AI and Cognitive Offloading with Leon Furze

On the podcast today, I talk with author and consultant Leon Furze about the ideas found in some of his recent writing on AI and learning. We talk about resistance as a framework for combatting cognitive offloading, the mental models that educators and students have of working with AI (and how they can lead us astray), and classroom activities that can help students build the kinds of expertise they need to use AI effectively.

Episode Resources

Leon Furze’s website

Leon Furze on LinkedIn

“Resistance as a Framework for Combating Cognitive Offload,” Leon Furze, March 22, 2026

“IYKYK: How Do We Know What AI Can Really Do?”, Leon Furze, March 18, 2026

“Resistance Training Toolkit: Expertise,” Leon Furze, April 22, 2026

“On the Sensibility of Cognitive Outsourcing,” Derek’s take on that MIT study, July 7, 2025

Lodge, J. M., and Loble, L. (2026). Artificial intelligence, cognitive offloading, and implications for education. University of Technology Sydney

Corbin, T., Tai, J., and Flenady, G. (2025). “Understanding the place and value of GenAI feedback: A recognition-based framework.” Assessment & Evaluation in Higher Education 50, 718-731

“It’s time to move past the notion of AI as an answer machine,” Derek’s elaboration on Leon’s comment about mental models of AI, May 15, 2026

Transcript

Leon Furze (00:05):
So I think we have to put in conscious moments of resistance and for our students. I think we have to create those moments of resistance for them and facilitate that because I don't think they'll necessarily do it themselves.

Derek Bruff (00:20):
Welcome to Intentional Teaching, a podcast aimed at educators to help them develop foundational teaching skills and explore new ideas in teaching. I'm your host, Derek Bruff. I hope this podcast helps you be more intentional in how you teach and in how you develop as a teacher over time. You may be familiar with author and consultant Leon Furze from his series of articles on the teaching of AI ethics or perhaps from the AI Assessment Scale, a popular tool for helping students understand what uses of AI are permissible on a given assignment, which Leon helped develop. I'm very excited to have Leon on the podcast today, though we didn't talk about his AI ethics articles or the AI assessment scale. We could have, but Leon is such a prolific thinker and writer that we focused on a few of his recent blog posts and the ideas they contain.

(01:12):
It was his March 2026 post titled Resistance as a Framework for Combating Cognitive Offload that caught my eye. In it, he uses the idea of resistance and strength training to help us think about the ways we can help our students resist the problematic cognitive offloading that AI offers to learners. I invited Leon on the show to talk about his resistance framework and we had a wide ranging conversation about equipping students to learn well in this age of AI. I had multiple light bulb moments during the conversation and I hope you do too.

(01:50):
Welcome to Intentional Teaching, Leon. I'm very glad to have you on the show and to I think have a really engaging conversation about AI and learning. I'm excited to have you here. Thanks for being here.

Leon Furze (02:02):
Yeah, thank you very much for inviting me.

Derek Bruff (02:05):
And I really want to make a joke about talking to tomorrow because you are many time zones away from me right now, but I'll avoid that and start with my usual opening question, which is Leon, can you tell us about a time when you realized you wanted to be an educator?

Leon Furze (02:22):
Oh yeah, wow. We'll go back through the archives. So I was a maths science kid in school, but my teachers didn't agree with me. And so I ended up at university studying English literature. And after three, four years studying English and American literatures as a dual honors course in the UK, I sort of fell out of the other side of that and thought, "God, what do I do with an English and American literature degree?" And a week after the applications had closed, I applied for a postgraduate certificate in education, which is a UK teaching qualification and then just sort of rolled through the first few months of that before I discovered that I actually quite liked teaching and then spent 15 years in the classroom.

Derek Bruff (03:14):
Nice. Nice. And what did you teach in the classroom?

Leon Furze (03:19):
For the most part, secondary English and English literature, typically with senior students, but I worked in a small regional Australian school for most of my career, so I've taught history, geography, religion, STEM, media, drama, pretty much everything except for maths because again, the math teacher, even in adulthood, still didn't think that I was qualified to do the maths.

