Online Learning in the Second Half EP 15 - Back to School Part 2 - A Letter to Teachers Sincerely ChatGPT
In this episode, John and Jason talk about a letter for teachers sincerely from ChatGPT.
Join Our LinkedIn Group - Online Learning Podcast
Resources:- Click here to read the full letter (and prompt) for teachers from ChatGPT, as well as letters to higher ed administrators, principles, provosts, edtech vendors, instructional designers, and students.
- Ethan Mollick on LinkedIn
- Teaching with AI recommendations from OpenAI
- AI Educational Video Series from the Warton School
We use a combination of computer-generated transcriptions and human editing. Please check with the recorded file before quoting anything. Please check with us if you have any questions!
[00:00:00] John Nash: That's what they've done. That's Conamigo's Other value proposition. You can forget using a zillion AI tools. Hey, now I only pay for
[00:00:08] Clicking: two,
[00:00:10] Jason Johnston: right? Me too. Now
[00:00:13] John Nash: three now, actually three. I paid for three. I went to Claude. I paid for Claude last night. And Oh, you did. Yeah, I decided to.
Oh, now you're on to four
[00:00:21] Jason Johnston: then. What's the fourth? Oh, I thought maybe you had mid journey. Oh yeah, I do. Shoot. Huh.
[00:00:28] John Nash: Yeah. Yeah. I pay for mid journey. Chat
[00:00:31] Jason Johnston: for Conmigo mid journey.
, it's like streaming services. It all started with Netflix. It was like, ah, this is great. I cut the cable. Now I only have Netflix. And now I've got
[00:00:43] John Nash: six of them.
Intro
[00:00:44] John Nash: I'm John Nash here with Jason Johnston.
[00:00:47] Jason Johnston: Hey, John. Hey, everyone. And this is Online Learning in the Second Half, the online learning podcast.
[00:00:52] John Nash: Yeah, we are doing this podcast to let you in on a conversation we've been having for the last two and a half years about online education. Look, online learning's had its chance to be great, and much of it is, but there's still some that isn't. And how are we going to get to the next stage?
[00:01:09] Jason Johnston: That is a great question. How about we do a podcast and talk about it?
[00:01:13] John Nash: That's perfect. What do you want to talk about today?
[00:01:17] Jason Johnston: I would love to talk about Back to School, Part Two. It sounds a little bit like an 80s comedy movie or something. Back to School, Part Two.
[00:01:25] John Nash: That's right, except Rodney Dangerfield's not in it.
[00:01:27] Jason Johnston: I'm glad you got the reference exactly. No, no respect here comes to Dangerfield but yeah, let's talk about Back to School, Part Two. I wanted to throw out one thing to you here.
So I asked the chat, GPT to write a letter to teachers for this fall. What advice ChatGPT would you give. Teachers this fall as they, they enter the new school year.
And so I wanted to read this letter from ChatGPT okay. And wanted to get you to respond from a professor teacher angle to see what you what you think of this. So here it is. I don't know. What's a ChatGPT voice? I'll just use my normal voice, so I think it might be too much. It's
[00:02:13] John Nash: you're chat
[00:02:13] Jason Johnston: G p t, Jason,
but it feels sincere, so I'll use a sincere voice.
Sure. I think that's right. Yeah. Yeah.
Letter: Dear Teacher
[00:02:19] Jason Johnston: Dear Teacher, I hope this message finds you well. As an AI language model, I wanna share three ways in which I can contribute to your classroom experience. Firstly, I can help you create engaging lesson plans, offering diverse content, ideas and activities tailored to your students' needs and learning styles.
Secondly, I can assist in generating interactive assignments and assessments, fostering a dynamic learning environment that encourages student participation. Lastly, I can provide additional explanations and resources to support struggling students and enhance their understanding of challenging concepts. To make the most of our collaboration.
I suggest utilizing me as a versatile tool for creativity, interactivity, and personalized student support. Sincerely chat. G B T.
[00:03:13] John Nash: I think that's, I think that's pretty good. It was, and I how was it that ChatGPTwas so scoped and succinct in these things?
[00:03:23] Jason Johnston: Yeah, that's a good question.
I, it wasn't a complex prompt by any means. I did ask it for three things, and I asked it to be encouraging and to be sincere. There you go.
[00:03:33] John Nash: I think it's a, I think it's a fair introduction to teachers who may not already be thinking of using it in their daily work. And I think those three things, lesson plans, generating interactive assignments and assessments and helping to think through explanations and resources of complex topics. The response the letter says you can help support struggling students, and I've done some work with Amanda, bigger staff and a webinar we did over the summer on differentiation of lessons for students in schools and how ChatGPT can do that with mixing some design thinking style prompts into that. So I think that these are good. And I think we know that fewer teachers than we think have really played with ChatGPT or even try used it. They're aware of it, but is it part of their daily work?
Probably not by the majority. Yeah, I like it.
[00:04:30] Jason Johnston: Yeah, I thought it was fairly impressive. Now, of course, it's generative. I think that it's also been, ChatGPT I think has been reading the same things we've been reading probably. Yeah. But maybe not, I guess this is based on 2021 and before, 'cause it's chat four.
So perhaps it's really ChatGPT's own ideas about the fall.
[00:04:52] John Nash: Let me ask you this, about this letter to the teacher, Jason. Is it a hundred percent true?
[00:05:00] Jason Johnston: Yeah. That is the one thing that we still know about ChatGPT is that it can be confidently wrong. And while this is more a matter of opinion, so it's not necessarily stating any kind of factual kind of things however I wince a little bit when when one mentions learning styles without any context around that just because more modern approaches-- not that there aren't different kinds of learners, but sometimes that prompts this kind of idea that these kind of Specific learning styles and a student is gonna slot into each one of those learning styles is still a thing.
[00:05:44] John Nash: Did I read recently somewhere that learning styles have been debunked?
[00:05:49] Jason Johnston: Yeah. Yeah. A couple years ago now, actually. Not that students, people have different ways of learning, but you can't say, oh, this student is a kinesthetic learner, so they have to have tactile things to, I'm a visual work and manipulate all the time.
I'm vi I'm a visual learner, so I can, can't listen to it. I can't listen to anything you say because I'm visual
[00:06:08] John Nash: "...and this podcast is crap because I'm a visual learner."
[00:06:11] Jason Johnston: So that's in terms of untrue. I would say that's the only thing in there that kind of just grabbed me a little bit of a red flag.
And then beyond the untrue,, sometimes it feels kinda like we know that it is it's predictive language model. And but we are predictive language model sometimes if people ask us yes, if people ask us an opinion about something, it's very easy for us to drop into teacher speak.
Yes. And generating interactive assignments and assessments sounds like teachers speak a little bit. It's very easy to talk about. "Oh yeah. So I'm highly into engaging students in my class." It's, whereas when we start to unpack that, let's talk about really what that means, "engaging students."
"I'm very dynamic in my 50 minute lecture."
Yes. Yes.
"...and that's my engagement, my student engagement" So, maybe not.
[00:07:06] John Nash: "I utilize constructivist approaches."
[00:07:08] Jason Johnston: Yes. So generating interactive assignments and assessments felt a little, it feels a little teacher speak. And I would wanna unpack that a little bit before, before just jumping on board with what an interactive assignment particularly means.
By default, I think a, an assignment is interactive because it forces a student to actually hand something in. So they have now interacted, they have now read something and they've responded to it. " What do you mean by interactive as assignment?" A writing prompt is an interactive assignment.
Yes, absolutely.
" What do you mean by actually interactive?" And is there back and forth going on here? Are they interacting with other students? Are they interacting with the teacher and the content or just the content?
[00:07:51] John Nash: Yeah. Nice analysis. I think that the letter is nice, but I think it also risks oversimplifying what can get done and in, for instance, yeah, if it, "oh, it can assist me in generating interactive assignments and assessments. Interesting. I will put that in as my prompt. 'Please generate interactive assignments and assessments for my fourth grade math class.'"
That prompt may generate a lot of stuff, and that's semi-interesting, but it's a little broad and and again, lacks context. Yes. And
[00:08:22] Jason Johnston: I think, and we may fall into that pit of algorithmic bias, whereas there are a lot of teachers out there that maybe think something is interactive that's not really the next level interactive. And so it just produces this thing that a lot of teachers already think out there about what interaction looks like. Yeah. Yeah. That's good.
I wonder if that answer in generating interactive assignments and assessments is that predictive nature of ChatGPT where it sounds good. It's oh yeah, that sounds fantastic. And then it's what is it? What is it actually talking about interactive assignments, like I could imagine some things, but I don't think it knows really.
It's just putting something together that sounds should be the top of the list, right?
[00:09:12] John Nash: Does. Yeah because a user could say, please assist me in generating interactive assignments and assessments for X. Is that sufficient to, so in what role do. Should that kind of prompt be used? For me, I would think it would, I would advise teachers to use it as a starting point for a brainstorm based upon their own expertise.
'cause what we know about using these models is that they are most beneficial when they are used in the context of something you already know a lot about.
Yeah.
And so if I was, maybe a new teacher or, yeah, I wouldn't wanna rely on just what it spit out the first time there. I'd want iterate, talk to it about what it meant or give it more refined information.
This a recent tweet that I archived from Mollick talks about, I think it's quite clever actually. Treating ChatGPT and other models for practical purposes as though it were a very smart, somewhat naive, slightly alien, well-educated gig worker. And so you give it this. Context.
Context so that you have a conversation with it about what you want to achieve, not a one shot hey, generate some interactive assignments, assessments for me, for a research class or a math class.
[00:10:30] Jason Johnston: Yeah, I think that makes a lot of sense. I saw somebody post just a little bit of negative pushback about Ethan M , which is the first I had seen.
Most of us have been praising his name for the last six months.
What did you see?
And it was in the space of they thought that he overemphasized the creative abilities of ChatGPT and the person that was responding had posted something because OpenAI has PO has posted a suggestion page for educators, and it included some very long prompts from Ethan Mollick. But your example about it being a sophisticated gig worker does not assume that this is a creative worker for us. It assumes that it is more like an assistant Yes. Or a TA or a, somebody who's going to only gonna be able to move forward with some good directions. In a little bit of maybe defense of Mollick, I feel like that his his examples kind of point more towards that. A lot of direction.
[00:11:36] John Nash: Have you ever used this online service Fiver? Where you for an inexpensive fee, depending on the range of the task, you can hire a stranger to do some kind work.
Yeah. Maybe creative work. Yeah. What kinds of things did you hire people on Fiverr to do?
[00:11:50] Jason Johnston: I've did a quite a while ago when it was first going, I was, I'm a very novice programmer and I had some P H P that I needed somebody to help me with on a website.
And yeah and they were able to achieve the goals I was looking for, essentially in terms of this one page I just couldn't get to fit correctly. And just to be able to say, Hey can you fix this for me? And without having to hire a full programmer for my website.
[00:12:23] John Nash: When you went to go get that work done, think back to how you had to prompt that, right? Yeah. Person. How, was it a one shot or did you have to be thoughtful about what you wanted to have them do?
[00:12:37] Jason Johnston: I seem to recall that I treated it like a one shot, but I was thankful that with the, at least with the person that I was working with, they didn't just do it.
They asked good questions about what it is that I was really looking for, because I didn't quite, I didn't know, quite know, I think what I was asking for. I knew what I was asking for, but I didn't know the language that I needed to use.
[00:12:57] John Nash: Yeah, I think that's where we're at .; And so last time I went on Fiverr was a couple of years ago, but as an elective project before I went to a conference I was the the head of a sig, a special interest group.
And I wanted to bring some fun posters for our membership to take home with them. And so I had some graphic designers on Fiverr, designed some posters that I could print on our 11 by 17 color printer and then take them to the conference. And that was a lot of fun. But yeah, I had to describe the tone that I wanted, the messaging that was behind the theme of the conference.
