My Cheat Code to Writing with AI

I Used to Avoid Using AI.

Here’s What Changed.

Like many academics, I’ve been quite resistant to AI, especially Large Language Models such as ChatGPT or Claude.ai.

Two things have held me back:

• The explosion of AI-generated papers in my classes.
• The work I put into writing my own books and articles.

It’s not just about the work itself, but the powerful way in which our brains change through the act of writing. I want that for me and my students in what we do both in the classroom and what we share out into the world.

Some people can envision their entire thought just sitting and thinking. I’m not like that—I literally think through my fingers (my handwriting is too slow). I often do not know how a sentence or paragraph is going to end up before I dig in there and start making meaning on the keyboard.

The act of writing becomes one of delightful discovery and growth.

That said, I have many tasks where my job is not to create new ideas from scratch, but to shape current information for a different format or for a different audience.

This content is largely known, cumulative across years if not decades of work.

The hardest part for me in writing for a new outlet is not having something to start with—starting from scratch each time is a heavy lift. A little spark of inspiration goes a long way in lightening that load.

Comprehensively editing drafts is also something I do on a regular basis for my graduate students and with my colleagues.

It’s one of my superpowers, and I appreciate being able to focus my best energy into this part of the writing process for my own work.

This is where AI-assisted editing comes in!

After some experimentation, here’s my five-step approach to AI-assisted editing:

1. I rely on my own documents and writing as source information.

   I’ll load my previous writing and research on a topic into a project folder. Sometimes I’ll include meeting transcripts or my own notes if that is relevant. I then limit my query to content within that project folder and ask for output in a bulleted list of key ideas.

2. I check the first response for accuracy.

   This step is key. I do not edit for style first but instead check for accuracy. If there are mistakes, or misinterpretations, I’ll re-prompt pointing out the error to make sure that the learning model on this project knows where I found a mistake. I don’t just delete the error and move on.

3. I ask for revised drafts in an iterative approach.

   Once the foundational information is created and checked for accuracy, only then do I get more creative and ask AI to make a version that meets certain style guidelines. This can be a game changer in formatting citations or grammar easily, saving tons of time.

   Sometimes I’ll also be more specific and ask AI to revise “in the style of a blog post by Stephanie Cawthon.” Because I have a lot of writing out online, there are recognizable trends in how I talk about key topics such as disability, accessibility, and teaching.

   AI tends to over-correct in its aim to simplify and distill my academic language. In this case I will say: “please go back and include the research references you deleted in this last version.”

   Another strategy is to ask for the direct quote or text used to create the synthesized bullet point. I like to do this both to see where the source information is, but also to have specific details available for when I do my own full edit.

4. Now it’s my turn.

   By this point I usually have enough to start my full edit.

   Sometimes I’ll delete whole sections, but most of the time I am already familiar with the overarching themes and can go in and do strategic wordsmithing, smoothing out (or unsmoothing!) sentence structures and making connections  for the reader. This is where I do the bulk of my creative work.

5. Grammar check.*

   I’ll do a final pass with AI to correct for or make suggestions on grammar and spelling, of course, much like I would do in Word or Grammarly.

So far, I’m finding AI-assisted writing to be an evolving process. I’m planning to offer this approach as a model for my students this fall. I hope that we can continue to learn from each other where AI is a support, and not a supplant, for that in-depth brain work.

Some food for thought:

• Where are you finding AI-assisted writing to be helpful in your own work?
• What strategies do you use to help keep yourself engaged?
• What can we do to support young people as they develop their own writing and thinking in this new world of AI-everything?

*I did use this AI assist in writing this newsletter copy but not steps 1-4.  New thoughts, new writing. And wow you can tell. So many words!!!!

Tune in next time for more thoughts on AI, disability, and accessibility!