Derek Bruff (03:47):
Okay. All right. But small school, you teach what needs to be taught. That's great. Okay.

Leon Furze (03:52):
Exactly.

Derek Bruff (03:54):
Now, however, you are doing a lot of work I think at multiple levels of education helping people grapple with AI and what it means for teaching and learning. I'm going to start with this question because this is a little unfair. I'm going to give you a post from Blue Sky without any context, but it does seem typical of a kind of post I see a lot and it read, "Study finds that students using ChatGPT learned and remembered less than students who did not use it. " And then they link to a study. And so given how often we just read a headline and move on with our lives, when you hear a headline like that, or what advice would you give to educators who hear a headline like that and are concerned about AI's impact on student learning?

Leon Furze (04:46):
Read the paper.

Derek Bruff (04:48):
Yes.

Leon Furze (04:49):
Please, please read the paper. Even that headline, I'd be 99% sure it was the MIT paper.

Derek Bruff (04:59):
It was not, but-

Leon Furze (05:00):
Oh, really? Well, the 1% that fooled me. So many of those studies, as we know, this version of this technology has been around since November 2022 in the public eye. Any studies published on it have gone through peer review, hopefully. And that takes maybe six months plus sometimes. If you're lucky, it takes six months. By the time the papers are published, the models that they're talking about are out of date, but I don't think it's possible just yet to make any big claims in one direction or the other about the long-term cognitive effects of using AI. There's so much supposition and assumption there.

(05:44):
I joked about the MIT paper before. Really, I think that paper, which was really incredibly famous and big splashy headlines, what it essentially said was if a learner uses ChatGPT to generate a response and then submits it without reading it, they don't remember what the output was, which frankly is common sense, not necessarily a spectacular research finding in the long-term effects of AI on cognition. The paper was fine. It was great. And the authors actually came out afterwards and said, "Look, there's a lot of people misinterpreting this paper and reading too much into it. " And that's true of the other side as well. We see a lot of papers that are promoting the potential benefits. One of them in the last week or so has been retracted from nature, really, really again, very popular paper that was very well received by the positive AI hypesters.

(06:43):
So it swings both ways.

Derek Bruff (06:45):
Yeah. I saw that retraction too. And that was one of my concerns with the presentation of that study that I mentioned on Blue Sky is the study finds that students using ChatGPT learned and remembered less than students who did not use it. Which students? How did they use it? Were they prepared to use it? Were they directed to use it well? I mean, your initial response, read the study is super important because I find that the study design says a lot about how we infer or what we take away from the results.

Leon Furze (07:19):
And do read it. Don't just use ChatGPT to turn it into an infographic and then post it on LinkedIn.

Derek Bruff (07:26):
Right. That is good advice. Speaking of posting on LinkedIn, I think that was where I saw your recent blog posts. How's that for a segue? You wrote a blog post called Resistance as a Framework for Combating Cognitive Offload. And that's kind of the heart of what I want to talk with you about today, because I found it really interesting the framework that you started to scope out there. And as I read it, you're not arguing that we don't use AI, but we use AI thoughtfully in ways that are going to support student learning. And so you use this metaphor of resistance in this context. So what do you mean by resistance in this context? How does that metaphor work for you?

Leon Furze (08:09):
Yeah. So I mean, that initially came from two directions. One was I was reading lots of research around AI and education and trying to just sort of look at the major themes that were coming out, which were the five areas that I ended up with in the framework. And then in trying to think of what kind of tied all of those things together, I've done some writing in the past about friction, but I thought friction's not really the right angle for these areas. So I had a bit more of a thing. And the resistance idea really came from the fact that I'm reaching an age now where when I get out of bed in the morning, it hurts me and I've got three kids and I pick them up and swing them around and inevitably injure my back. And so I go to the gym reasonably often and I do strength training or resistance training.

(09:03):
And the reason I do that is not particularly because I enjoy it, it's because I have to. And if I didn't do it, my body would fall to bits. And I think it's the same with this technology. It's so seductive, it's so compelling and easy to use. It's increasingly ubiquitous. They really want us to use it and to sort of justify its existence. It is going to be increasingly difficult to avoid it. So I think we have to put in conscious moments of resistance and for our students and of course I mostly deal with younger students in K to 12. I think we have to create those moments of resistance for them and facilitate that because I don't think they'll necessarily do it themselves and adults, not just kids.

