Catching Up: Student Influencers, AI’s Mirror Problem & The Power of Early Work Experience

In this episode of Catching Up, Nate McClennen and Mason Pashia explore the ideas shaping the future of learning and work—from the link between school innovation and enrollment demand to the rise of student influencers as a new form of district storytelling. They also dig into AI’s growing role in education, including bias in feedback systems, the homogenizing effect of generative tools, and how AI may free teachers to spend more time on uniquely human work. The conversation also highlights the importance of agency, internships, and real-world learning, while ending with reflections on poetry, humility, and the power of noticing the world around us.

Outline

• (0:00) Intro & Episode Preview
• (10:14) Student Influencers
• (12:51) AI Bias & Homogenization
• (27:01) Deep Dive: Agency & AI’s Role in Education
• (50:17) What’s That Song?

Intro & Episode Preview

Mason Pashia: Today we’re talking about high schoolers becoming influencers for their school and district, which is kind of a fun new thing being tried by some districts. We’re further looking into the question of, is AI bringing us together or pulling us apart, both relationally and also just as, like, a species POV?

And then we’re going to talk a little bit about dams as a sign of humility.

Nate McClennen: Oh, can’t wait. Listeners, wait till the end for that one. It’s really good. Yeah, on my end, I’m really thinking about we know that the public sector is going to struggle with enrollment in at least a, a, not a large number of schools, and so how does that relate to the innovation programs and supply versus demand?

We’re going to do a really deep dive on agency. Everyone’s talking about agency right now. It seems like the important word that everyone’s putting into their mission statements and their portraits. We’re going to dive a little bit into internships and work and the power of those things on young people’s lives in terms of early or better chances of getting jobs. And then we’re going to end with a Mary Oliver poem that I wanted to share, and about a fox eating a rabbit. So stay tuned, everybody. It’s going to be a great show.

Mason Pashia: What was your first job, Nate?

Nate McClennen: My first job was picking dandelions for my father on the front lawn. He gave us a penny a dandelion.

Mason Pashia: Beautiful. Mine was trying to sell samples of cinnamon rolls in a grocery store.

Nate McClennen: Cinnamon rolls yourself? Was that yours, or they were others?

Mason Pashia: It was my grandma’s recipe. My aunts made them, and I ate so many. Like, truly so many. Yeah, it’s great. All right, this’ll be a fun one.

Nate McClennen: Yep. Excited.

Nate McClennen: Hey, both of us had some wedding experiences over the last week. So your wedding was Colorado?

Mason Pashia: Colorado, yep, just outside the Rocky Mountains. My brother got married. Congratulations to Brock. Wonderful. One of the highlights, maybe the most… I don’t know. A couple highlights would be that the bride surprised everyone with a dwarf Highland cow that had a bunch of the drinks on a saddle on its back, and it was, like, the bartender. So it was very cute.

Nate McClennen: Well, mine was for my niece, the first of that generation to get married, and we did not have any dwarf Highland cows in attendance, but it was a good time for family gatherings and a good reminder of how important family is. Well, good. Well, we haven’t had a chance to catch up.

We’re kind of on our three-week schedule, and that’s okay. And I think we should just jump right in. I think we have… We wrote down a lot of things here, so I have a bunch to share, and I know you have some comments as well. So let’s just jump in. How about that?

Mason Pashia: I love it. Let’s get going.

Nate McClennen: All right.

Enrollment, Innovation & Demand

Nate McClennen: So here, first one is… this is a great report.

I had thrown it in Slack the other day, and I was listening to it again this morning, by Tyton Partners. And so Tyton Partners is both an investment firm, but also they do a bunch of thinking about education. And this project was sponsored by Walton Family Foundation called Choose to Learn 2026.

And it was basically… they were doing a survey of schools over the last decade, systems and schools over the last decade, and trying to understand a relationship between enrollment, so did the enrollment increase in those schools, and then looking at enrollment increase, decrease or stay the same, related to number of innovations.

And this is focused on high school specifically. And so by innovation, it was changing the bell schedule, staffing innovation, pathways, mastery, all the things that we talk about for changes and transformation work in schools. So we have the number of innovations, the enrollment and then the third variable was participation.

So student participation in those.

Mason Pashia: Our own heart.

Nate McClennen: I know. It’s like it checked all the boxes, green flags all over the place. So worth reading, but let me give you the summary because it’s fascinating. So number one is over the last decade, and this doesn’t come as a surprise to me, but it’s a reminder of when we actually look at the data, there’s something different happening than sometimes what we perceive.