And you're right, it was a back and forth a little bit. Some of the offers amongst the workers on Fiverr were very clear about what kind of information they needed so that there wasn't a lot of back and forth. 'cause that can raise the price, for instance. If you have changes or you weren't clear in the way in which you asked for your product, they have to come, they make it.
And then you look at it and you go, ah no. Good. So they'll even limit the number of revisions you get, which I think is an interesting thing with large language models. They're tireless and they'll redo the revisions forever. But to this idea of thinking about how to help our education brothers and sisters out there, as they start to use this language it's like thinking about how you task someone to do these things on Fiverr.
You have to be really clear. Or if you have a teaching assistant or a speech writer or whomever that's gotta be explained if they've never really hung around you before or is it kind of making sense across
[00:14:28] Jason Johnston: Yeah, it makes sense to me. Yeah. Yeah. And there's a joke among creatives right now that even with the AI coming on, they're like, we're gonna be fine, because the client still needs to explain what they want.
Yes.
This is the, this is both the job of the creative, but also the great frustration sometimes of creatives is that the client often has a very difficult time explaining what it is exactly that they want. And so a good creative that knows how to work with clients, knows how to prompt, get the conversation going, to ask the good questions, to specify exactly what's going on, which is, something we run into over and over again here as we're developing courses to really get a clear idea of what it is the instructor is trying to accomplish, figure out that scope, and then try to move forward on whatever it is that and that we're trying to accomplish together.
[00:15:26] John Nash: But if you do one shot then I think you, the system will presume that whatever problem you have is the one that you must solve and needs solving.
[00:15:34] Jason Johnston: Let me ask you this question then, and maybe this will get us back to what this directly, what this podcast is about, which is how does that then relate to teachers approaching classes, say this fall.
For the first time in, in full light of OpenAI ChatGPT large language models.
[00:16:03] John Nash: Instructors don't know what they don't know yet about these models. I've run across some very interesting and compelling prompts written by Pedagogues that are very powerful at helping instructors think about whether their learning objectives are any good or just even if they put in what they want to achieve in a course, that can take you through a very intelligent conversation to help you set up better learning intentions, high impact activities to hit those intentions.
And I'm not sure that's really hit out there, that kind of sophistication exists and is in within reach, very easy reach to use. So I think that's one of the implications. I think as we hit the fall and more and more instructors get exposed to the way in which you can create a lesson plan or, do some explanations or give some resources.
It can actually be a very helpful assistant in making much more active human-centered activities that learners like and want to do that will help you hit your learning goals.
[00:17:06] Jason Johnston: Yeah. Yeah. And I think you're right in terms of they just don't know what they don't know. And I wonder if it is likened to are, Pedagogical approaches, there are instructor approaches to any technologies on the first pass.
You think about what we're talking about here, which is online learning. Our kind of first pass is oh, look at how I can just stick a bunch of stuff up into the LMS. This is exciting. Next pass is let's get a little more organized and let's put it together in this kind of way.
Third pass might be, okay, how do we really take this to the next level? Make it interactive and make it human? And that might come through. Stripping down some things that might come through, finding some specific things that are really working well and expanding them or scaffolding them or breaking them apart , or figuring out a thread that kind of maybe can go through your whole course.
So I feel like I, I feel a little bit like. That we're at the first pass right now for many people. What do you think? Yeah,
[00:18:08] John Nash: I think that's a great analogy. And I think that all this, our conversation right now about instructors not knowing what they don't know may be much ado about nothing because six months from now or a year from now, the models will be very different in terms of the way we interact with them or what capabilities they have. So I, yeah, I don't know. I think you're right. And so maybe we'll iterate on what you're suggesting is there'll be another first pass coming or the next generation of instructors that decide to go into this because they didn't take it up this year they'll be engaging these models in a different way than we are right now.
[00:18:47] Jason Johnston: Yeah. And I think about too, just even in terms of my own learning and we've been talking about Ethan Mullikin will share some links. Obviously both of us follow a lot of his postings. They've got, he's got some great videos. I don't know if you've taken a look at those videos that he, and I'm assuming as his wife put out from the Wharton School just to try to get people up and going for this fall.
