Latest / Azadeh and Guy on AI / Amplify, Don’t Export

Transcript

Amplify, Don’t Export · Azadeh and Guy on AI

• 0:00
  SPEAKER_01: Hi and welcome to Azadeh
• 0:02
  SPEAKER_01: and Guy on AI.
• 0:04
  SPEAKER_01: And today we have with us Nick Hughesby who find this opening very funny.
• 0:11
  SPEAKER_01: It was very coordinated.
• 0:12
  SPEAKER_01: It's very adorable.
• 0:13
  SPEAKER_01: Almost like we've done this before.
• 0:16
  SPEAKER_01: And Nick Hughesby is an Associate Professor of Literacy
• 0:19
  SPEAKER_01: and many other interesting things here at the University of Nebraska-Lincoln.
• 0:25
  SPEAKER_01: And Nick and I have been on podcasts together because we podcast on our own podcast
• 0:34
  SPEAKER_01: about academic life.
• 0:35
  SPEAKER_01: If you want, please check us out.
• 0:37
  SPEAKER_01: Not that kind of doctor.
• 0:39
  SPEAKER_01: And we're going to talk though,
• 0:41
  SPEAKER_01: not about that,
• 0:42
  SPEAKER_01: but we're going to talk about the AI and more to the point
• 0:45
  SPEAKER_01: because Nick has been really using AI in his classroom,
• 0:50
  SPEAKER_01: educating the next generation of teachers.
• 0:53
  SPEAKER_01: He's done a lot of experimenting and a lot of trying things and making mistakes
• 0:59
  SPEAKER_01: and fixing the mistakes and all of that, that we're all doing right now.
• 1:02
  SPEAKER_01: So I thought it was a great opportunity to ask some questions about that.
• 1:05
  SPEAKER_00: Yeah.
• 1:05
  SPEAKER_00: Before we start, we haven't introduced ourselves for a while.
• 1:09
  SPEAKER_01: Oh, okay.
• 1:10
  SPEAKER_00: Yeah.
• 1:11
  SPEAKER_00: Go for it.
• 1:13
  SPEAKER_00: Hello, everyone.
• 1:14
  SPEAKER_00: Hello again.
• 1:14
  SPEAKER_00: I'm Azadeh.
• 1:15
  SPEAKER_00: I'm a first-year PhD student here at University of Nebraska-Lincoln,
• 1:20
  SPEAKER_00: and Guy is my amazing PhD advisor.
• 1:23
  SPEAKER_01: That's why we're doing introductions.
• 1:24
  SPEAKER_01: Yeah.
• 1:26
  SPEAKER_01: So you can feel better about yourself.
• 1:29
  SPEAKER_01: Because I've been lonely today.
• 1:31
  SPEAKER_01: I work from home most of the day,
• 1:32
  SPEAKER_01: and now we're in the studio, so I need some positive reinforcement.
• 1:36
  SPEAKER_01: Anyway, I'm Guy Trainin.
• 1:37
  SPEAKER_01: I'm a...
• 1:39
  SPEAKER_01: I'm working with some excellent graduate students.
• 1:43
  SPEAKER_00: That's even better.
• 1:45
  SPEAKER_01: And innovative learning technologies,
• 1:49
  SPEAKER_01: mostly AI lately, not surprisingly, because of where we're at.
• 1:54
  SPEAKER_01: And so, all right.
• 1:56
  SPEAKER_01: So that's with introductions.
• 2:00
  SPEAKER_01: So let's start with something really general, kind of trying to get the lay of the land.
• 2:04
  SPEAKER_01: How are you helping your students use AI well in your classes?
• 2:10
  SPEAKER_02: So I think it depends on what you think of as well, right?
• 2:19
  SPEAKER_02: There's a lot of different ways to come at this.
• 2:23
  SPEAKER_02: And for me in particular, I think of using AI well as a way to extend our thinking.
• 2:36
  SPEAKER_02: Like the raw materials that we're working with,
• 2:39
  SPEAKER_02: the ideas, those have to come from a human being.
• 2:44
  SPEAKER_02: And if they don't, we're going to run into problems, right?
• 2:49
  SPEAKER_02: Because these consumer models of AI that we're working with,
• 2:53
  SPEAKER_02: whether it's CoPilot or Cloud or ChatGPT, they're so good at indexing.
• 2:59
  SPEAKER_02: They're so good at mapping, of holding multiple things in parallel all at the same time.
• 3:07
  SPEAKER_02: That it allows them to do a really nice job of clarifying ideas
• 3:10
  SPEAKER_02: and spotting patterns and naming strengths and weaknesses in the material
• 3:15
  SPEAKER_02: that we provide.
• 3:18
  SPEAKER_02: But they're not really good at ideas on their own, right?
• 3:25
  SPEAKER_02: They're not really great at, I think about, I've been thinking a lot about wisdom.
• 3:35
  SPEAKER_02: And current consumer AI models are really,
• 3:40
  SPEAKER_02: really good at providing and organizing information,
• 3:44
  SPEAKER_02: but they're not good at producing wisdom.
• 3:47
  SPEAKER_02: And working in pre-service teacher education,
• 3:53
  SPEAKER_02: one of the things I'm always thinking about is that pedagogical content knowledge,
• 3:58
  SPEAKER_02: which is a form of wisdom.
• 4:05
  SPEAKER_02: So when it comes to AI,
• 4:08
  SPEAKER_02: I'm kind of in the middle of,
• 4:11
  SPEAKER_02: you know,
• 4:11
  SPEAKER_02: you've got kind of the AI is going to destroy everything kind of frame,
• 4:19
  SPEAKER_02: and you've got the AI is going to lead to abundance for everyone kind of frame.
• 4:23
  SPEAKER_02: I'm kind of in the middle.
• 4:24
  SPEAKER_02: Like, it's...
• 4:27
  SPEAKER_02: It's both going to improve some stuff,
• 4:29
  SPEAKER_02: but it's also going to cause some larger problems,
• 4:33
  SPEAKER_02: and particularly when we're thinking about de-skilling, that kind of stuff.
