course this could be answered in this meta right here, but it might be a bit too far from the original point of this question. \$\endgroup\$
    – Matthieu
    Apr 26, 2023 at 6:14

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .