The traditionally recognized Big Five publishers are:
Penguin Random House
Simon & Schuster
The novel must be purported to be written by a human, but turn out to have been mostly* written by AI.
What counts as "mostly" written by AI? As noted in the comments, it is unlikely that we would ever know exactly what percentage of the words in the final draft originated with the human author vs the AI, so there is a very high likelihood of considerable ambiguity. I will use my best judgement, taking into account opinions expressed in the comments, to resolve this fairly.
If it is revealed that the book is written by AI before publication and the publisher decides not to go through with publishing it, this will resolve to "No"
Scenarios that would likely result in a "Yes" resolution:
The author created a plot outline and character sheets, and used an AI tool to write all or most of the prose.
The author used an AI tool to write the initial draft and then subsequently revised and edited the prose.
The author used an AI tool to write the initial draft, and then added a few additional human-written chapters.
Scenarios that would likely result in a "No" resolution:
The author wrote a complete draft or drafts and then used an AI tool to 'polish' the draft with line edits.
The author wrote a complete draft or drafts and then used an AI tool to add a few additional chapters, clearly amounting to under 50% of the length of the book.
The author wrote themself into a corner and asked ChatGPT for ideas for how to get themselves unstuck.
Edited the description to try to provide more clear and fair resolution criteria
I wish the resolution criteria were “a significant part” and not 50% only because this is something that is difficult to factually measure.
What I’m imagining is likely to happen is that a major news organization runs a story about a published novel being written largely with AI. The author claims they invented the individual scenes, but had some sort of AI-autocompletion (like GitHub Copilot) write most of the words.
@MatthiasPortzel That's reasonable. Originally, I was just going to say either 'mostly written' or 'substantially written' by AI, but I wanted a more specific criterion. I guess specificity doesn't do that much good if, as you say, you can't factually measure how many words were AI written.
I'll think about removing the 50% part. Interested to see if anyone else wants to weigh in.
@JoshuaFosse I think something like “significant AI assistance” is much better, and also more likely to be the sort of thing we’d be able to conclude from public reports. The author isn’t going to highlight which words they wrote, but they might say that they built the plot outline, had an AI generate a draft of each scene, and edited/combined/added to that draft, leaving many sections untouched. A scenario like that feels like it should resolve YES.