But he's six months down the road with his prompts. So he's trying to explain this to people. And so we're, it's like my first attempts at ai, art, it's "make this," and that's the whole prompt. And now, through a variety of means, I'm getting a lot more specific.
I figured out what language I. It responds to and how it can do certain things, and the tool's gotten more sophisticated as well. And then there are other tools out there that actually help me with my prompt, get my prompt better when I go back to mid journey to, to create the art that I'm looking for, whatever.
Yes. So I think that, yeah, I think in terms of this kind of first pass that in six months is probably gonna look really different for a lot of a lot of instructors I
[00:20:01] John Nash: I don't know what sort of the short hit is on that, except that you're right. The prompts are you're learning over time how to make your prompts more sophisticated because you're happy to do that out of necessity. I take it because you're not satisfied with the quality of the output from the model.
So you have to continue to have the, continue to have a conversation with it. Correct it, modify it, guide it, push it, just like you would a sophisticated, semi-intelligent assistant who was a human. Yeah.
[00:20:31] Jason Johnston: Yeah. And I wonder about how, like this Hattie bot now does mean that. You and I don't have to come up with a Hattie bot, somebody else has come up with a Hattie bot. Really. Even if it's just a matter of putting the prompt or creating a, some sort of a chat. Yeah, a pre-program chat guided guard, railed chat bot for us to as a Hattie bot to jump into.
And I think that we will, I think that we will find our ourselves in some various places for putting in original kind of prompts and just that kind of open-ended kind of relationship. And then also in various Levels of interaction with these bots that are, that have been highly tuned and programmed for us and for teachers that really don't either have the time or interest to really get that deep into something just like we have templates already for, for whatever it is that we're trying to do in Canvas or in, in with our syllabus or anything else.
[00:21:36] John Nash: I, yeah, for a couple of decades now, teachers have been googling how to do stuff and it's less contextualized. Yeah. That's more of a treasure hunt than it is a a directive to develop new content. But it's similar, isn't it? Yeah. Gonna be going out to look for stuff. Yeah.
. Jason, this was a great conversation across these two back to school episodes A lot of changes in front of us. I'm making a lot of changes. And I think that the AI models are going to make a lot of changes.
[00:22:09] Jason Johnston: Yeah, and teachers be looking for your letter, sincerely chat GPT coming into your mailbox soon. I'm sure chat will be writing a letter with some a little bit of guidance for this fall. So be looking for that. We'll put the, that letter. We also had chat, write a bunch of other letters to various people in education. And so we'll stick those in the notes and I'll link out to them so people can read them and respond if they would like to check those out.
But yeah, and check out online learning podcast.com. You can see the notes from this podcast and links out to various things.
[00:22:47] John Nash: What I'm concluding here is that a lot has changed over the summer particularly with generative AI, but a lot has not changed, and in a good way. Teachers are still gonna have to bring their best game to the classroom, and we need great teaching.
We need great instructional design and that will trump AI every time. And I'm looking forward to figuring out how to be a better teacher in the face of AI, and also think about how AI can support. Us in doing that work.
[00:23:21] Jason Johnston: Yeah. And as we think about it, really any new technology that comes along so much of it is. is being reflective about it, having a plan, trying things out, making mistakes, getting feedback from your students, all the things. It's really no different this fall except that we've got, large language models trained on trillions of pieces of information, backing us up and pushing us forward.
[00:23:44] John Nash: That's right. Otherwise, nothing's
[00:23:46] Jason Johnston: that is it's business as usual in this back to school time, right?
[00:23:50] John Nash: That's right. Yeah.
[00:23:52] Jason Johnston: And join us at our LinkedIn community as well. Just look us up online learning podcast. And feel free to drop us a message on LinkedIn. We'd love to hear from you. Thanks And good luck out there to all the teachers. Go get them.
[00:24:05] John Nash: Yeah. Go get 'em.
End of Episode