Unsurprisingly, this new article from New York Magazine is getting a lot of writers and teachers talking!
Here are some quick thoughts about my experience teaching some writing-heavy music classes this year. It’s not so hopeless!
This past year at NYU, I’ve taught 4 sections of an advanced pop music theory course, and 5 of an advanced pop aural skills course. These classes have a lot of writing in them, and I feel good about the methods I’ve used to get students to really engage with that material.
In terms of ChatGPT and other LLMs, I don’t outlaw them. Being a cop in that regard opens up too much of a can of worms for me, particularly as an adjunct.
But in music contexts especially, I feel I’m good at picking out LLM prose and ideas. First, the made up facts about the readings are quite notable. Second, LLMS spit out generic corporate-speak, which sticks out like a sore thumb when writing about art. Because it spits out THIS bad writing, the assignment submissions that use it just get bad grades. So that’s a big disincentive to use it. In general, my feedback for those submissions is “LLM-y language,” whether they actively used it to spit out an answer in this case, or have just used it so much already that their writing sounds like an LLM. From there, the idea is to show what good (not-LLM) writing and thinking looks like.
On the positive feedback front, I encourage word vomit, parenthetical asides, bullet points, just really informal writing in general. I emphasize that these responses are my opportunity to get to know how you think, whatever that may be. The relationship building is as important as the content produced. Classic process over product. Theory class has 2 informal writing assignments per week, and Aurals has 1 per week that pairs with a transcription. Those assignments are preparation for a seminar discussion, so even if the ideas aren’t there in the responses, the discussion helps model the kinds of critical thinking that leads to good insight. Also, since the students get along well and are really supportive of each other, a really good discussion comment will get peer praise, making that a memorable model of what a good interpretation sounds like.
By the end of this semester, I’ve seen overall usage go down. There are a few “super users” who are still using it to crank out plausible prose, but, the ideas are much better. It feels like to me that they’re thinking critically at the start, and using the LLM as a prose generator/editor from there. Still not perfect, but much better at the core.
Part of why this works for me is:
A) I’m a fast reader and can grade these responses quickly
B) I’m trusting that the real learning and feedback comes from the discussion rather than my feedback, so I don’t put too much pressure on myself on that front
C) this is the end of the line for the theory and aurals sequence, so I don’t have to worry about getting students prepared for the next step in the sequence, skills wise
D) I’m an adjunct, but have a very supportive chair, a union, and got to design much of the courses, so it’s almost all material that I personally enjoy
Again, this won’t be applicable everywhere, but it’s a start.
In general, I’ve felt that this has been a particularly satisfying semester, teaching-wise. I’ve gotten a sense from the students that music is a real refuge where things still make some sense. And because they really want to immerse themselves in music, there isn’t as much of a desire to outsource that immersion to LLMs.
Kevin Laskey 