For the first time in my life I’ve had the support of a research assistant and it’s been a fantastic experience. We’ve just finished the first draft of a piece where we’ve been working together over a couple of weeks, and I wanted to document some of that process.
This writing project started with an invitation from a journal editor, asking me to submit a position paper on a very broad topic, after an initial discussion where we’d explored some related ideas. I went back to my assistant and asked for his thoughts on writing something together in this broad area and we agreed that there was probably something worth exploring.
So I went away, drafted some notes on a few areas with a narrower focus, read a few papers, updated the notes, and then had another session with my assistant where we went through the notes together, highlighting the lines of inquiry that we thought had the most potential, identifying issues we’d need to address, and then did some more focused brainstorming, where we ended with a rough outline of a paper.
I took the outline and did some more writing over a couple of weeks, pulling in notes from my library, bouncing ideas of the assistant every now and again, but generally just expanding sentences in the original outline. Once it had grown to a few hundred words, I sent it back to him and asked for his thoughts. He identified a few problems with some of the main concepts that I hadn’t anticipated, and I was back to the drawing board.
After another few round of back-and-forth iteration I gave him the updated outline and asked him to flesh it out. I wasn’t happy with the first version, or the second, or the third. In fact, the first few versions weren’t very good at all and I had to give a lot of input, coaching, and guidance. But over time I noticed that the writing improved. We spent a lot of time talking about the ideas in the paper, and to be honest, I feel like I also grew as a writer and critical thinker during these sessions. I was expecting that I’d simply be handing off the writing to him, and then asking him for revisions when it came back. But the reality is that his writing caused me to think differently, which changed the direction of the piece. And as much writing as he did on the manuscript, I feel like I probably did as much in terms of providing feedback and guidance.
Every time I gave him back a piece that we’d worked on together and asked him to write another draft, it came back better. It took longer than I expected and was harder than I expected, but I definitely feel like it was worth it. I could probably have written the same number of words on my own, and maybe even a bit quicker, but the iterative nature of talking through the topic with my assistant led to a better final draft.
Of course, Claude was my assistant, and the final draft was almost entirely generated by it. I’ve gone through it carefully and edited it lightly but if I’m honest, the bulk of the words are Claude’s. Or are they? I can’t shake the feeling that there’s a lot of ‘me’ in those words, informed by the many interactions we had in the writing process. All the rounds of feedback – in both directions – where we questioned and corrected and pushed back against each other, means that my thoughts and ideas and yes, even my humanity, are embedded into that final draft.
I imagine that some of you reading this will baulk at the idea that I could call something mostly written by Claude as ‘mine’ (or maybe, ‘also mine’). But if you didn’t think it was a problem when I was talking about ‘him’ then you might need to reflect on why it’s a problem when I talk about ‘it’. Typically, research assistants aren’t credited as co-authors for any writing they do as part of research projects (unless the ‘assistant’ part of the work is wrapped up in a postgraduate degree programme as well), and are generally acknowledged for their contributions.
Incidentally, this is aligned with the guidance coming out from many journals now; generative AI can be acknowledged for contributions to writing, but not credited as co-authors.
If it’s OK for my human research assistant to work under my guidance and supervision to prepare manuscripts for me, that I publish under my name, then surely the same must be true for AI assistants?
What am I missing here?