• 4:40
  SPEAKER_02: But I'm not as worried in my courses about that existential stuff.
• 4:46
  SPEAKER_02: That's really frontier model kind of work versus these consumer spaces that...
• 4:58
  SPEAKER_02: Really are these interesting spaces to help think through and organize that information.
• 5:05
  SPEAKER_02: And so if we're using AI to amplify rather than export our thinking,
• 5:13
  SPEAKER_02: I think that's well or right now.
• 5:16
  SPEAKER_01: And I really like the fact that you're separating the frontier model
• 5:21
  SPEAKER_01: that we don't have access to.
• 5:23
  SPEAKER_01: We're far from them at least a few months, if not forever.
• 5:29
  SPEAKER_01: And the things,
• 5:30
  SPEAKER_01: the commercial products that most of us and definitely our students have access to.
• 5:35
  SPEAKER_01: Because looking at what's happening around us,
• 5:39
  SPEAKER_01: most universities are still not signing on to wholesale.
• 5:45
  SPEAKER_01: Every student will get access to all of these devices, although just recently...
• 5:52
  SPEAKER_01: Google let all students, at least in higher education, have notebooks.
• 5:57
  SPEAKER_01: Google Notebook, not Notebook LM, which is the same thing.
• 6:02
  SPEAKER_01: Is it a Gemini notebook now?
• 6:03
  SPEAKER_01: Yeah, Gemini notebook, whatever.
• 6:04
  SPEAKER_01: It's like they rebrand every like two months.
• 6:07
  SPEAKER_00: Yeah.
• 6:08
  SPEAKER_01: For no good reason, but you know, whatever.
• 6:12
  SPEAKER_01: But I'll keep complaining.
• 6:13
  SPEAKER_01: That's what I'm promising to do.
• 6:16
  SPEAKER_01: Okay.
• 6:16
  SPEAKER_01: I'll keep complaining about Google's.
• 6:19
  SPEAKER_01: But they gave this.
• 6:20
  SPEAKER_01: But institutions have not, for the most part, in the United States.
• 6:24
  SPEAKER_01: I just heard reports from other places where there is wholesale, let's give it to everybody.
• 6:29
  SPEAKER_01: So there's equity and there's high quality.
• 6:33
  SPEAKER_01: But it's still not the frontier.
• 6:34
  SPEAKER_01: So, yes, I have serious existential worries about the frontier, but this is not what I'm using.
• 6:42
  SPEAKER_01: I'm using, you know.
• 6:43
  SPEAKER_01: Claude, Opus 5.5 or whatever it is.
• 6:46
  SPEAKER_01: And it is fantastic, but it is not the frontier.
• 6:49
  SPEAKER_01: It is not breaking into websites and doing things.
• 6:53
  SPEAKER_01: As far as I know, at least mine isn't.
• 6:55
  SPEAKER_01: Maybe when I go to sleep, it does things.
• 6:58
  SPEAKER_02: Don't give it things to do
• 6:59
  SPEAKER_01: when you go to sleep.
• 7:00
  SPEAKER_02: I'm like, here's all the things I need you to do by the time I wake up in the morning
• 7:04
  SPEAKER_02: and drink my coffee.
• 7:05
  SPEAKER_01: Go.
• 7:06
  SPEAKER_01: All right.
• 7:07
  SPEAKER_01: So when...
• 7:10
  SPEAKER_01: When a pre-service teacher uses AI to draft a lesson plan,
• 7:14
  SPEAKER_01: at what point does it stop being their plan?
• 7:16
  SPEAKER_00: Yeah, I was going to ask that question.
• 7:18
  SPEAKER_00: Okay.
• 7:23
  SPEAKER_03: So I think when...
• 7:25
  SPEAKER_02: I think that authorship shifts the moment that ideas originate with AI and not with the human.
• 7:37
  SPEAKER_02: I think if...
• 7:40
  SPEAKER_02: You just say,
• 7:44
  SPEAKER_02: draft me a lesson on fluency instruction and there's no input from you,
• 7:53
  SPEAKER_02: that's not you authoring anything.
• 7:56
  SPEAKER_02: And I want to differentiate two things there before I get too far into
• 8:01
  SPEAKER_03: this answer
• 8:02
  SPEAKER_03: because I do think that we have to
• 8:10
  SPEAKER_03: Rethink authorship in an age of AI.
• 8:15
  SPEAKER_03: I think we need to
• 8:18
  SPEAKER_03: stop the
• 8:21
  SPEAKER_02: witch hunt for,
• 8:25
  SPEAKER_02: oh my gosh, I saw two emdashes in this peer review.
• 8:28
  SPEAKER_02: He used ChatGPT.
• 8:31
  SPEAKER_02: No, I just know grammar.
• 8:33
  SPEAKER_02: Like, why am I being punished for knowing grammatical structures and punctuation?
• 8:40
  SPEAKER_02: But we're living in,
• 8:44
  SPEAKER_02: particularly in academia,
• 8:47
  SPEAKER_02: and I think in schools to some extent,
• 8:49
  SPEAKER_02: of how do we stop students from just using AI to do their essays?
• 9:00
  SPEAKER_02: And we're so caught up in that question that we're not necessarily thinking
• 9:06
  SPEAKER_02: about How could we reinvent that?
• 9:09
  SPEAKER_02: How could we rethink that?
• 9:11
  SPEAKER_02: What does that look like now?
• 9:16
  SPEAKER_02: And I mean, we had the same conversations when Spellcheck came out or Grammarly came out.
• 9:22
  SPEAKER_02: These are all,
• 9:24
  SPEAKER_03: I mean,
• 9:25
  SPEAKER_03: not to the same extent of AI,
• 9:27
  SPEAKER_03: but
• 9:30
  SPEAKER_03: we have to think through as a field,
• 9:35
  SPEAKER_02: what does authorship look
• 9:36
  SPEAKER_01: like now?
• 9:38
  SPEAKER_01: So let me take it to something very concrete.
