Raw LLM Responses

I find the most distinctive thing about AI content, aside from the bland writing style we all know and hate, is that it fails to engage with anything that happened in the class. But this distinctive red flag only works if the thing that happened in the class was distinctive. Like, if you do something or looked at something or had a conversation that didn't exist on the slides or isn't something generic, and then you ask students to include it in their writing, ta-daa, you at least require people to think about how to write a complex prompt. If you're lecturing and writing assignments in a way that allows people to not attend class and still hand in an assignment based on the prompt alone, you're being a bad teacher. All assignments should: 1) Closely reflect the class material 2) Have specific requirements that reflect skills learned in class, e.g. a specific analysis technique 3) Include an element that reflects something that only happened in class 4) Require high quality citations and analysis, including material that was read or discussed in class 5) Be scaffolded so the progression of the assignment is clear (and require students to report major changes in their plans). 6) Be comparable to some kind of writing work completed in class (created in such a way that ChatGPT cannot be used), so the true abilities of the students is known prior to the final assignment. People should really struggle if they didn't attend the class, even if they're writing the assignment themselves. The class's content must be absolutely central to the assignment. If it's not... what's the point? It's just an essay.