But there has been, at the high school level, a large increase in redesign and transformation efforts. So more and more schools are looking at different types of approaches to learning at the high school level than in the decade prior, right? So there’s a big change there.

But participation in these has been low, and what they found related to enrollment, because enrollment is, we know, a big deal. We’ve been writing a lot about demand and writing a lot about the challenges to enrollment with privatization, with ESA, with the new federal tax credit, with demographic shifts and declines.

All those kinds of things are going to challenge public school enrollment, which will affect budgets, etc. So they found that it wasn’t just offering transformation experiences, but it was participation in those, and when there were larger numbers of students participating in these redesigned systems, then the enrollment in those districts tended to increase, right?

So it’s sort of an interesting idea. It’s not if you build it, they will come, and then enrollment will potentially change. It’s if you build it, you better make sure that they come, and then you may have a corollary effect where you might have enrollment increases, which is something that’s going to be really important in the next two decades as we think about these larger shifts.

So this ties into the work, the things that you and I have been talking about a lot, which is this demand side. We are doing a lot on the supply side, and Getting Smart does a lot of this, but also so many other intermediaries and districts and schools. But we’ve got to pay attention to the demand side, and we can’t just have empty programs.

Or the other thing they noted was programs that are just small. So it might be a single CTE pathway that’s very innovative and has a mastery-based approach rather than a whole-school approach, which might be redesign principles or wall-to-wall pathways where every kid could have them and is required to take one.

So just a really interesting thing that validates some of our thinking and extends some of our thinking.

Mason Pashia: Yeah, that’s super interesting. I mean, it really echoes what we’ve been hearing. I think we’ve said before on the show, some of the most innovative schools that we know of really don’t have a waitlist. It does not mean that people are going to come.

Even if it has a series of decades that have… like, it’s been around for a long time, that doesn’t necessarily mean that it is one of these kind of enrollment-driving innovations. And maybe that’s because of the scale. Maybe that’s for this other reason. I got the chance to look at the report for, like, 5 seconds before this call, but I noticed something at the end in the key recommendations where they were advising funders specifically to fund beyond the pilot, so to…

Nate McClennen: Yeah.

Mason Pashia: …fund non-pilot activities.

And I think that this is something that I keep seeing really in every sector right now, where it’s, like, this emphasis on sustaining and maintenance. That is, like, the hardest part to fundraise for after you’ve actually tried something. I’m just curious how that sits with you. I know we’re doing a lot of work that is starting these initiatives as well as trying to help people kind of shepherd them along. But how does that feel to you, this pilot versus sustaining spectrum?

Nate McClennen: Yeah. I think that’s really interesting. Now, I saw that as well, and I think that was the theme of this paper a little bit, is where funders should focus their efforts. And I see, I guess, three things, three stages, right? So you have your early design sprint, PDSA-type cycles, continuous improvement cycles, where you’re learning about the efficacy of a particular approach, ideally driven at the classroom level.

We’ve talked a lot about bottom-up innovation, and our podcast with Rebecca Wolf was all about that, and the research supports that that’s important to set a spark. You then might have a pilot that’s making it more broad than just the individual classroom design sprint or things like that.

But then, of course, you have long-tail implementation, and how does it sustain over time? And I do think there… it probably matters what stage the school is in, in terms of their transformation efforts, and sometimes they just need sparks at the R&D level.

But when a school gets beyond that, I think sometimes the energy dissipates. It’s not as exciting. You’re trying to increase participation in this great thing that you’ve built. That’s not as exciting, perhaps, as building the great thing. And so we might have some interest lag from all stakeholders.

And so I think it’s an important piece, and often because it’s not as flashy, it probably runs out of money, right? It runs out of interest from a board, runs out of interest from a superintendent or a school principal or out-of-school, whatever the case may be. So I do appreciate that recommendation, and I think foundations should spend more time thinking about what’s the long tail once something is proven to have efficacy.

So.

Mason Pashia: Yeah. It also makes me think about local foundations and regional foundations getting involved too, where you have a little bit more of this proximate relationship to the results rather than somebody who’s funding it from somewhere else and kind of dropping in to fund and then moving on.

Nate McClennen: Totally, totally. And it just… I think funders really focusing on how do you increase demand, that actual question, is a really interesting one that you and I have wrestled with. It is not enough to just build these things. It needs to be that every student really wants them, because we know that when we build these transformative models, they’re based on learning sciences.

So if you take mastery-based, we know that mastery-based is important and it has better outcomes in the long run. But if students don’t want it or they don’t see the benefits of it, it won’t matter, and they won’t get the benefits of that. So I think we need to continue to push hard on demand and how do we build demand?