• 9:41
  SPEAKER_01: In your class, how do you think about that and how do you communicate that to students?
• 9:46
  SPEAKER_02: Well, so instead of saying, you know, draft me a lesson plan on figurative language, go, right?
• 9:55
  SPEAKER_02: Like the first step to anything is here's the template.
• 10:02
  SPEAKER_02: I'm going to make you fill it out in class, in pen, in pencil.
• 10:06
  SPEAKER_02: You take a picture of it.
• 10:09
  SPEAKER_02: You upload it, and then you have AI help stress test it for you, right?
• 10:18
  SPEAKER_02: Because there, you're using AI a bit like a dictionary, right?
• 10:25
  SPEAKER_02: If it's looking at what you're saying you're going to model,
• 10:29
  SPEAKER_02: and this is really hard for beginning teachers,
• 10:32
  SPEAKER_02: like thinking through the defossilization of something That's pretty sophisticated,
• 10:39
  SPEAKER_02: like reading or writing.
• 10:43
  SPEAKER_02: It's really hard for them to think through, what are the steps here?
• 10:46
  SPEAKER_02: Because they've done it for so long and so naturally,
• 10:50
  SPEAKER_02: quote unquote, but they can't think through the steps.
• 10:53
  SPEAKER_02: And Chagibiti or Copilot or what have you is a little better at thinking through,
• 11:02
  SPEAKER_02: oh, I see what you're doing here.
• 11:03
  SPEAKER_02: Do these steps make sense?
• 11:05
  SPEAKER_02: What else could go there?
• 11:08
  SPEAKER_02: And so you're really using it as,
• 11:11
  SPEAKER_02: I like to think of forcing them to use it as like a writing coach and not as
• 11:16
  SPEAKER_02: a ghostwriter.
• 11:17
  SPEAKER_02: Like we are not Britney Spears trying to write our memoirs.
• 11:20
  SPEAKER_02: Like that's not how that's working.
• 11:24
  SPEAKER_02: But it's always a question of like, what's the function, right?
• 11:31
  SPEAKER_02: And all of the research that's coming out around AI and AI usage is pointing
• 11:37
  SPEAKER_02: to It's really effective,
• 11:41
  SPEAKER_02: or I shouldn't say really effective, that's the wrong quantifier, qualifier.
• 11:46
  SPEAKER_02: It's more effective when you've done the thinking first.
• 11:54
  SPEAKER_02: When you have something that you've done, and if we're talking about learning as...
• 12:02
  SPEAKER_02: Sustained practices that transcend a collection of individual experiences
• 12:08
  SPEAKER_02: and courses with those ideas,
• 12:11
  SPEAKER_02: that impacts learning more than do it for me, AI, right?
• 12:17
  SPEAKER_02: And it changes that authorship.
• 12:20
  SPEAKER_02: So if you're handing me a ChatGPT-designed
• 12:25
  SPEAKER_02: lesson plan, I'm going to know, right?
• 12:29
  SPEAKER_02: Because again, ChatGPT is not great.
• 12:32
  SPEAKER_02: At certain things.
• 12:32
  SPEAKER_02: It's not great at wisdom.
• 12:37
  SPEAKER_02: And so where's all this stuff from class?
• 12:39
  SPEAKER_02: Where are the things that we're talking through in courses?
• 12:45
  SPEAKER_02: Like we did...
• 12:47
  SPEAKER_02: So modeling was one example.
• 12:49
  SPEAKER_02: Another example might be like child-friendly definitions for vocabulary words.
• 12:53
  SPEAKER_02: Incredibly hard.
• 12:55
  SPEAKER_02: Yeah.
• 12:56
  SPEAKER_02: Right?
• 12:56
  SPEAKER_02: And so we spent a lot of time internalizing.
• 12:59
  SPEAKER_02: I think it's Louisa Motes who has a really nice four-step process, a heuristic.
• 13:06
  SPEAKER_02: You start with the synonym and then you think about what makes the synonym,
• 13:10
  SPEAKER_02: what makes the word different from the synonym and where is that la la.
• 13:16
  SPEAKER_02: And so we do that in class.
• 13:17
  SPEAKER_02: And then we ask ChatGPT to help us think through,
• 13:23
  SPEAKER_02: what are we missing in this child-friendly definition?
• 13:30
  SPEAKER_02: And it really gets at that notion of teacher as curator, teacher as someone who...
• 13:42
  SPEAKER_02: Is really spending time thinking through what is the function of AI aside from,
• 13:48
  SPEAKER_02: oh, it's going to help me get it done faster.
• 13:51
  SPEAKER_02: So if they hand in a lesson plan that's just ChatGPT, I'm going to say, I need your lesson plan.
• 14:00
  SPEAKER_02: This is not your lesson plan.
• 14:02
  SPEAKER_02: It happens to have your name on it, but it has none of your thinking.
• 14:06
  SPEAKER_02: It has none of what we've been talking about in class.
• 14:12
  SPEAKER_02: That's not okay.
• 14:14
  SPEAKER_02: There has to be something of their thinking there.
• 14:19
  SPEAKER_02: And so I spend time modeling that.
• 14:23
  SPEAKER_02: Every assignment comes with kind of AI guidelines.
• 14:26
  SPEAKER_02: Here's how you can use it.
• 14:28
  SPEAKER_02: Here's what I want you to avoid.
• 14:31
  SPEAKER_02: And so I haven't really come across...
• 14:37
  SPEAKER_02: A lot of lesson plans that were flat out AI because of that.
• 14:44
  SPEAKER_00: Can I ask a question?
• 14:45
  SPEAKER_00: I really want to expand on that question because we recently received a lot
• 14:50
  SPEAKER_00: of questions from students.
• 14:51
  SPEAKER_00: I'm co-teaching a course with And students ask a lot of questions
• 14:56
  SPEAKER_00: about how often do we need to use AI?
• 14:59
  SPEAKER_00: Do we need to use AI for everything when we are doing the lesson planning?
• 15:03
  SPEAKER_00: So I'm curious, how do you guide the students for their interactions with generative AI?