Derek Bruff (09:52):
Yeah. I've been working this year with a number of faculty who are experimenting with pedagogical uses of custom AI chatbots. And so part of that involves creating system messages for these bots to direct them to interact in certain ways. Maybe you're tutoring, maybe you're coaching, you're simulating a patient or something, right? And we're often working against the design of the AI platforms that want to be fluid and easy and overhelpful and we're like, "That's not how learning works. We need them to do other things than that. " So I've been using a gym metaphor myself lately and I think it's compatible, but I wanted to run it by you. So I've heard folks say now many times that using AI can be like taking a forklift to the gym. It will move the weights around, but that's not the point. And I've been arguing that that is true, right?

(10:56):
It can be a forklift in the gym and that's an unhelpful use of AI for learning purposes, but it can also be like a treadmill. When it's cold and rainy outside and I don't want to run in my neighborhood, I can go to the gym and I can run on a treadmill. It's creating an opportunity for resistance in a sense.

(11:15):
Does that work in your metaphor? I think it does.

Leon Furze (11:17):
Yeah, I think so. I think one of the pieces of research that I referenced in that original article was a paper from University of Technology Sydney, Jason Lodge and Leslie Loble produced a big report on cognitive offload and sort of the lay of the land really with the emerging research. And they talk in that paper about the potential for beneficial cognitive offload because sometimes we don't want to do certain things and it does free up room for bigger and brighter things. I think about my own use of AI is obviously quite broad and diverse, but probably my number one use of the technology is recording articles verbally and then using whisper transcription models to transcribe them, but then also running them through something like Claude to do the, put in the subheadings and paragraph them and pop them into nice tidy HTML that I can just copy paste into my blog.

(12:26):
Because I could sit there and write directly into the blog, but I find my attention wanders if I've got multiple tabs open and things I can concentrate more on writing an article if I'm out on the farm where I live, walking around kind of verbally speaking through it. And that for me frees up a lot of the tedium. I also have an AI built into my accounting platform that does my bank reconciliations, which is great.

(12:55):
So there are opportunities for beneficial offload, I think. I don't need to be spending three hours a quarter doing my bank reconciliations anymore.

Derek Bruff (13:02):
Right, right, right, right. Yeah. Yeah. And I think that's an example, like your approach your writing sounds like an example of my treadmill metaphor, right? It's using a little bit of technology to help you engage in the work in a meaningful way to you that might be hard to do otherwise, right, but it's not doing the writing for you. It's your thoughts, it's your ideas.

Leon Furze (13:25):
Yes

Derek Bruff (13:26):
You mentioned in the blog post these kind of five pieces of your framework. Can you kind of walk us through those five pieces and what they mean?

Leon Furze (13:35):
Yeah, sure. I mean, like anything, you can identify as many themes in research as you want really. You got to narrow them down at some point. I was joking to my wife yesterday that I think half of my time now is spent just my attention wanders for a bit and I kind of get distracted with what I've been writing and then I focus on something new and start inventing frameworks. And I think inventing frameworks is something that researchers and educators really enjoy doing just as a hobby. So the five areas that really stood out to me when I was reading around were the evaluative judgment or evaluation, which is something that a number of Australian academics have written a lot about.

(14:25):
That's really the capacity to evaluate the quality of work including your own and now the work of artificial intelligence, fact checking, hallucinations, all of that kind of work. Feedback was another area and for me, that's more about students learning how to seek feedback, how to use AI well to get feedback for their own sake, much less about educators using AI to provide feedback because the research is suggesting that students really, really don't like that and they don't respond well. The importance of expertise, this is something I've been writing about for a long time now, but I'm really of the opinion that AI works best in the hands of someone with existing disciplinary or domain expertise, which kind of makes it paradoxical to learning almost because by definition you don't know what you don't know.

Derek Bruff (15:23):
I have a slide in most of my presentations that says expert AI use requires expertise.

