New Things ChatGPT Is Doing (and Not) With Me

I don’t use any one AI platform consistently. I use different tools for different things, and I’m also often testing different things to see how each tool behaves. Sometimes ChatGPT does something better, other times Gemini does it better, etc. I often put something through AI to imagine what my students would see if they did it through AI, so I can modify my assessments. Which is what I was doing this time.

I read about a story by Omar El Akkad called Busy on Brian Merchant’s newsletter, and I found the article on Vice. So I put the link on ChatGPT and asked it to summarize the story and its main themes (I know students do this instead of reading, even though it’s a v ery short story). It has an intro about a published anthology book which has the story in it, so ChatGPT said there was no story, but that it would tell me about the anthology. I then told ChatGPT to read properly, the story is there. It still could not find it but said it was building on authoritative reviews of the story to summarize it for me. I then told it, no, the story is there, it starts with “On my way to cut God’s tongue” and then it summarized it for me. Then it was like “you’re right, I’m sorry… here is the summary…”. What is really interesting here is that when I went back to check the chat again now, in order to write this blogpost, I realized I had put in the wrong link! (see our chat) I did not give it the link to the Vice version of the story (which introduce the book at the top then has the story) but rather I had indeed given it the link to the book itself. So while I thought it was messing up at first, it was actually not, and when I insisted it was messing up, it apologized and confirmed what I was saying (even though I was wrong). I mean, we know AI did this, but this one was a bit funny. Objectively, the story was not in the link I gave it, but perhaps when I gave the intro sentence, ChatGPT was able to find the story elsewhere… but was… uhh…. too polite(?) to tell me it had to go elsewhere to find it?

Now the themes that came out from ChatGPT were partly relevant but because it is a story, there is so much detail you miss when you read the themes of ChatGPT, including two very important details IMHO, which I will not write here 🙂 but in any case, ChatGPT’s themes were OK but not great, and I kind of know now how to focus a discussion around this story if I assign it in class. Would students read the story if I promise them it is a very very short story? Or do I ask them to read it in class?

What I found disturbing is what ChatGPT asked me after the summary and themes. I know that for some time now, ChatGPT has sort of “gotten to know” its users in such a way that it can tailor things to your interests, even if you do not directly ask it to do that. This is what ChatGPT wrote at the end:

If you’d like, I can also:

• connect this story to UBI debates,
• read it through a trauma-informed or metarelational lens, or
• help you turn it into a discussion or assessment prompt (which I know is very you).

First of all, I didn’t know UBI meant universal basic income, but makes sense that I’d be interetsed in that given my interest in justice. The second is really suprising because yes I work on these things but I did not myself see that connection with the story, not directly. And finally, the one at the end where it says “which I know is very you” is hilarious. It actually did a good enough job with all of the things it was offering. It did the worst in the trauma-informed part, but still added maybe a couple of useful ideas. It was not perfect, of course not, but not a bad start, given all the topics. I actually noticed it also started including things in the suggestions related to my work on equity and care, without having explicitly said it would. I know my ideas when I see them. Of course, it does not reference anything on any of the work it is doing, so that’s problematic, right?

It offered in the discussion part some useful prompts (not too different from what I would have come up with, so no added value there, really, except that to realize I’m not that creative if AI can do it) and then it offered some really good assignment options, and I think these would be really cool to do relatively early in the semester. I’ll comment on them after I share them:

Option 1: Embodied reflection
Ask students to:

• observe one week of “busyness” in their own lives or institutions,
• identify where activity replaces meaning,
• reflect on how it feels in the body (not just cognitively).

Option 2: System redesign (kinship lens)
Students redesign one system from the story (or their university) using:

• dignity without productivity,
• care without surveillance,
• contribution without coercion.

They must explain what beliefs had to change first.

My personal adjustment to all of these is that I think we might use these assignments and this reading to reflect on education, their education in the past in school and at AUC and where things felt like “busy work” without meaning, and to think about how we might change that.

I also have ideas related to aspects of the reading that ChatGPT seems to have missed entirely, but I don’t wnat to blog about them out loud until after I discuss them with students in class.

This is all very basic use of ChatGPT, I haven’t done anything phenomenal, but I’m just sharing it because I think 2 years ago ChatGPT would not have integrated my interests and identity in this way. And I am someone who does not use it often! I only use it once or twice a week max, for 5-10 minutes at a time.

I found the featured image from my WordPress library – it’s just a common one representing AI with a human head surrounded by zeros and ones, and it’s from Pixabay.