• 15:09
  SPEAKER_00: Do you ask them to give AI a persona?
• 15:11
  SPEAKER_00: For example, you are a teacher with 20 years of teaching experience.
• 15:16
  SPEAKER_00: Help me improve my lesson plan or something like that.
• 15:19
  SPEAKER_00: Or how does it work?
• 15:21
  SPEAKER_02: So I typically have them give the, most of them use ChatGPT.
• 15:28
  SPEAKER_02: So I make it ChatGPT specific, but I give them the profile of an instructional coach.
• 15:34
  SPEAKER_02: So I've built the profile.
• 15:36
  SPEAKER_02: I've used it in multiple courses around like, you're an instructional coach.
• 15:40
  SPEAKER_02: Here's the things you know.
• 15:41
  SPEAKER_02: Boom, boom, boom, boom, boom, boom.
• 15:43
  SPEAKER_02: Here's how you give feedback.
• 15:44
  SPEAKER_02: Boom, boom, boom, boom, boom, boom.
• 15:47
  SPEAKER_02: And that's what they feed.
• 15:54
  SPEAKER_02: And so then when
• 15:58
  SPEAKER_02: they go in and they ask, talk to me about this as a child-friendly definition.
• 16:07
  SPEAKER_02: Where might kids get tripped up as they're learning the meaning of that new word?
• 16:13
  SPEAKER_02: And it doesn't necessarily do the work for them,
• 16:17
  SPEAKER_02: but it will lead them to Here's what I'm noticing.
• 16:22
  SPEAKER_02: Here's what could be the trick.
• 16:24
  SPEAKER_02: What might be the address to that, right?
• 16:28
  SPEAKER_02: So it's really meant to be more dialogic and thoughtful than a straight-out answer.
• 16:37
  SPEAKER_02: And I've found with, like, if I had a persona that was, you're a 20-year veteran teacher,
• 16:46
  SPEAKER_02: That feedback did not always go as well.
• 17:07
  SPEAKER_02: It was a lot of,
• 17:08
  SPEAKER_02: you did this wrong,
• 17:08
  SPEAKER_02: this is how kids are going to be, this is how kids are going to be here.
• 17:11
  SPEAKER_02: It wasn't that same back and forth where I really wanted that coaching piece to happen.
• 17:21
  SPEAKER_02: So I provide that to them,
• 17:23
  SPEAKER_02: they upload it into their chat GPT, and that's the persona that chat exemplifies.
• 17:30
  SPEAKER_00: Mm-hmm.
• 17:31
  SPEAKER_00: Awesome.
• 17:32
  SPEAKER_00: Yeah, thank you.
• 17:33
  SPEAKER_00: And do you model that in the classroom in front of students?
• 17:36
  SPEAKER_00: I mean, just walking them through those steps or?
• 17:39
  SPEAKER_02: Yeah, it's the first time that we do it,
• 17:42
  SPEAKER_02: I show them how to upload the profile, you know, just adding the file.
• 17:46
  SPEAKER_02: And then chat goes, oh, yep, this is what I'm going to do.
• 17:51
  SPEAKER_02: And then we talk a little bit about prompt engineering to give some sense of like,
• 17:59
  SPEAKER_02: Here's the function of this.
• 18:00
  SPEAKER_02: Here's the outcome I need.
• 18:01
  SPEAKER_02: Here's the context.
• 18:04
  SPEAKER_02: And here's kind of the background information that you might need now.
• 18:06
  SPEAKER_02: Here's what I need feedback on.
• 18:09
  SPEAKER_02: And it goes.
• 18:10
  SPEAKER_02: And the persona is written in such a way that it'll prompt for those things
• 18:14
  SPEAKER_02: if students don't offer them up in the first place.
• 18:16
  SPEAKER_02: It'll be like, oh, hey, I have no clue about the context here.
• 18:20
  SPEAKER_02: Help me out.
• 18:23
  SPEAKER_02: Which is always helpful, right?
• 18:26
  SPEAKER_02: But it helps,
• 18:27
  SPEAKER_02: I have found that that instructional coach piece helps surface things
• 18:30
  SPEAKER_02: that they haven't thought of.
• 18:34
  SPEAKER_02: And again, that curation part becomes really, really important.
• 18:40
  SPEAKER_03: Yeah, thank you.
• 18:43
  SPEAKER_00: Oh, next question.
• 18:47
  SPEAKER_00: Let me see.
• 18:48
  SPEAKER_00: So what are the most exciting things you are hearing about AI
• 18:54
  SPEAKER_00: from pre-service teachers or even from your colleagues?
• 18:58
  SPEAKER_03: What are the most exciting things I'm hearing?
• 19:03
  SPEAKER_02: I mean, here's the kinds of exciting things that I get really excited by.
• 19:13
  SPEAKER_03: When AI is leveraged to do something
• 19:20
  SPEAKER_02: much more efficiently or much more smoothly.
• 19:26
  SPEAKER_02: So, for instance,
• 19:31
  SPEAKER_02: I now have an AI bot thanks to a friend who manages my Zotero.
• 19:38
  SPEAKER_02: It goes through.
• 19:40
  SPEAKER_02: Checks to make sure there aren't any duplicate entries every once in a while.
• 19:45
  SPEAKER_02: It'll check my downloads folder.
• 19:46
  SPEAKER_02: And if those aren't already imported, it will import them for me.
• 19:50
  SPEAKER_02: Just a second.
• 19:50
  SPEAKER_02: I think I need to leave and make an agent.
• 19:54
  SPEAKER_02: I mean, look, it's a beautiful thing.
• 19:56
  SPEAKER_02: It's really a beautiful thing.
• 19:59
  SPEAKER_02: And that is something that would take me time.
• 20:02
  SPEAKER_00: Oh, sure.
• 20:03
  SPEAKER_02: Yeah, I mean.
• 20:04
  SPEAKER_02: It would take me time and not necessarily thinking time.
• 20:07
  SPEAKER_00: Yes.
• 20:09
  SPEAKER_02: That kind of stuff I love.