Leon Furze (15:29):
Yeah, absolutely. And this is where the areas start to overlap. You need to be able to evaluate the quality of AI and in order to do that, you need some expertise in what you're doing. And the last two areas are metacognition, which is a bit of a buzzword in education, just like cognition in general, it's almost its whole separate little field of education studies. For me, metacognition is less about thinking about thinking or learning about learning and more about understanding the feeling of learning, if that makes sense. So what does it feel like to grapple with a difficult concept and then push through and come out the other side? Because I think AI flattens a lot of that experience and I think one thing that's going to be increasingly difficult to convince learners is that that is a good thing because AI is so fluent and so confident sounding, so frictionless I think it's going to be really difficult to convince people that struggling through hard thinking is a good idea.

Derek Bruff (16:43):
It was hard to convince students of that before we had AI often.

Leon Furze (16:46):
Oh yeah, absolutely. Let's not be naive as well. I've been around the block for a few years now myself and I've taught kids and adults. I've taught postgraduate teaching qualifications and other aspects of higher education as well and nobody really wants to learn when they're feeling tired or we're heading into winter here in Australia and the days are getting shorter. There's other things that we'd rather be doing often than struggling with dense concepts. And I think AI again is very seductive there. So teaching students to kind of just sit with those feelings for their own sake is I think really important.

Derek Bruff (17:32):
Yeah. I think I've shared this story on the podcast before, but I had a transformative moment my junior year of college. I was doing a study abroad program for math majors. I was in Budapest, Hungary. I was there with about 45 American math majors taking a whole bunch of upper level math courses all at once. Super geeky, but it was hard. It was challenging. And I remember there was this one course called Conjecture and Proof and they would give us these math problems that we had to work through and I was pretty good at math. I was good at maths. Everyone told me I was good at math. And so I get there and I'm like, it's never taken me more than like half an hour to work through a homework problem in the past. And these problems take four hours, six hours, eight hours, right?

(18:20):
Sometimes there's three of us working it together and like even putting our heads together, it takes hours to kind of get the insight that helps us move forward. I never had that experience before, but having gone through that and realized that yeah, sometimes it takes six hours to solve a math problem, knowing you've done that once helps you approach other hard problems very differently. And it is a feeling, like you say, it's less about knowing that it really is more about having that confidence to go into something hard like that and to persist.

Leon Furze (18:51):
Yeah. And I mean, that story has just triggered a memory in me, which is kind of illustrative of all of the stuff that AI can't do. The fact that you've just shared a personal anecdote and then it's conjured up one of the memories from the depths of my brain as well. When I was doing my own postgraduate teaching course, we went on a field trip to Florence, Italy and we were English media drama teacher graduates, teacher students. And for one of the days we had to follow a small theater troupe around the city of Florence as they very enthusiastically acted out the Medici story and we had to get involved. We had to join in with some of the improv and incredibly daunting for a group of young, inexperienced, pre-qualified drama teachers to stand in the streets of Florence and

(19:55):
Act with a theater troop in the middle of Italy. But that experience absolutely 100% made me a better and more confident drama teacher and that I think as you've just said with your maths experience, just having to go through that process of sort of baptism of fire to improve your skills to kind of hone and sharpen your knowledge, that's not going to come from interacting with a chatbot. And there are lots of ways to gain knowledge from using AI. I'm not on the side of the fence that says that they can't be used in education at all, but we have to look for the moments where the more human, relational, like your experience and mine there, where we can bring those to the fore.

Derek Bruff (20:46):
Yeah. I think you had one more part of your framework.

Leon Furze (20:50):
Yeah. So the fifth aspects sort of necessarily last is this idea of sort of stretch to sit with that resistance metaphor just to push further now. And the analogy that I used on the blog post originally was you don't just walk into the gym and start lifting heavy things because you're going to injure yourself in much the same way you're not just going to walk into your most powerful AI platform and expect it to start learning for you. We do all of the other things first. We build evaluative judgment, learn how to seek and use good feedback. We apply our own understanding of learning and our metacognition. We develop expertise and then we apply the technology to move us into new uncharted territory, make weird connections, go down rabbit holes. And I think once you've got that basis, that foundation, it's definitely possible to use AI in that way.

