A Modest Literary Analysis of "Use of AI" Sections on TTRPG Kickstarters, Part 1

As you are no doubt aware, the use of AI-generated content in TTRPGs is increasing. I was unfortunately reminded of this fact by @Ettin64, one of the creators of the fantastic Hard Wired Island.

Well, this sucks. There are a bunch of TTRPG crowdfunding campaigns and 500-1000+-page (!) books on DriveThruRPG that are obviously just AI-generated slop: https://t.co/2gJit8okGK

— lo-fi ettin to study/relax to (@Ettin64) December 13, 2023

The associated reddit post contains 13 links to DTRPG pages, of which about half have been delisted at the time of writing, and 4 links to completed Kickstarter campaigns that declare they have used AI.

I decided to revisit my November crowdfunding stats as well as my ongoing December crowdfunding stats to see exactly who is using AI and how. The full stats will come later (around the beginning of January), but right now I, like Ettin, am interested in the rhetoric these campaigns are using to talk about their use of AI.

Just so that we're all on the same page here, this is what Kickstarter has to say about using AI in projects on their platform:

Projects using AI tools to generate images, text, or other output must be:

• Open and Honest: Disclose in your submission and on your project page
  • what AI technology you plan to use in creating your work 
  • how you plan to incorporate the AI-produced content in your project

AND

• Original and Creative: Disclose in your submission and on your project page
  • the extent to which your project is your original work, and 
  • the specific elements you will be creating using AI output

Seems simple enough. So let's see some of the examples of this in action.

Standard Fare

First up is the standard stuff. I won't link any specific campaigns here, suffice it to say that it boils down to "We used Midjourney. We have a license to use Midjourney images." (Substitute Midjourney for whatever AI image creation was actually used). There's not much to say here. It's simple, it's to the point, it fundamentally misunderstands the question that they're allegedly trying to answer:

Do you have the consent of owners of the works that were (or will be) used to produce the AI generated portion of your projects? Please explain.

Almost by definition, artists whose work has been used to train these LLMs cannot consent because the training data is scraped from the internet. The one exception seems to be Adobe Firefly, which they claim is only trained on Adobe Stock and public domain images.

The question of credit raises its thorny head again in the next category of answers...

In the Current State of AI

Several campaigns use shockingly similar language (well not that shocking, two of those campaigns come from the same creator):

In the current state of AI we do not think there is a way that you can credit, but if such a thing occurs in the future we will be sure to implement it. 

Notice the use of "we think" and "if such a thing occurs in the future," creating a rhetoric of plausible deniability. "Well gee, we recognize that this is stealing from artists, but unfortunately there's just no way to get around it. Aw shucks, well if we can steal more ethically in the future we'll be sure to do that."

These campaigns also have identical answers to what parts of their campaign will use AI:

For the illustration of the pages we are using Midjourney, then use Photoshop to edit the illustration for the desired result.

So they're using Midjourney, but human labor is entering the picture to get the "desired result." If I had to speculate, this labor involves mostly eliminating the telltale signs of AI-generated art, but I will grant that there could be other goals here. Speaking of other goals...

Our Artistic Vision

One creator, Mizo Games, has a fascinating take on why they use AI art in all three of their campaigns:

The art generated for this project was created using Midjourney AI exclusively in response to our own unique prompts and direction. We did not use any pre-existing prompts or works from other Midjourney users as the basis for our art. While Midjourney's algorithm inherently draws on its training data, we made original creative decisions in iteratively guiding the AI output through descriptive prompting and deliberate selection. The resulting images reflect our artistic vision and choices alone.

There is a strong focus on "artistic vision" here, a real attempt to show ownership of the generated art by tracing its creativity through their human brains. Interestingly, they show more concern for the property and creative license of "other Midjourney users" than the artists whose work was used as training data in the algorithm. By focusing on how "unique" their prompts are, they seem to feel they can claim artistic legitimacy, ignoring presumably that anyone could use those same prompts and achieve exactly the same results. Still, this presents a fascinating rhetorical turn by virtue of recognizing that there is some illegitimacy tainting the use of AI-generation. If there weren't, there would be no need to defend it this way. Artists don't generally feel the need to disclose that their art was produced by their own unique thoughts and actions, as it's generally understood to simply be true. When it isn't (plagiarism, tracing, clear inspiration that isn't credited), people tend to take notice.

This concern about artistic legitimacy takes on a new form in the next category...

Specifically Not Writing

Too many campaigns to count have some variation of the following in their Use of AI section:

Covers are made with AI. All interior art is made by artists. No AI was used for writing any of the encounters, rules or any part of the product itself.

This one in particular is fascinating as it seems to admit that only humans can be artists (making AI what exactly?). But it's the latter half that bears closer inspection. "No AI was used for writing any of the encounters, rules or any part of the product itself." Well, if it's good enough for cover art, why is it not good enough for the writing? It appears to be almost a recognition that AI generation of writing in a TTRPG project would be illegitimate. The many campaigns that have some version of this all betray the same fundamental concern: that use of AI will render their products corrupted in some way. And while Mizo Games wrapped their use of AI into their overall creative vision, these other creators have instead divorced visual art from their TTRPG product. By separating visual art (cover or otherwise) from the writing, they are creating two categories of creativity in TTRPGs: the core (writing, pure creativity that cannot be touched by AI) and the periphery (visual art, incidental concerns that can be tainted by association with AI because they won't affect the perception of the core material).

Next time

 And with that, I'll leave the rest of the categories for next time. These are:

• Stretch Goals
• The Economics
• Outliers

As a final note, there is a small category of campaigns that report using AI tools like Grammarly to polish their writing. I don't know where I personally stand on this, except to say that a human editor is going to be able to do that job much better. On some level, I commend those campaigns for disclosing that information in the first place as it is certainly nowhere near the level of AI use that these other ones are.