• 20:10
  SPEAKER_02: I mean, the joke is always, where's my AI who's going to fold my laundry?
• 20:16
  SPEAKER_02: And that feels like we're several years off.
• 20:19
  SPEAKER_02: But anything that will take something that gives me time back to do thinking work,
• 20:32
  SPEAKER_02: to do intellectual work is super exciting.
• 20:36
  SPEAKER_02: Yeah.
• 20:38
  SPEAKER_02: I've been playing a lot with APIs in Canvas and trying to think through how
• 20:44
  SPEAKER_02: to make it help me run a class better.
• 20:49
  SPEAKER_02: And that's been, there have been some really good things there.
• 20:52
  SPEAKER_02: There have been some not so good things there.
• 20:54
  SPEAKER_02: But those are the kinds of things that get me excited because if it's like coding,
• 21:00
  SPEAKER_02: that's above my pay range.
• 21:03
  SPEAKER_02: And if it's like,
• 21:04
  SPEAKER_02: give me recommendations for the best sweater I can buy for
• 21:08
  SPEAKER_02: the upcoming sweater season, which seems like we had it in, now we don't.
• 21:13
  SPEAKER_02: That's just Google.
• 21:14
  SPEAKER_02: Like, why are you doing, like,
• 21:16
  SPEAKER_02: why are you using Chatty or Claude to do something you could simply ask
• 21:23
  SPEAKER_02: a search engine to do, right?
• 21:25
  SPEAKER_02: Like that, don't want to do that.
• 21:28
  SPEAKER_02: But if it can help me do something more efficiently.
• 21:32
  SPEAKER_03: Absolutely, yeah.
• 21:33
  SPEAKER_02: I'm totally down for it.
• 21:35
  SPEAKER_02: Love that part of it.
• 21:37
  SPEAKER_02: So those are the kinds of things that keep me excited.
• 21:42
  SPEAKER_01: And so AI is changing,
• 21:44
  SPEAKER_01: and you talked a little bit about this when you talked about lesson plans
• 21:47
  SPEAKER_01: and authorship, but it changes what it means to read or to write.
• 21:52
  SPEAKER_01: Because the minute we have an AI there that is summarizing,
• 21:57
  SPEAKER_01: and you have to prompt,
• 21:58
  SPEAKER_01: and you talked about prompt engineering a little bit,
• 22:01
  SPEAKER_01: although I personally don't like the engineering part of it.
• 22:04
  SPEAKER_01: You're just prompting.
• 22:05
  SPEAKER_01: Why does it have to be engineering?
• 22:07
  SPEAKER_01: Well, you know why.
• 22:09
  SPEAKER_01: It's fundable.
• 22:10
  SPEAKER_01: Yes.
• 22:11
  SPEAKER_01: And it's fundable and it's catchy, like vibe coding.
• 22:16
  SPEAKER_02: I'm engineering.
• 22:17
  SPEAKER_02: I'm engineering my prompts.
• 22:20
  SPEAKER_01: And so you've got to do that.
• 22:22
  SPEAKER_01: You've got to evaluate the output, which is probably the hardest thing.
• 22:26
  SPEAKER_01: And that's where most of us or most things go wrong is when you don't evaluate.
• 22:31
  SPEAKER_01: And then you submit something that has factually wrong things and a You know,
• 22:39
  SPEAKER_01: references that don't exist and all of those things.
• 22:41
  SPEAKER_01: And also how comfortable you feel in revising some text that was created by AI
• 22:48
  SPEAKER_01: because the style may be very different than yours.
• 22:50
  SPEAKER_01: And so you suddenly,
• 22:52
  SPEAKER_01: you find yourself,
• 22:53
  SPEAKER_01: and I found myself a few times,
• 22:55
  SPEAKER_01: like I've got something from AI and I'm trying to put it in.
• 22:58
  SPEAKER_01: And then we go over it a few times and we basically change it completely because
• 23:02
  SPEAKER_01: that tone, something there.
• 23:05
  SPEAKER_01: You know, doesn't mesh with the rest of the tone of the piece.
• 23:07
  SPEAKER_01: So it was a good solution for a while, but it actually didn't work.
• 23:12
  SPEAKER_01: So what should we teach here when we think about reading and writing?
• 23:16
  SPEAKER_01: We're thinking about,
• 23:17
  SPEAKER_01: you know, we work in elementary schools, but we're also thinking about the teachers.
• 23:21
  SPEAKER_01: What should we teach them?
• 23:24
  SPEAKER_02: Is this kind of getting at like a critical literacy,
• 23:27
  SPEAKER_01: literacy's frame?
• 23:29
  SPEAKER_01: It could get to new literacies.
• 23:31
  SPEAKER_01: With critical literacy,
• 23:32
  SPEAKER_01: critical literacy I find always difficult because it's a very wide concept.
• 23:37
  SPEAKER_01: Right.
• 23:39
  SPEAKER_02: And I think, again, just as we need to rethink things with authorship,
• 23:46
  SPEAKER_02: I think we need to rethink One,
• 23:50
  SPEAKER_02: attaching literacies to everything because financial literacy,
• 23:54
  SPEAKER_02: sound literacy, soda literacy, AI literacy.
• 23:59
  SPEAKER_02: It's a literacy palooza and it means nothing.
• 24:02
  SPEAKER_02: And that's just incredibly frustrating.
• 24:05
  SPEAKER_02: But aside from my personal gripe, which thanks for entertaining my personal gripe.
• 24:10
  SPEAKER_01: We have soapboxes here.
• 24:12
  SPEAKER_01: I have nothing but soapboxes.
• 24:14
  SPEAKER_01: It sometimes feels like...
• 24:17
  SPEAKER_02: But I think we're at this critical juncture where
• 24:24
  SPEAKER_02: the author here has changed.
• 24:27
  SPEAKER_02: When we're talking about AI, the author has changed.
• 24:30
  SPEAKER_02: And for most of human history,
• 24:32
  SPEAKER_02: at least since the invention of the written word,
• 24:37
  SPEAKER_02: the text has been a connection between two people, right?