How do students see that? I know some of your work… you’re doing work with students and trying to have them understand and trying to get them to elicit, like, what’s a good school? What does a good learning environment look like? And maybe that kind of work builds demand because they have ownership over something that they previously didn’t have ownership over.

Mason Pashia: For sure.

Student Influencers

Mason Pashia: That’s actually a really nice segue into my next point, which is a short and quick one, but it’s just a story I really enjoyed. It was on EdWeek, I think a month or so ago, but it was about how in Detroit they’re starting to recruit students to become student influencers.

So they’re seeking out students who have kind of a communications following or a knack, and they are essentially giving them a title of student influencer, inviting them to all these events that are sometimes, like, an assembly or something they’d be at anyway, but sometimes it’s at a fundraiser or sometimes it’s at a thing that’s more board-sponsored. And the result has just been this really interesting buy-in from these students to be like, “I’m invited to these places that seem important,” and the city is actually taking notes. So one of the details in the piece that got me excited was there was an employee from the city’s transportation department at one of these school events who came up to the student and was like, “Hey, would you mind actually covering this event?

We’d love for you to post about it.” And so this… it’s, like, a really interesting way in a time when every student you ask what they want to be when they grow up, they say a YouTube personality or an influencer.

Nate McClennen: Right, right, right.

Mason Pashia: And it’s a really great way to lean into that and invite them in on this kind of important role for building demand of your organization, told through the eyes of someone who experiences it every day and in a super unique way. So I’m all for it. I think a fleet of student influencers is not the worst thing for these funders to consider from a demand-side perspective, but also for every district to consider.

Nate McClennen: All right, so here’s the demand-side formula I’m thinking about right now based on this conversation. So we have this fleet of influencers as variable one. We need to increase those. Student perspective on school design, so the work you’re doing in Issaquah and other areas, so that’s number two. That’s the second project.

And then third would be… let’s see… the third one would be how do we popularize our innovative schools? So this is the Hollywood play, or this is the… how do you promote it to the general public?

The general public. Getting Smart is great. We have a lot of audience members who are educators in that space, but it’s not necessarily parents or the general public who are not involved in education. So we need that third variable about something that we’re writing and elevating and amplifying stories, and I think some of the stories we’re doing in Getting Smart with the stories of transformation could get at that because they’re interesting enough for the general public to hear.

Then you get student voice plus student influencers. I’m going to take those three as our recipe right now for increasing demand.

Mason Pashia: Pretty good recipe. I like it.

Nate McClennen: Yeah. All right, let’s see. I’m going to pivot.

AI Bias & Homogenization

Nate McClennen: We always have to talk about AI, and so this was an interesting study. Actually, it’s a year or 2 old now. But the premise of the study is around feedback styles given by AI when submitted a piece of writing, okay? So it’s a writing piece there, and it’s about bias. And so what they did is they took the exact same piece of writing, a response to a prompt, so that was a fictitious student’s response, and then when they fed it to an AI evaluator, they told the AI evaluator some characteristics of that student.

So was it male, female? What was the race? Motivation. So a couple other demographic-type statistics, and they made variability on that. Oh, and then the other one was about motivation, so student motivation, low or high. And so they did a bunch of different personas, and they fed it into the AI assessment.

And not surprisingly, the evaluation that came back, the written evaluation that came back, used different language based on the different personas that they were fed, even though the essay they were given was the exact same every time. They being the AI, right?

And there was a bunch of different examples in the paper I’ll put in the show notes because it was super interesting, but for example, if the AI was fed that the persona was a Black, Hispanic or Asian female, unmotivated, with learning differences, they received less constructive criticism and more praise, reflecting both feedback withholding, right?

But on the other hand, if it was someone who was a male, white, motivated, high-achieving student, they were treated more distantly, so positioned as more capable of handling critique, etc. So when AI has these biases built in, I think it’s just a big heads-up that more and more educators and school leaders are supporting this, are saying, “Hey, we can use tools to evaluate,” because it gives rapid-cycle iteration way faster than a teacher can, but the biases that exist in humans are being replicated in the large language models.

Mason Pashia: Interesting. I mean, this is just another drop in the bucket of AI helping us understand ourselves better, right? This bias has actually existed forever. We also are always going to be kind of not objective when giving feedback, just based on assumptions or who we think the person is.

