The Ethical Tightrope. AI-Assisted Writing in the Digital Age
Working With Claude 3.5 Sonnet
What is the ethical use of GenAI? Only a few months ago, it was pretty simple to distinguish between largely human-generated versus largely AI-generated. This is no longer the case, and as Joanne Penn says, these are the worst GenAi tools you will use moving into the future.
I use AI assistants in my writing processes, but the actual writing is all human me. The ethical use of AI should be important to all writers. AI is not going away; it’s how we use these tools that matters.
Not AI Revolution - More AI Evolution
For the last week, I’ve been working with Anthropic’s Claude 3.5 Sonnet on a recommendation from Joanna Penn of The Creative Penn via her Patreon community. She recently released a Patrion-only video demonstrating how she used Claude for her latest book, “Blood Vintage,” a folk horror. It's fascinating.
Using the new Projects function, now available in the paid version, she set up a project for the book with highly detailed custom instructions. The instructions covered so many aspects that Claude would be left with no doubt about how to respond to any prompt. Needless to say, I ‘borrowed’ the format to assist me with writing my first book:
The first part asked Claude to take the role of an expert in writing my specific genre.
The title and a detailed synopsis of the book.
Main character descriptions and motivations.
The anticipated sales blurb for the book.
A section on the writing style and tone.
Character and psychological depth.
The custom instructions are also a work in progress and will be expanded upon as the book comes together. Then we come to the prompts and the produced ‘artifacts’ as they are named within Claude.
The Ethical Dilemma - Where We Draw the Line
This is where the potential ethical dilemma arises: What and how much of the generated content should be used in the finished draft? The temptation, especially for a first-time novelist, is to use large chunks of the generated text, tweak it slightly, and then move on to the next scene. The problem is, of course, that it wouldn’t be my book. It would be the same as having a reasonably detailed outline and handing everything over to a ghostwriter. It defeats the purpose of the exercise, doesn’t it?
So, how am I using it ethically?
It isn’t so difficult; all it takes is a bit of self-discipline. The first step of the process is to write the vomit draft version of the scene. The scene is uploaded to a new prompt, and Claude is prompted to take it and produce its version. For example:
Can you please help me expand this scene? First Person POV. Produce 1300 (as required) words focusing on the following beats. (A list of the relevant beats).
In the Scrivener project, I duplicated the scene and split the editor, with the original version on one side and the Claude-generated version on the other. It’s straightforward then to step through the two versions and, where Claude came up with something interesting, incorporate it into my version using my voice and editing as required.
From a 1300-word generated text, I could use between three and five ideas and expansions in my version. It doesn’t sound like a lot, but the point isn’t to pad out my scene; it’s to enhance and improve it. One fault I’m already aware of is how little scene setting I’ve been including. Now, while I’m not always a fan of the overly verbose language produced by Claude, it does show me where I can improve things.
I already have several scenes from the beginning hook that are not completed but are close to being the first-draft-ready. I fed these into Claude and, keeping an eye on the custom instructions, asked it to produce a 2000-word version (it was already 1700 words). What it created was really cool, highlighting clearly where my prose was lacking.
An AI Writing Assistant - A Digital Soundboard
Once it was pasted into Scrivener, the split screen was incredibly useful. It enabled me to quickly scan through both versions and mark up the areas in the original scene that needed more work. It’s like having my own personal editor or writing assistant. This is especially useful for my situation, where I don’t yet have any writer buddies to use as a soundboard.
Within a half-day, I’d processed the first, second, and third scenes. Naturally, it’s not even half the job; I still must rewrite them all and expand my markup notes. It’s a great start. As I continue through the book, I’ll be fine-tuning and expanding, where necessary, the custom instructions. As soon as I’m happy with the first sequence, I’ll export it and add it as content for Claude to handle my writing style better.
Before I worked on those first scenes, however, I had an idea for the fourth scene. A detailed 500-word outline was quickly produced, uploaded to Claude, and expanded to around 1500 words. I must say it did a damn good job; if I was lazy, I could almost use it verbatim. But, of course, I won’t be doing that. As I said earlier, it would defeat the raison d’etre for writing the book in the first place.
Breaking Down The Ethical AI Writing Process
So what are the next steps? It gives me a lot to think about for a start. It is not only good ideas but also helps me identify the areas where I’m failing and where my knowledge of the writing craft is lacking. I was telling, not showing, and the descriptions of the settings were (are) bare-boned and uninteresting. That is one of the critical areas that I need to attend to. Bare-bones settings are not good enough; worse, they impact the story’s development. Without the much-improved descriptions, I wouldn’t have had the idea for the fourth scene. I have a better sense of where the story is going.
This, then, is the breakdown of the process:
Write the scene to the best of my ability.
Feed it to Claude and ask Claude to generate its version.
Copy and paste into a duplicate of my scene.
Compare them side by side in a split screen.
Add notes and mark up the original, taking some ideas from the generated text.
Rewrite the scene.
This is only the start; as the characters develop and Claude becomes more familiar with them, I expect to see (and imagine myself naturally) deeper character interactions and points of view. Settings will become richer, which in turn will make the story more compelling and engaging—not only for the reader but also for the writer.
This is where the fun starts.
Applying AI to The Everyday Solopreneur
All this got me thinking. What about the Everyday Solopreneur newsletter? I could use a similar process, which I’m doing now. I dictated most of this newsletter while riding my bicycle this afternoon. Otter.ai didn’t let me down. I can’t recommend it highly enough. By the time I arrived at my destination, the transcription was just over 800 words—more than enough to get me started when I eventually got home late this evening.
I’ve rewritten and expanded the text to over 1300 words (I write the newsletter in Ulysses). Once the spelling and grammar check is complete, I will feed it into a Claude Project specifically for the newsletter. I’ve already compiled all posts from this year into a PDF and uploaded them to the project custom instructions. I’m curious about what Claude is going to produce. Will any of it be usable?
The Late Bloomer
I’ve been slightly concerned that, perhaps, writing a first novel so late in life (64 years old a few weeks ago) really was too late. Then I stumbled upon a Writers, Ink Podcast (Episode 205), interviewing Marsha DeFilippo, the former personal assistant to no less a celebrity author than Stephen King.
After working for Mr. King for 33 years, she retired and began writing herself. She was born in 1954, making her around 70 now. Assuming she’s been retired for several years, she still started writing very late in life.
So, if Marsha can start her writing journey in her golden years, there's hope for this old dog yet. After all, age is just a number, right? (At least that's what I keep telling myself every time I look in the mirror.)
Final Thoughts
In this brave new world of AI-assisted writing, I'm reminded that the key is to use these tools as just that: tools. They're here to enhance our creativity, not replace it. The ethical use of AI in writing is all about finding that sweet spot between leveraging technology and maintaining our unique voice and vision.
It will always be tempting to let the AI do all the heavy lifting, but where's the fun in that? For me, the joy of writing lies in the struggle, the discovery, and, yes, even the frustration of crafting a story from scratch.
So, whether you're 24 or 64 (or 70, like Marsha), remember this: nothing can stand in your way if you want it badly enough. Not age, not technology, not even the occasional bout of writer's block.
Enough for now; I've got a novel to write, with a little help from my assistant, but mostly with good old-fashioned human determination.
Hopefully, you enjoyed this post. If you want to say ‘thank you, ‘ the best way is to get involved in the comments. And my promise to you…If you get in touch, I will answer! So comment away… (a subscription is also nice)