• 24:43
  SPEAKER_02: And I'm thinking a lot about Harari's Nexus, which Guy got me turned on to.
• 24:49
  SPEAKER_02: Fabulous book if you have not read it.
• 24:52
  SPEAKER_02: Still contemplating it in my brain.
• 24:56
  SPEAKER_02: And when we're thinking about an AI written text,
• 25:02
  SPEAKER_02: when AI is on the other end of that connection, that's a non-human member.
• 25:11
  SPEAKER_02: Like it's not just...
• 25:15
  SPEAKER_02: It's the fact that it was created by an algorithm that none of us understand.
• 25:26
  SPEAKER_02: I don't even know that the engineers at Anthropic understand it.
• 25:31
  SPEAKER_02: And that's categorically different than anything else that we've encountered before.
• 25:40
  SPEAKER_02: Traditional critical literacy,
• 25:42
  SPEAKER_02: regardless of what model you're relying on,
• 25:47
  SPEAKER_02: assumes a human author and assumes intentions and assumes some understanding of
• 25:56
  SPEAKER_02: the human condition around which that text was created that you can interrogate.
• 26:04
  SPEAKER_02: And it works because Those author's tools are shared, right?
• 26:11
  SPEAKER_02: We all have access to that same grammar.
• 26:13
  SPEAKER_02: We have access to the events or the histories or what have you.
• 26:20
  SPEAKER_02: Sometimes not as fully as the author might, but in swaths, right?
• 26:26
  SPEAKER_02: We can get there, right?
• 26:28
  SPEAKER_02: Syntax, rhetorical moves.
• 26:31
  SPEAKER_02: And we can ask, really importantly,
• 26:35
  SPEAKER_03: who wrote this?
• 26:37
  SPEAKER_03: What individual?
• 26:38
  SPEAKER_02: And what are their intentions?
• 26:42
  SPEAKER_02: And what do they stand to gain if you think about that piece of writing as persuasion?
• 26:54
  SPEAKER_02: And with an algorithm, it's not comprehensible that way.
• 27:01
  SPEAKER_02: We don't know what's actually at work.
• 27:04
  SPEAKER_02: And so that opacity is really part of the frustration right now, right?
• 27:10
  SPEAKER_02: Like that notion of do we actually know what's happening?
• 27:16
  SPEAKER_02: And so we can't just extend these existing models because they assume a legibility
• 27:23
  SPEAKER_02: that just is not there.
• 27:25
  SPEAKER_02: It's just not present.
• 27:28
  SPEAKER_02: It feels like this is an opportunity for people Some critical practices for texts
• 27:35
  SPEAKER_02: whose authors are AI.
• 27:38
  SPEAKER_02: This isn't a spot where we can retrofit.
• 27:40
  SPEAKER_02: This is not a remodel.
• 27:41
  SPEAKER_02: This is a new build.
• 27:44
  SPEAKER_02: And that's hard because it's changing so fast.
• 27:51
  SPEAKER_03: But it feels like that's what we have to do
• 28:01
  SPEAKER_02: Particularly if people are outsourcing their thinking to the AI,
• 28:05
  SPEAKER_02: and that is all what they're doing there, how do they member check that?
• 28:13
  SPEAKER_02: How do they know to member check that?
• 28:17
  SPEAKER_02: And that's a kind of separate problem.
• 28:21
  SPEAKER_02: Yeah, and I think about this
• 28:23
  SPEAKER_01: in a way.
• 28:24
  SPEAKER_01: I think about...
• 28:28
  SPEAKER_01: The work you do when you write.
• 28:30
  SPEAKER_01: When you write,
• 28:32
  SPEAKER_01: it becomes really hard to edit yourself after a while because You have written this.
• 28:38
  SPEAKER_01: It made sense in your brain somewhere and you've worked really hard.
• 28:42
  SPEAKER_01: So there's also effort there.
• 28:44
  SPEAKER_01: But you definitely, it becomes really hard for you to critique.
• 28:48
  SPEAKER_01: So one of the things that we do in writing,
• 28:51
  SPEAKER_01: and we write together quite a bit,
• 28:53
  SPEAKER_01: we wrote together too,
• 28:54
  SPEAKER_01: but differently,
• 28:55
  SPEAKER_01: is we read it out loud because that creates that distance that allows you
• 28:59
  SPEAKER_01: to hear what works and what doesn't work.
• 29:02
  SPEAKER_01: But I think with AI, it works the same way for a different reason.
• 29:06
  SPEAKER_01: It works the same way because the probabilities, because it's a probabilities game.
• 29:12
  SPEAKER_01: And it always chooses something that is fairly probable.
• 29:15
  SPEAKER_01: That's why it sounds so smooth,
• 29:17
  SPEAKER_01: because it's like we could have predicted that word, that next word.
• 29:20
  SPEAKER_01: So it goes, it almost...
• 29:23
  SPEAKER_01: Is effortless in that way and therefore becomes really hard
• 29:26
  SPEAKER_01: to figure out what's gone wrong there.
• 29:28
  SPEAKER_01: You need to take a tremendous amount of care to figure out where AI went wrong and
• 29:33
  SPEAKER_01: to go,
• 29:33
  SPEAKER_01: wait a minute,
• 29:34
  SPEAKER_01: this seems really fluent, but now that I'm looking at it, is it saying anything?
• 29:41
  SPEAKER_01: Yep.
• 29:42
  SPEAKER_01: And our brains are built to do exactly that, to predict what the next word is.
• 29:46
  SPEAKER_01: And, you know, I love poetry for a variety of reasons.
• 29:50
  SPEAKER_01: But in poetry,
• 29:51
  SPEAKER_01: the whole point is to surprise,
• 29:53
  SPEAKER_01: in many ways,
• 29:53
  SPEAKER_01: to surprise your listener and to give them something where they go, wait, what?
• 29:59
  SPEAKER_01: And that's not what AI does.
• 30:01
  SPEAKER_02: That's not what AI can do, right?
