A few days ago I wrote a post describing my interactions using Claude as a writing partner, as part of an ongoing attempt to normalise the practice of writing with AI (as opposed to having AI write for you).
I wasn’t sure how that would be received and a friend of mine commented that it was certainly a controversial position, especially in academia. I can see how that might be the case, and yet, for my purposes, this is just a natural part of how I think about writing now.
We need to recognise that using AI to support writing (and other forms of creation) is more nuanced and non-binary than we’re currently treating it. Creating with AI is going to become more collaborative and seamless, especially as the user-interfaces for these systems mature.
For certain kinds of writing having The Button is fine because we just want to generate some copy. But for others, like the kind of writing I described in my post, we want a more intuitive back-and-forth, iterative interaction, where a simple button to generate text just won’t cut it.
For example, this post is 100% ‘me’ with no AI contribution. Which makes sense because I want to share my thoughts. Having an AI write this post makes no sense. I might ask it for feedback and to suggest ways to improve it, but ultimately this kind of writing only makes sense to do yourself.
Other posts are more collaborative. For example, in this post yesterday the article summary is AI-generated, and I only wrote a short comment that included my opinion. But the purpose of that post was to highlight a new feature in Copilot, and AI is well-placed to provide an overview for readers. My only contribution was a to include a short annotation with a point to note. That kind of post isn’t about including ‘me’ in the writing; it’s just a signpost towards something else.
Other kinds of writing will be 100% AI-generated. For example, the new Deep Research research tool from DeepMind, or a Perplexity Page, will likely be all AI. Which is fine when you’re just looking for information on a topic. That’s not to say we won’t interact with that information in different ways, but the initial content can be all AI-generated.
My point is, we need to start developing a sense of taste for when to lean heavily on AI for content generation, and when we want to engage with it on a spectrum with either more, or less, AI-generated input.
We also need to help students develop this sense of taste, and we need to adjust our assessment tasks to take it into account.
Whether we write with AI shouldn’t be an on/off-type question and yet, too often in academia, this seems to be the conversation we’re having.