My PhD student is being advised left and right to let Claude do her lit review, write her qualifying presentation, summarize the books she needs to read to prepare. She is holding fast to the conviction that this slow, frictionful work is the work she signed on for. Immensely proud of her.
@ablerism Love it! I’ve been thinking about friction a lot, especially in knowledge work, and worry about a future where we smooth off the edges of productive struggle.
@jaheppler love it in the sense she’s resisting, not loving she’s being advised by others to use a chatbot
@ablerism Having a PhD that still has that new baby smell (July 2025), it’s weird to feel like there’s already a techno-generational gap. I only had access to LLMs during the revising stage—using it as a fancy thesaurus or tool for concision suggestions. It wasn’t an option that was widely used when I was still in the drafting stage (for the most part).
@ablerism As we all know, doing the lit review helped us to see the gaps better, find new connections, and refine our work. I can’t imagine that happening through AI processing it for us.
@ablerism I guess (1) that’s the way of the world today, but (2) there are conscientious objectors. I’m proud on your behalf.
@JudsonGreene I expect you’re exactly right, the LLM is (I assume) structurally unable to recognize gaps in knowledge. It would be interesting to try some prompts along the lines of “what are some topics in NT scholarship not being discussed much right now?” I don’t think token prediction can get you there, and it’s not a shortcoming of the training data. It’s just a different question entirely.
This post is roughly about this, using the Rumsfeld Matrix as a guide for scholarly energy.
@isaacgreene Thanks for the recommend—amazing stuff. I’m pairing it with Suzanna Black Roberts nice’n’ranty AI takes on her Substack with Alastair. She’s doing a little series.