• 30:05
  SPEAKER_02: It's not very good at...
• 30:09
  SPEAKER_02: Surprising us outside of like, when it's like, leave your
• 30:12
  SPEAKER_03: wife, like that whole story.
• 30:15
  SPEAKER_02: But in terms of the writing, it's really kind of a mediocre model, right?
• 30:22
  SPEAKER_02: When we're thinking about it.
• 30:24
  SPEAKER_02: And so if we move into new authorship models,
• 30:29
  SPEAKER_02: like what does it look like when your Chatty or your Claude
• 30:38
  SPEAKER_02: Is an extension of that.
• 30:39
  SPEAKER_02: Like, what does that work look like?
• 30:43
  SPEAKER_02: So that it doesn't sound uber generic, right?
• 30:48
  SPEAKER_02: And so much of that is around meta language,
• 30:50
  SPEAKER_02: around how you like to construct an argument,
• 30:53
  SPEAKER_02: that kinds of language you like to use,
• 30:55
  SPEAKER_02: the kinds of sentence structures you like to use,
• 30:58
  SPEAKER_02: the journeys you want readers to go on that don't often match up.
• 31:06
  SPEAKER_01: And just to take it to, again,
• 31:12
  SPEAKER_01: to writing,
• 31:13
  SPEAKER_01: because I think this comes to light mostly about writing,
• 31:17
  SPEAKER_01: because that's what AIs do, is they're very good at writing and also writing code.
• 31:22
  SPEAKER_01: But those are the two things they started doing immediately.
• 31:25
  SPEAKER_01: And smoothly enough that it has become a problem in teaching.
• 31:30
  SPEAKER_01: But the way I think about it is for many years,
• 31:33
  SPEAKER_01: our strategy in teaching writing,
• 31:35
  SPEAKER_01: especially in quite a bit in elementary,
• 31:38
  SPEAKER_01: even more so in secondary, was here's a template, write me a five-paragraph essay.
• 31:43
  SPEAKER_01: And then everything tends to sound the same, which is exactly AI.
• 31:48
  SPEAKER_01: And what we have to do is really what a lot of college writing teachers talk
• 31:53
  SPEAKER_01: about is let's take the template out and let's write it ourselves with a lot less,
• 32:02
  SPEAKER_01: you know, this is how you do this and this is how you need to sign to sound.
• 32:07
  SPEAKER_01: But that means that it forces us to break practices that we've had for a long time.
• 32:11
  SPEAKER_01: Well, and let's be real that that practice
• 32:13
  SPEAKER_02: of like,
• 32:14
  SPEAKER_02: You use this template for,
• 32:16
  SPEAKER_02: I don't know, what, 10 years in schools from grade two on?
• 32:22
  SPEAKER_02: Like, that's bad practice.
• 32:24
  SPEAKER_02: Yes.
• 32:24
  SPEAKER_01: It is bad practice, but it is happening.
• 32:26
  SPEAKER_02: It's a Gotskian tool.
• 32:28
  SPEAKER_02: Yeah.
• 32:29
  SPEAKER_02: We should be reducing it so that we finally get at the notion of what is this model for?
• 32:37
  SPEAKER_02: It's for making meaning.
• 32:38
  SPEAKER_02: And this is one way to make meaning.
• 32:42
  SPEAKER_02: So I think part of the pushback around AI,
• 32:47
  SPEAKER_02: particularly in education,
• 32:49
  SPEAKER_02: is it does not care that your bad pedagogy was how you were running your class.
• 32:58
  SPEAKER_02: Doesn't care.
• 32:59
  SPEAKER_02: Doesn't care.
• 33:02
  SPEAKER_02: And it's making that very apparent.
• 33:04
  SPEAKER_02: It's making it very clear that
• 33:09
  SPEAKER_02: Oh, AI can reproduce the student writing that your students have produced?
• 33:14
  SPEAKER_02: Well, then was it very good?
• 33:16
  SPEAKER_02: Like, was that worth it?
• 33:19
  SPEAKER_02: And it's really pushing on that.
• 33:22
  SPEAKER_02: Yeah.
• 33:23
  SPEAKER_02: And I think that is a thing that, as professionals, we need to grapple with.
• 33:31
  SPEAKER_02: Like, because that's what it's doing.
• 33:33
  SPEAKER_02: It's exposing some really uncomfortable truths.
• 33:36
  SPEAKER_03: Mm-hmm.
• 33:37
  SPEAKER_01: And just to,
• 33:39
  SPEAKER_01: and we're almost out of time,
• 33:41
  SPEAKER_01: so I think that it is important to recognize that some of it is driven,
• 33:47
  SPEAKER_01: as has been shown many times, by assessment time.
• 33:50
  SPEAKER_01: So if the assessment has a series of expectations and you give an example,
• 33:55
  SPEAKER_01: Then all the teachers preparing kids to pass that specific assessment are going
• 34:01
  SPEAKER_01: to basically have them practice that very thing that is the example
• 34:06
  SPEAKER_01: of how you do well on the exam.
• 34:08
  SPEAKER_01: So there is a need also to go back and say,
• 34:13
  SPEAKER_01: if AI can do this assessment that well,
• 34:16
  SPEAKER_01: is this a Really good assessment of literacy.
• 34:20
  SPEAKER_01: It can be a very good assessment of math.
• 34:22
  SPEAKER_01: I think that certain skills in math are skills in math, much like spelling, right?
• 34:27
  SPEAKER_01: You need to learn this,
• 34:28
  SPEAKER_01: and there's a way to test it,
• 34:30
  SPEAKER_01: and AI will always be fairly good at it,
• 34:32
  SPEAKER_01: despite the early struggles that AI had with exactly those kinds of problems.
• 34:37
  SPEAKER_01: But that was a solvable problem.
• 34:40
  SPEAKER_01: But the minute we get to skills like writing and thinking through writing,
• 34:45
  SPEAKER_01: That's a whole different thing.
• 34:47
  SPEAKER_01: And I would argue reading comprehension as well.
• 34:50
  SPEAKER_01: And that's a whole different thing.
