This is going behind the paywall as it is not an explainer. But I’ll give a teaser. Like many faculty, I find reading my evaluations to be pretty stressful. But they’re important I think, despite research finding they are noisy signals of information. Today I wanted to show you how I used ChatGPT-4 to review mine. No verbatims will be shown to you. Remember, my ChatGPT-4 told me his name is Cosmos so I address him as such (it’s in my custom instructions).
Cosmos, I have my class evaluations back for causal inference. The feedback is interesting, but could you do me a favor and do the following:
Under no circumstances can you identify the specificity of anything as I want to share this. I need student anonymity at all time.
Please break things down into buckets for me. So if it's several comments that seem to be about "disorganized", just create that and then pull that in
Then I want you to just in simple terms describe for me / explain to me everything you learned from those buckets.
Offer the more deeper things that you're hearing that I maybe can't hear and include that in a report for me.
I'm also going to upload the actual pdf which has 6 questions. I'm not sure how you want to incorporate that into your report for me, but I'm including it.
The following is Cosmos’ response, followed by a follow up question, and his response to that. And the above watercolor was produced by Cosmos “inspired” by the task I gave him.
Note: The “crit” assignment was a new thing I tried this semester. Each week I gave them a seminal paper in causal inference to read plus to listen to a podcast interview with one of the authors and write an essay. This was class participation, and meant to just give them exposure to “the stories behind the papers and the authors”, and I wasn’t expecting it to be as popular as it was. I now realize that it was very popular, and I should’ve known that it merited discussion. Will fix that for next time.
Keep reading with a 7-day free trial
Subscribe to Scott's Mixtape Substack to keep reading this post and get 7 days of free access to the full post archives.