AI As a Finishing Agent

And he's worked in both public and private settings with students from middle school to graduate school. He is got a wealth of knowledge about these various institutional spaces and student social, emotional, and academic development across this age range, as it relates to developing responsive AI systems. So definitely a very interesting conversation. Nick, welcome to a Vision for Learning. So great to have you here.

we've been talking a lot about the Apple Vision Pro because that's what has been released recently, but I'm excited today to talk all about AI and how it works. And I look forward to you, hearing this conversation, and we'll get to that conversation with Nick in just a moment.

Nick, I think a good place to start out is talking about the idea of plagiarism. it's still something that people are worried about and are still feeling stress about and feeling like it's, that's the thing we need to worry about with ai and I think it is really down at the bottom of the barrel of things that we. Actually need to worry about it. To me right now, plagiarism is no different than plagiarism has always been, except that it's a lot easier and that's the only difference.

kids are still gonna be motivated to do it. Kids are still gonna want to do it, to get out of hard work. that's all the issue. Where are you coming down on the idea of AI and plagiarism right now?

In particular was impressed by a post by Alberto Romero, late in the summer of 20, when he was talking about how these detectors introduced sort of a surveillance culture into our classrooms. And that's where, that was the tipping point for me. when I decided that I had to really change the way that I was teaching, I had to start to think about, If I was gonna continue to lean into, into long form essays as a English teacher, then I had to, start to structure them differently.

I had to collect them more incrementally. I had to open up, the essays to more AI in integrated processes. And make that part of the workflow, because, we have to adapt to the conditions on the ground, I think, as opposed to, just holding onto the mythical past. so yeah, I like the way that you have framed it as it's a leveling that this is, it's all plagiarism. we need to start with those conversations as.

statistical print printouts, which, if you go down that road, you'll very quickly, your students will have their own statistical printouts to show you that what they're doing isn't AI generated. So it's it's just you're, you go into that Cold war mentality and then inevitably nobody's learning. there's no trust. There's no enthusiasm in the classroom, and you get to the point where you're just wondering, what are we doing here? it's definitely not education. So that's where I'm at.

Nick. Tell me about, about your idea of how you're changing your teaching as you're focused on integrating AI more now. what does that look like? there, there's the idea that you don't, Have students take home and write essays solely at home, and that's, that would be like an obvious, easy first response. But what else does that look like where you're actually integrating AI work into your kids' work?

so we're entering into sort of an experimental space, I am hesitant to go all in the sense of having a fully integrated AI classroom. I know that some teachers are doing that and I'm glad that they are doing it because I'm learning from their experiences. But I'm running sort of two parallel tracks within my class where.

There are times where, we are still just working, in a very traditional way, just generating, generating ideas without, using ai, tools, x and E. Hello, for instance, just coming up with those ideas from scratch. I think, post pandemic kids really need that. because like throughout the pandemic, they had so many assistive tools that were bumping them in terms of here, do this next thing, do this next thing. particularly those two years when they were mostly on screens.

so we have to, I'm finding that we have to. Find ways within the curriculum to, to create those x ni, hello skills. at the same time the, yeah.

Because if something happens, I know you can just freeze my grades and wipe away everything, and we don't have to take tests this one year. And like we sent a very strong message to them that we can just cancel school at the drop of a hat. and a lot of people don't tear me say that, but that's the reality of what we did.

One, one method is to, ignore the, these AI tools, and just kind of, just lean into the hard work, the, the struggle, make everything as, as hard as possible. every, brainstorming has to be, just pulling up all of the ideas from yourself. And, our kids are really struggling with that kind of work. And so I'm trying to meet them, I'm trying to meet them, in, maybe 20% of the time with some AI responsive lessons. so to make it a little bit more concrete.

So we're working on an essay on, west Moore's. great. Great book, the other West Moore, and it's a, it's an, eighth or ninth grade classic. so a ya version of, the, a longer piece that he did. So he's now governor of Maryland and just a tremendous, story of, of never giving up and.

So we're going into kind of some thematic questions and like we've already done what I was talking about is like the one track where we've done some brainstorming X knee hello, with a different essay, but, so now, we've dipped our toe into AI a little bit and they have, access to some models and I've, I've. I laid the groundwork through parent permissions, to get them that access.

And and we've, I've given them some very, carefully structured prompts and, they've accessed some potential outlines and, and I've also had them use some brainstorming to create some thesis formula. so it's. The idea is to use AI more Socratically. and I got the idea from, con Conmigo, which is a, it's a commercial product out there that, con Conna Academy is, has put out and is rolling out in a number of schools.

And what's distinctive about their product is that it doesn't just like dump answers. like most AI models it, it will ask more questions. so I've created, prompts for my students that will. Not jump out, here's the thesis, but it says, here's like a formula that you have to then populate with your own ideas. so we're doing all of that, like on the front end of writing, and then we shift to a more of a concentrated human space where they're gonna be doing the writing and drafting.