(21:48):
I certainly use AI in that way in things like coding and web design and areas where I'm right at the edge of my comfort zone.

Derek Bruff (21:57):
Yeah. So let me ask a couple of kind of things that I think fit in the framework, because I really love this approach of trying to say, okay, we don't want students to just rob themselves of learning opportunities, but more specifically what does that mean? What are the components of that? And I think you've done a great job of enumerating and identifying these ... The part that you have to have the ability to evaluate AI output. I hear that a lot, right? But the notion that learning is hard and it's going to take some struggle and I need to know what that feels like so that I know when I'm not doing it, right? I don't hear that articulated in quite that way very often. There's a couple of other elements. So as I think about my own AI use, I feel like maybe I'm off here, but I feel like I'm getting better at knowing for which tasks AI will be helpful to me and my process and for which tasks it will not.

(22:57):
And that's a little bit about the AI and a little bit about the task and a little bit about me and how I want to work, right? And so where does that fit in your framework? It doesn't sound like feedback, but it has a kind of similar vibe of like when to go to AI and when not to go to AI.

Leon Furze (23:14):
Yeah. So I mean, in both the feedback area and the evaluation area, I've seen both research around this and then anecdotally and in my own experience in using the technology, as you say, developing that awareness of when to go AI and when to go human essentially. And with students, I'll look at feedback as the example here. Students are telling us that they want human feedback for some things and that they do not like receiving AI feedback from human teachers. They recoil if they even get the slightest hint that a teacher has used AI to give them feedback, they revolt. And on the almost sort of contradiction is they are using AI on mass to give themselves feedback.

Derek Bruff (24:06):
Right, right. Okay.

Leon Furze (24:08):
Yeah.

(24:09):
And then there's research coming through that says even if the students, if they don't know whether a given piece of feedback is AI generated or human, and then you tell them that it's AI generated, even if it's written by a human, they don't care that the knowledge or the supposition that it's AI generated is enough to totally devalue the feedback. So for me, there's something there and it's not the quality of the feedback that matters because AI feedback can be quite high in quality. It's the quality of the relationship. So one paper I'm thinking of, Tom Corbin, Joanna Tai, maybe another author escapes my mind at the moment writing a piece last year, I think about different types of feedback and they talk about the importance of recognition and the sort of the mutual recognition and vulnerability, the shared vulnerability between an educator and a student.

(25:09):
The student reaches out for feedback, which puts them in a position of vulnerability. The teacher provides feedback and kind of says, look, here, I'm the expert, trust me. And that kind of mutual recognition is something that AI can't do. They use the term extra recognitive, which I've told Tom I don't like because I can't say it. So I've asked him to change the name. So if he ever hears this podcast, then Tom, please do change that in future posts. But that extra recognitive feedback that AI can give can be just as good quality in terms of identifying strengths and weaknesses and all of that stuff, but it doesn't have that reciprocal human relationship component. So like you said, you over time can identify what's a good way to use AI and what isn't. And I think more broadly in education, we need to come to those conclusions as well because we see students and educators using AI in ways which are probably not great.

Derek Bruff (26:15):
Yeah. And I think a lot about how some of that, it could vary from student to student. For this student, this use of AI is going to slow them down, but another student that may actually be a helpful thing because I think sometimes as educators, we want to have a kind of single solution for everything, but our students are very different and that relationship piece is key. It's almost what you said, the students will easily go to AI themselves to get feedback on their work, but if they get essentially the same feedback from an AI in the place of their instructor feedback, they recoil and it feels almost like it's analogous to the uncanny valley, right? That's supposed to be a human, but it's not obviously not a human and I'm freaked out by that, right? There's

Leon Furze (27:06):
An aspect there as well that speaks to the value of the relationship between the teacher and the student because the student doesn't want the educator to use AI to give them feedback. And then on the educator side, I've seen this cause a bit of tension because the reasons that teachers are using AI to give feedback are not because they don't want to engage with the students or form relationships, it's because of their workload. So we have students using AI to write work and then the teachers feeling totally flat from being overworked, marking all of this AI generated stuff. So then they start using AI to give feedback and the students resent them for it and then the teachers resent the students for resenting them and that whole cycle kind of goes around. I was working face to face at a school in Melbourne on Monday this week and this came up with the heads of faculty.

