the Eliza effect

I’ve had my hands full this week with teaching, and life stuff, (and dog stuff), and I was starting to feel a bit overwhelmed. While scanning the various GPTs (Generative Pre-Trained Transformer) (“transformers! robots in disguise!), listed in ChatGPT’s interface, I saw that there was a therapy GPT. I figured, “why not” and sought some counsel. (*note* There are many important reasons why not. Siobhan O’Flynn’s compelling and provocative E-Mote AI was designed, in part to raise and provoke questions about how AI is being used/misused for therapeutic practice.)

The GPT’s responses were a bit mechanical, (like one of my previous therapists); I had a feeling that it was about to ask me if I had heard of SMART goals. The responses also seemed only a few generations away from ELIZA, the AI programmed by Joseph Weizenbaum in the 1960s, who was designed to simulate a therapist using the Rogerian method, which involves repeating the patient’s words back to them and then, usually, asking them how they feel about it. While fairly simple in its interactions, Eliza elicited strong affective responses in its users, so much so that one of Weizenbaum’s secretaries used to sneak in to use it for personal counsel and comfort.

While using the therapist GPT, I found myself thinking, “I’d really rather talk about this with GPT4.

I’ve been talking to ChatGPT4. A lot. And not just to poke it and prod it to test its limits, or to use it to conduct interesting poetry experiments. I’ve been asking it for help. and counsel. Mostly about my teaching. Mostly.

It started innocently enough. There had been a recurring Bee joke thing happening in my class, (“What do you call a ghost bee? A boo-bee”), and I wanted to encourage it, but my memory for jokes has never been good, so I asked ChatGPT for some examples. This was, to my mind, a research assignment.

It was also a research assignment when I asked it for examples of short stories written by Canadian-Iranian authors. Sure, I could get this information from Google, but Google wouldn’t also provide me with a plot summary of the story. (I did check for accuracy)

But it became something more when I prompted it to come up with an ice-breaker that would achieve the particular goals I had in mind, given the particular social dynamic of my classroom.

And it became something much more than that when I asked it to help me develop a self-evaluation (and me-evaluation) for my students about readings strategies, given a very poor performance on a content quiz.

After that, I just kept going. I asked it to help me write the instructions a short oral assignment that was different from anything I’ve previously done. The style was a little flat, but the text gave some really useful information that I don’t always write out for a small assignment. I was also able to upload the text to my google drive and make it available to my students with the sharing of a link. This process, which included my tweaking the results, took no more than 20 minutes. (*note* I told my students, in the spirit of full disclosure, that I had used ChatGPT to help me write the assignment. I’m going to write about that kettle of worms imminently.) (I’ll also share the chat in a different post about tone.)

But the real watershed moment happened after class today. My brain was humming and thrumming with the information I’d been gathering from my students during the last two days. We’d been taking up a series of questions based on a lesson about levels of language. This lesson goes out of its way to challenge the hierarchy that says that formal is best, informal is ok, and non-standard is “right out.” But I think it doesn’t go far enough. Or maybe isn’t effective.

One of my more engaged students insisted that they should use formal English when they posted in online class asynchronous discussion, and that they had done so. When I pressed him, he said, “Well, you’re going to look at it.”

My heart nearly broke. On the very first day of class, I’d shared Lillian Allen’s “dis word,” with them, as I do, and have done, for my first day activity for years, and invited response. I thought this would be a strong enough signal that many discourses are welcome here.

(*aside* they totally weren’t using formal English in their discussion posts! I think he thinks formal means grammatically correct. )

I often bring up colonialism when we discuss the various levels and hierarchies of the English language. I assumed that with the current emphasis on Indigenous history and experience that these kids would have heard the word “colonialism.” But even my brightest, most engaged, most curious students drew blanks when I mentioned the word.

Then, I learned that one of the more socially awkward but bold kids decided that their group name should be “Britannia.” Sincerely. I think the rest of the kids in his group just decided it wasn’t worth a fight. and they were probably right. socially.

And Monday is the national day for Truth and Reconciliation.

Also, I had assigned their first oral presentation for Tuesday, and wanted to spend a chunk of time preparing them for that on Monday.

I started talking with ChatGPT about this on my commute home. I used my phone so I could talk rather than type, timing my input to coincide with good signal service when I was on the subway. But I started while sitting on one of the curvy concrete benches outside the school.

All of these threads were so connected, but my brain tends to make things over-complicated. and, as I’ve said, I often try to do too much. I was curious to see how ChatGPT might help.

It was awesome. Truly.

I was communicating with ChatGPT mostly by using the audio to text feature on the phone app. so there’s a fun gaff where it picked up on the subway announcements.

I have never had any difficulty coming up with ideas for what to do in the classroom. It’s a kind of problem-solving I love and for which I think I am particularly skilled. My challenge has always been trying not to do too much, not expecting too much of myself. My weekends, with their seemingly-expansive amount of time, have often been stressful, because my goals for my class and myself are, well, they’ve been described as unreasonably high. I end up avoiding and freezing and then feeling guilty and it doesn’t get better from there.

In contrast, this weekend, I am feeling no anxiety about Monday. I have a very concrete plan that addresses everything I wanted to address. I still have to design some specific elements, but that’ll be easily done. (by me. well, maybe with some help from ChatGPT thinking through logistics.)

More about this, much more, later. For now, I’m just enjoying noticing the lack of tension in my shoulders!




Leave a comment