And then we'll come back after we have the drafting and we'll get some commentary from ai. I have this, theory about AI that it's really only effective at the beginning and at the end of the writing phase. Like you, you have to, particularly at the end, you have to fill up the AI bucket, for it to start to do anything interesting. and I'm hoping, like I'm working with ninth graders now, so the AI literacy just beginning, but.

I'm hoping as I work with, students that are older, we can start to get a little bit more sophisticated in the sense of, I would love to show them, how you can enter in information at a very low level and get a certain kind of response, but if you really want AI to help you write something super specific, then it's a much more like. Involved process and what actually is revealed is like, there's a lot of specific kind of writing that takes place between you and the model.

And I'm thinking like down the line, like what is writing instruction gonna look like? Like a lot of people are fearful that, okay, AI is gonna out mode English, as a subject, or writing as a discipline. I, in my conversations with folks, I'm like, no, it's just writing's changing. it's, we're just gonna be learning how to write more functionally, and operationally.

And I think that this is a very interesting challenge because especially as you're talking about these kids who were in school during the pandemic and probably did not have a ton of writing opportunities and an opportunity to develop that writing, they don't have a voice. And so when they ask the AI to write it, then. It's a really big deal because it produces something better than what they could write. As far as they can tell.

When I ask the AI to write something, it is never as good as what I could write myself, at least in my opinion, right? and so I think, I wouldn't use that word, I wouldn't write it that way, et cetera. And so we need to change. Our perspective about how we are writing things and be aware that is changing.

And I think that piece is the part that's really hard for teachers to grasp and appreciate because I. There's this nostalgic way that we used to do it, and then there's this new way that's scary and we don't understand it, and we don't, nobody really understands how these, transformers work so that, so that it actually produces the text. Even the people who are creating the algorithms to make it work, they don't even understand exactly how it's working and that part. It's scary for everybody.

I totally get that. So I like what you're saying about how you're going through

that process. How, the AI is only helpful in the beginning and in the end. can you talk a little bit about your. Your writing and how much you use it, and if it is, something that is helping you or hindering you as you're writing your sub substack posts that are so good and detail oriented. what's your experience with your own personal writing as opposed to your teaching writing to your students?

and And that kind of would plant sort of a seed of doubt in me, about, ultimate purpose and about how it was gonna land with audience. And, and then I'm not saying that AI is ultimately doing all that finishing for me, but for some reason kind of me synergizing with ai. lifted a little bit of those that worry and those concerns. in to, with, I've done a PhD and I've grappled with my own writing demons now for I. 20 years.

So I mean like me com compared to how anxious of a writer I was in undergrad compared to how I am now. I've just gotten a little bit quicker each year and a little bit less critical. But now like with AI in tow, it's like something's just happened with me. I'm truly grateful for it, because now when I like sit down, it's just, I know that I just, I need to give, these AI systems like the sort of the logical connections, between my ideas, like those that really is what it needs to do.

A lot of its sort of polishing work. And so I just go in there and I'm like, I have these connections of the ideas. Like the ideas were always there for me. I'm just like slamming the ideas out, right? and I'm not worrying so much about the articulation. And then I get into a conversation with AI about the, like. What it looks like. And I let it do a first look through, and it will give me some, initial sort of, here's some things to consider, as you do some editing.

So I'll allow it to give more like a laundry list of things to, to do as opposed to actually going through and, Kind of going through and actually revising all of the text. And then, with those directives in hand, usually I'll get to a place where it's almost, polished. I don't know, I've just become much more, more efficient and freer as a writer. and I'm in with doing the substack now once a week. I'm just in, in like maximum.

performance zone, as a writer, where it's like I'm just feeding off the audience and I'm like nestled into like this, like amazing group of writers right now on Substack. like Alejandro and, Nat and Michael Rutenberg, like these, we, I mean we're a small e in terms of the AI community on there. we've, we're just chopping up ideas and doing guest posts and helping lift each other and, it's just, it's like social media before social media,

As you've been using AI for your own work, your own writing, how has that helped you relate to and s helped students see where the AI is beneficial or not for what they're doing?

We're in, in co-pilot mode, where I'm learning from their use as much as, they're learning from my experiments up on the board. so it, it's a much more, collaborative, a collaborative energy, collaborative atmosphere. and I think the. And this is something that I encourage, the teachers that I encounter is to have a, to have a culture of transparency around use. if you, as a teacher, if you're using it to generate any classroom materials.

it's, I think it's a good thing to be forthright with students that you have leaned into, AI to, to generate those materials. Um, you, we get into sort of a weird space in, in K through 12 where you have some teachers who are prohibiting, student use of AI and yet using it to make their tests You.

I. But every time I copied and pasted it, I just never felt like it was any good or that it was worthwhile. I'm in a doctoral program right now, and the temptation is certainly there, especially on certain assignments to just have it do it because. The assignment is basically pointless and like it doesn't really matter. And on those situations, I definitely want to use AI because when the assignment is pointless, then it's, it's really difficult to give it time and attention.