(28:02):
We said, "How do we deal with this? Because we want to use AI in ways which reduce our workload, but we don't want to do it in a way which threatens the relationship with students." And I said, "Look, I mean, I would just cut out the middleman. Instead of having teachers using AI to give students feedback, teach the students how to seek feedback from AI better." Students just throw their work into AI and say, "Is this good?"

Derek Bruff (28:34):
Right?

Leon Furze (28:34):
Then they get a very sycophantic, enthusiastic response from something like ChatGPT, which tells them they're a genius and there are ways to use the technology better and also places where the technology is best avoided. So teach them those behaviors and aptitudes, cut out the middleman and then anytime that you save, use it in just having conversations face to face with students.

Derek Bruff (28:59):
So feedback is not just when to go to AI and when to go to a human, but how to seek feedback from different sources and when to seek it.

Leon Furze (29:08):
We have methods of feedback. We have methods of formative assessment, programmatic assessments, all of these things which we know are effective, but they don't scale very well and they require a lot of time investment and they're much more relational and that's really, if we're being honest, that's the problem. It's not the AI that's the problem. It's speed, efficiency, scale and cost.

Derek Bruff (29:37):
Yeah. Well, I want to kind of move into some practical strategies for teachers and maybe I'll do this by saying in this framework there are these areas where we want to build resistance for students. If we're going to teach resistance to students, what do we have maybe to unteach them and In terms of their existing uses of AI.

Leon Furze (30:04):
Yeah. Students and educators and everyone. I've got another series of posts on my blog at the moment, which is broadly called If You Know IYKYK. Based on this idea that if you know what AI can do, it's quite easy to use it in ways which are fairly effective. But if you don't know what it can do, the technology doesn't help at all. And one thing that I would love people to unlearn is a design fault I would say of OpenAI or a conscious design decision maybe, not a fault. When they released ChatGPT, the first ever very big publicly available commercial chatbot style, large language model, they made it look like Google. And that I feel was a very deliberate move to make it familiar seeming and comfortable and easy to use. But essentially you go to chatgpt.com, it looks like Google, it feels like Google.

(31:07):
It even to an extent, especially now that Google's just chock full of AI anyway, it feels like Google. And it is the worst way to use the technology, particularly from the early days before these things had reliable internet connections and the ability to search.

Derek Bruff (31:24):
In 2022, that was a terrible choice to think

Leon Furze (31:26):
Of. Oh yeah, absolutely. And so what I've written about in that series of articles is the first experience of AI for most people seems to set the expectation going forward. It sets their ceiling on what they see AI as the kind of the mental model around it. So if your first experience of ChatGPT was the thing that kind of looks like Google, so I'll use it to do a search, the mental model becomes AI is a search engine. If your first experience is Microsoft Copilot and using it to summarize an email from the boss, it becomes the thing that summarizes text and a lot of people's mental models, students and other are stuck on that. And I think the only way to deal with that is to ignore the way that the technology companies are actually presenting their technology and just kind of start experimenting and doing weird stuff with AI.

(32:28):
The example I've used recently, I'm working with a group of curriculum leaders at the moment is thinking that artificial intelligence, large language models are really, really proficient in writing code and therefore they can do a pretty good job of simulating anything that a computer can do because they can write and execute code. Even ChatGPT on the web can write code and run code and access tools and use stuff. So what does it look like if instead of thinking of AI as a search engine, we think of it like a sort of Swiss Army knife app for working with a computer. And then people start doing really weird stuff. Like one of the educators had mentioned they downloaded a whole bunch of PDFs and things with dates and stuff on and had ChatGPT converted into an iCal file. And that's such a kind of a niche specific use, which has nothing to do with how most people use AI.