• 34:52
  SPEAKER_01: And that's where we need to also rethink some of our assessment.
• 34:58
  SPEAKER_02: Yeah.
• 34:58
  SPEAKER_02: And we need to rethink some of our pedagogical choices just writ large, right?
• 35:05
  SPEAKER_02: Like, how do we use AI to help support that comprehension in ways that are useful?
• 35:15
  SPEAKER_02: And we've got to get our heads wrapped around that.
• 35:20
  SPEAKER_02: Because all of this is now, I mean, it's, to quote Aladdin, it's a whole new world.
• 35:28
  SPEAKER_02: And we've got to take that magic carpet ride.
• 35:32
  SPEAKER_01: Yes.
• 35:33
  SPEAKER_01: And we also, unlike Aladdin, we don't have a choice about this.
• 35:37
  SPEAKER_01: The magic carpet ride is here and we're all on it,
• 35:40
  SPEAKER_01: even if we close our eyes and say it's not here.
• 35:43
  SPEAKER_02: Right.
• 35:43
  SPEAKER_02: Like you're...
• 35:45
  SPEAKER_02: People who say, oh, I don't use AI because of, you know, a variety of different reasons.
• 35:52
  SPEAKER_02: And be what they may.
• 35:55
  SPEAKER_02: Environmental impact, et cetera, et cetera, et cetera.
• 35:59
  SPEAKER_02: It's here in ways that we don't fully always become aware of, right?
• 36:08
  SPEAKER_02: And so...
• 36:10
  SPEAKER_02: It's already at play in our lives.
• 36:14
  SPEAKER_02: Whether it's, I logged into my banking account this morning to be like, how poor am I now?
• 36:21
  SPEAKER_02: And it was like, oh, hi, I'm the AI assistant.
• 36:24
  SPEAKER_02: Can I help you?
• 36:24
  SPEAKER_02: No, absolutely not.
• 36:26
  SPEAKER_02: I don't need you anywhere near any of my financial information.
• 36:29
  SPEAKER_02: And then I sent an email that was like,
• 36:31
  SPEAKER_02: hey, I don't appreciate the AI bot talking to me
• 36:35
  SPEAKER_01: about my financials.
• 36:36
  SPEAKER_01: Get it out.
• 36:36
  SPEAKER_01: Yeah.
• 36:37
  SPEAKER_02: Right.
• 36:38
  SPEAKER_02: But our teachers,
• 36:40
  SPEAKER_02: ourselves,
• 36:40
  SPEAKER_02: we are this is going to be something that infiltrates our lives and we can ignore it
• 36:51
  SPEAKER_02: or we can engage in ways that help us have more control over not only product,
• 36:56
  SPEAKER_02: but also data, also personal information.
• 36:59
  SPEAKER_02: Like there's all that stuff, too, that we haven't talked about, but it's out there.
• 37:03
  SPEAKER_01: It is out there,
• 37:04
  SPEAKER_01: but I think that our responsibility,
• 37:07
  SPEAKER_01: especially as teacher educators,
• 37:09
  SPEAKER_01: is to make sure that our teachers have enough knowledge, have enough capacity.
• 37:13
  SPEAKER_01: So this generation of kids have an understanding of AI when they go out there.
• 37:19
  SPEAKER_01: They know how to use it and when not to use it and think through all
• 37:23
  SPEAKER_01: of these problems because they're growing into this world.
• 37:28
  SPEAKER_01: It's our responsibility to prepare them
• 37:30
  SPEAKER_02: for it.
• 37:31
  SPEAKER_02: Yeah, no, totally.
• 37:32
  SPEAKER_02: It's just like anything else.
• 37:33
  SPEAKER_02: There has to be scaffolding.
• 37:34
  SPEAKER_02: There has to be mentorship.
• 37:35
  SPEAKER_02: And so in my own teaching, I've been trying to be really
• 37:42
  SPEAKER_02: Explicit and direct about what that looks like for me.
• 37:49
  SPEAKER_02: And those best practices are constantly shifting as new models come out.
• 37:53
  SPEAKER_02: That's the other kind of exhausting thing about AIs.
• 37:56
  SPEAKER_02: It's like, oh, you changed again and now you no longer work.
• 38:00
  SPEAKER_02: Why?
• 38:00
  SPEAKER_02: Why?
• 38:01
  SPEAKER_02: So there is that part of it.
• 38:03
  SPEAKER_02: But it's also a really interesting...
• 38:12
  SPEAKER_02: I don't
• 38:19
  SPEAKER_02: think there'll ever be another moment where the stakes are this high
• 38:23
  SPEAKER_02: and potentially impactful.
• 38:27
  SPEAKER_02: It feels like we're on a precipice and at some point we're either going to drop or fly.
• 38:34
  SPEAKER_02: I'm hoping for the flying.
• 38:36
  SPEAKER_01: I'm hoping for the flying.
• 38:37
  SPEAKER_01: Look to the magic carpet.
• 38:39
  SPEAKER_01: Back on the magic carpet.
• 38:40
  SPEAKER_01: Yes.
• 38:40
  SPEAKER_01: All right.
• 38:41
  SPEAKER_01: So thank you very much, Nick.
• 38:43
  SPEAKER_01: I am so happy you remembered to say yes to this interview.
• 38:47
  SPEAKER_02: I mean, look,
• 38:48
  SPEAKER_02: it's amazing what that first cup of coffee on a Monday morning will do to you
• 38:51
  SPEAKER_02: because you'll be like, oh, I never responded to that.
• 38:54
  SPEAKER_02: Yes.
• 38:55
  SPEAKER_01: And so you did respond and you came and that's fantastic.
• 38:58
  SPEAKER_01: Yeah.
• 38:59
  SPEAKER_00: Thank you so much.
• 39:00
  SPEAKER_01: And we'll see you next time on Azadeh and Guy.
• 39:03
  SPEAKER_00: On AI.
• 39:04
  SPEAKER_01: Oh, but that time you did it individually.
• 39:07
  SPEAKER_01: See, in the intro,