And so if the assignment, what I'm doing in those situations where there's an assignment is I'm trying my hardest to figure out what the real purpose of the assignment is. How I can make it worthwhile for me in a different way that matters. so that's. That's one thing. The other part of that is that when I do copy and paste it with one exception, which is I use an app called audio pen.

Audio pen.ai, and what this tool does is it will take you speaking, which is something that I'm very comfortable with doing right, and it will take your speaking and turn that. Rambling, incoherent thought into something that is clear. That app, because of how I've set it up, has done the best job of taking my rambling, incoherent thoughts and turning them into something that is publishable.

And I just need some help. I've had this problem historically throughout my writing career where it's just I hit sort of a block in my thinking, like how can I unwork this? And, when I was writing my dissertation, those sort of blocks would halt my writing process sometimes for. Five, six minutes, and now with sort of an AI tool, I can get a couple different options to, to play around with pretty quickly.

yeah, but mostly it's I like how, I mostly work with, chat GPT, version four, which has been acting really quite strange the past couple weeks. But, um.

And they know Japanese so they know what it is, but they didn't wanna translate it themselves. They put it into chat, GPT, and it translated into a story about somebody killing someone else instead of teaching you how to, how to fold this flower correctly. and what's so fascinating is that if this goes back to that whole literacy piece, that if you don't know what you're doing already. Then you could really be in a lot of trouble with it. and that's really what learning is all about, right?

That we wanna teach people to know how to do the things that we're doing. And in, in one of your subsets, I don't remember which one, you linked something from, AI for education, io that talked about, about one, there was a flow chart and one of the things said. If you're skipping the learning, you're doing it wrong, consider using AI in a different way and like that piece is just so important. We, we shouldn't be skipping the learning. any thoughts on that? On the skipping the learning part, I.

It can be, multimodal, and it's our job as educators and particularly, writing and structure instructors to invite students into a deep, critical engagement. ethical aesthetic, philosophical, Rhetorical, these sort of deep engagements with texts and whatever form they come. we, and you can stage that in a number of different ways depending upon where your students are developmentally. but if, I think there's gonna be a tipping point, as we think about implementations and integrations.

where these processes, will start to encroach on some of that critical thinking. and that's where we have to be, have that overseeing or that, and that, Also hold on to a vision, that I think are going to ultimately, like the only way that we're gonna be able to maintain it is by establishing some principles or, and some outcomes that we wanna see from these, syner synergistic kind of spaces. so like in my most recent post, I was starting to plan out some ai.

Learning outcomes, that, that were, that, I wanna continue to think about. and there are also, there are a lot of like isti, and there's quadruple a, which is a big a ai, promotion. Organization, but they have an educational wing, like they're a company or organizations that are building these like deep, literacy programs, that I think are gonna help us to structure our practices, and help teachers build up some sort of watch points or criteria to, to.

think about, is this particular implementation maintaining critical thinking? Are we giving away too much? it's gonna, I think a lot of it will just come down to sort of like, um, costs and and benefits for particular implementations. so when I was at a conference out. Philadelphia. A couple weeks ago, I did a training with teachers from across the Eastern seaboard where I walked them through a AI writing process and, we used AI to do brainstorming, and then initial drafting and then editing.

And I had them go through like a five scale, like a valuation of, How did AI do you know, and I, the major like criteria that we were working on, was like, knowledge and skills generation, like was there new knowledge being generated through the integration of ai? Were there new skills being generated? I think we need something like that. like some sort of like solid criteria that we can use to evaluate use. Because like right now it's. It's pretty ambiguous.

I don't know, like the, at the conference there was like, the value they were throwing out was like, human centered, right? is our integration? Is it helping us keep education human centered? And that's, that's like a good big idea. But it seems, there's also like other things underneath that, like what, what exactly does. so that's gonna be the work I think of the next couple months, maybe year, of figuring out those criteria, for evaluating implementations.

and I think like part of our vision for this, these training materials are just kind of. coaxing leaders just to sit down and use the products, as part of an acculturation process. what can these actually do? I think once you sit down, when you sit down immediately and you ask or right away and you see it, that it can create a sonnet, like that's, that sort of blows you away and you're like, oh, wow. the. there's no nothing that you know, that these machines can't do.

But, as you prompt away with it for, a half hour, then you start to see the limitations, the repetitions. You start to see the hollowness of the text. you start to ask it questions about the text, and three prompts in that, the level of understanding just. Maybe it gets a little bit, blurry. and I don't know, you, you start to understand the way it works and, once you get

However, I think real people are going to start writing that way 'cause they're gonna be mimicking what they see. Just as a little side note, not to get us too off course, but, so as we close up here, Nick, this was awesome. in the show notes, if it's okay with you, you put a. A great outline of your process of how you've gone through learning about this and a few different links to yours and other substack that were just really powerful.

I'm gonna leave those in the show notes so that people can check that out. We didn't have a chance to go into all of those things, but they, I thought they were really good and really interesting, and so I want people to be able to check those out. How can people connect with you and learn more about the stuff that you're doing? Nick?