(33:25):
So my advice for students would be get a little bit weird, stop thinking of ChatGPT as a search engine because it's a horrible search engine and start thinking more of it as just a kind of multipurpose computer application that speaks the language of computer.

Derek Bruff (33:43):
Yeah, because what you just described or like your own use of using AI to take your kind of raw blog post and format it with headers and such.

(33:58):
Just yesterday I was taking someone's syllabus, which was a well formatted Word document and we wanted to put it up on our website as an HTML document so it would be more accessible to screen readers and such. I did all that by hand. I don't know why I did that now. I shouldn't have, right? But it is. It's your first exposure. And I've used AI tools for that kind of thing in the past, but it's weird to have a Swiss Army knife that does such strange things. It's hard to get in the habit of realizing, oh, here's something that could do. I mean, I had an actual Swiss Army knife. I still have it somewhere in my house and it has tools I've never used and that I don't know what they do, right? But I know what most of it does. Yeah, that's a good metaphor.

Leon Furze (34:48):
Reaching beyond those initial mental models is really hard because this is in the blog post, but in user experience terms, there's this problem called the discoverability problem, which means like how easy or how difficult is it to discover the latent features of a piece of software? So you look at Microsoft Word and you can discover most of the features just by clicking through the ribbon. Everything's laid out. There's a few tabs. If you want to add comments, you can find that if you search around enough AI doesn't have any feature toolkit. And the companies have tried, I think, to do some of this. If you look now at ChatGPT or Google Gemini, it's got the dropdown box, you can turn on research mode, you can click a button to generate an image, but those are very narrow feature sets and the abilities or the capacity of AI is so broad that it would be impossible for it to have a toolbar like a traditional app.

(35:45):
You've got to kind of feel your way through and create your own mental toolbar.

Derek Bruff (35:49):
Oh yeah. That is a high load for user interface.

Leon Furze (35:55):
Oh yeah. Terrible software.

Derek Bruff (35:58):
Right, right. And again, they're going for the simple, the easy, the Google, the clean.

Leon Furze (36:05):
The Silicon Valley boring.

Derek Bruff (36:08):
Right, right, right, right, right. I do want to ask about some classroom practices because you've got another blog series you're calling resistance training toolkit. And so these are activity structures that are kind of targeted at some of these areas in the framework. And so can you share a couple of examples of what those classroom activities might be that would help students develop this kind of resistance?

Leon Furze (36:28):
Yeah. I mean, so these for me, they come from the world of things like Ron Ritchart's Making Thinking Visible, the Project Zero, Harvard University, Visible Thinking Routines, those kinds of toolkits, the Black and Williams formative assessment book, they have lots and lots of examples in them of really short classroom activities. So I wanted to expand that resistance idea into that territory. And what I've started to do is build out a few examples section by section. So I started with expertise and had to think about like, okay, so what's a good activity to use in a classroom either before, during, or after introducing AI and then also link it to that emerging research around the technology. So there's a paper that suggests that if you interact with AI as if it's a novice learner and then you teach it what you know, almost like a teach the teacher kind of approach, then that can be useful for identifying gaps in your own knowledge.

(37:36):
So one of the activities is essentially an adaption of teach the teacher and the student interacts with the AI kind of info dumps on them everything they know about a topic asks questions and then goes backwards and forwards and iterates for you that a few times. Another one is a close reading activity essentially. So what can a student do without AI before they start using AI? We do some close reading, we do some annotation, we take things really slowly and then when we start to introduce the technology, we start looking for opportunities to add complexity or layers to the things that the student has already noticed. So we don't just default to using the AI in the first instance we actually showcase or develop or explore our own expertise first.

Derek Bruff (38:33):
Yeah. So both of those are designed to help foster that expertise in the discipline or the domain in part by, I think what you were saying earlier is we're having students use the chatbot not in their normal mental model, but in other modes, right? One thing a chatbot can do is it can pretend to be other people of particular types, right? Not always well, right? But it can pretend to be someone who doesn't know about cell biology and then you have a chance to then teach it what you know about that. I had on my podcast a while back, Heidi Nobles, who is a English teacher at the University of Virginia and she has a lot of experience as a book editor. And so she had all of these uses of AI that were kind of fundamental model of a book editor. What does a book editor do?

(39:20):
What kind of feedback do you get structurally, conceptually, strategically? And again, that's not the kind of go- to model, but it's one that points towards the students developing expertise perhaps if they are adding that complexity and nuance in that way. Well, I hope you continue writing that blog series or maybe put all of that in a book.

Leon Furze (39:42):
My intention is to ... I purchase domains almost kind of compulsively. So I've got my Leonfurst.com. I've got practicalaistrategies.com. I've got teachingaiethics.com. I think somewhere I've probably already purchased a domain that would suit this series. My intention is normally to publish this kind of stuff, open access and just throw it out into the wild, which is I think that's the teacher in me more than the business person in me. Produce all of this kind of stuff, put it out into the wild, get a bit of feedback and see what people think about it. So yeah, I'm definitely working on all of those in the background.

Derek Bruff (40:21):
We are almost out of time. So I will ask what's next for you. Do you have any new projects coming that you're excited about?

Leon Furze (40:28):
Always. So my PhD is officially finished.

Derek Bruff (40:33):
Congratulations.

Leon Furze (40:34):
Confirmation of that. And I'll walk across the stage in June for The Silly Hat. I'm going to continue with the writing. I've got a book sitting with the editor at the moment, the imaginatively titled Practical AI Strategies two. And I'm just working with lots of really interesting people. Recently, I've had conversations with ACARA, which is the Australian Curriculum Assessment Reporting Authority that does the professional standards for teachers over here. So there's really high level conversations going on in Australia, but I'm also running lots of individual school, universities, one-on-one with teachers, that group of curriculum leaders that I mentioned before. So I find that really enjoyable. I've always enjoyed teaching teachers for over a decade now running professional learning. So I'm going to continue writing and continue teaching.

Derek Bruff (41:28):
That's great. That's great. Well, all good work and we thank you for doing it and thanks for coming on the podcast today. This has been a really fun conversation. Yeah.

Leon Furze (41:38):
Great.

Derek Bruff (41:38):
Yeah. Thanks so much.

Leon Furze (41:39):
Thank you very much, Derek.

Derek Bruff (41:42):
That was author and consultant, Leon Furze. I keep thinking about some of the ideas he shared in our conversation. For years in my presentations, I've said that AI tools like ChatGPT are more like wordsmiths than Oracles, which is a great metaphor, I guess. But Leon helped me see that I was actually trying to redefine the mental models we have of working with generative AI. Our mental models of AI don't just inform our understanding of AI. They shape our use of AI and there are a lot of ways to use AI poorly. I think we need to move away from the mental model of AI as Oracle, as Google search, as answer machine, and we need to help our students do the same. Leon mentioned a recent lit review by Jason Lodge and Leslie Loble and that review points to studies showing that A, default uses of AI by students are largely unhelpful to their learning, but also that B, when students are provided better guidance on using AI, then their use of AI can actually improve their learning.

(42:46):
I think changing student mental models and in some cases, educator mental models of AI is key here and I wrote a whole blog post about this idea, a link to which you can find in the show notes. Also in the show notes are links to the blog posts by Leon Furze that we referenced in our conversation today, as well as links to some of the studies and resources that Leon mentioned. If you like a podcast episode that gives you lots to read after you listen, this is the episode for you. And if you liked what you heard from Leon here today, please do visit his website for lots of ways to keep up with his work. He's got newsletters and podcasts and LinkedIn connections to make. He really is a prolific thinker and writer about AI and learning.

(43:30):
Intentional teaching is sponsored by UPCEA, the Online and Professional Education Association. In the show notes, you'll find a link to the UPCEA website where you can find out about their research, networking opportunities, and professional development offerings. This episode of Intentional Teaching was produced and edited by me, Derek Bruff. See the show notes for links to my website, the Intentional Teaching Newsletter, and my Patreon where you can help support the show for just a few bucks a month. If you found this or any episode of Intentional Teaching Useful, would you consider sharing it with a colleague? That would mean a lot. As always, thanks for listening.