So it would be so cool if after all of this weird AI experimentation that we’ve kind of just been thrown into over the last few years, if people were able to be like, “This is the way I like to get feedback, and I’m going to put that in at the beginning.” If you can actually define the things about yourself that are kind of ineffable or we’re not asked to spend time with, this deep internal reflection, I think that could be really, really powerful, and it would help the AI actually be accurate and useful. So I don’t know, that’s very rosy-colored glasses, but if what comes of this is us understanding ourselves better so that we can have the machines not be biased and not hallucinate, that would be really great.

Nate McClennen: Yeah, I mean, you’ve talked about this before, and I think it’s really poignant, right? This idea that AI is a mirror of human existence, at least the part of human existence that’s been written down, recorded in audio or video, which we know is not an entire set. And that is full of all sorts of flaws that make us human.

And so this is reflecting back and demonstrating that. I think making sure that people, that we writ large recognize that and then reflect on it… so where’s the reflective process as all these big AI companies are going public, and there’s more and more prevalence, etc.? How do we use it as a mirror, I think, is really important.

So really good point on that. So that’s just an interesting one, just further affirmation that we need to be reflective and need to be thoughtful about using this really powerful tool that’s emerging.

Mason Pashia: Totally agree. Kind of on the same theme, but maybe from a different direction, Rebecca Winthrop published a good piece in The New York Times about a week ago that was really a little bit under the hood of this idea of homogenization of AI, which in some ways your piece, or the thing you just shared, is sort of counter to that.

It’s talking about this more personalized, bifurcated…

Nate McClennen: It’s more divergent, right? Yeah.

Mason Pashia: And this is actually talking about how it brings it all toward the same point. So basically this… her piece is kind of a think piece on a few different studies, but one of them is this study at Georgetown led by Adam Green that’s been going on for about 8 years, and his team examined 370,000 students and, particularly, their admissions essays for college. And so basically the study found that after ChatGPT became available, their essays got very widely diverse in terms of the language used. So style, much more floral in some cases, much more simple in some sentence structure, all that stuff, the formal elements super distinct, and yet they all kind of converged around the same few ideas.

So it was not very original in terms of the content at the center, but the kind of makeup around it was very creative, and it actually fooled the judges. The judges came in and said that these essays were more creative, even though students were all more or less saying the same thing, just in different ways. In a separate study, they basically found that human-written essays offer up to 8 times more new ideas than AI ones, which is super interesting. I can’t tell if that’s a prompting problem or an AI problem, but that is a fascinating trajectory. And in continued versions of this study, they just keep finding that while the language and the kind of dress-up changes, the content is less and less dynamic if you’re using an AI model. And I just found that super fascinating. I think that’s something that we’re just going to have to keep living with, and it’s sort of like a fool me once, fool me twice problem where just because it sounds good doesn’t mean it is good. And there’s a little bit of a… I don’t know, maybe a person in me that’s like, maybe there are just not that many creative ideas for an admissions essay, right?

Like, maybe that’s a failure of the question if the content is always kind of circling the same themes, and maybe there are actually only a few things to write about, and it is how you voice the thing that is distinct.

Nate McClennen: It’s this idea of convergent but nice-looking, right? So I found that with my own writing, as AI has entered my own writing, as I have played with different ways to do it. Like, do you take an outline and then have it create its first stab at it, or do you have it reflect on it afterward?

And I have found that in writing blogs, if I start with my outline and say, “Hey, help develop the blog,” that is way more convergent and boring than if I write the whole thing first and then send it to the AI to say, “Hey, can you check my grammar? Can you check the way that it comes together?”

And that’s not unusual. I think we’ve been seeing that in research and how teachers are using it. But the cognitive offloading is going to happen, and it seems like that’s the case with these essays. College essays is a perfect place to cognitively offload stuff. And more interesting to me is that the judges, the human judges, were fooled by the flowery language, right?

So we’re actually getting essay slop, right? Just like we talk about work slop and all the other slops that are out there, we’re getting college essay slop. And it would be interesting if they could run that question that you said. You take 370,000 essays and see what are the 10 most dominant themes and what % of the essays had those themes.

Because I bet you it’s 99% of the essays have the same 10 dominant themes. And that may have been true before computers as well, because there are things about humans that they write about for college admissions essays.

Mason Pashia: No, I mean, I think it’s this weird thing where this is as much about the person who drafts the response as it is who drafts the question.

Nate McClennen: Yeah.

Mason Pashia: As we sit more and more with these technologies, people are going to have to get at, like, what is it we actually want to get at with this question?

Like, if you are writing the college admissions prompt or question for the person to respond to, are you trying to figure out their identity? Like, who they are…

Nate McClennen: Right?

Mason Pashia: …in the world? In that case, language is actually a really good indication of sort of how they navigate the world and these kind of syntactical patterns they’ve picked up. Or are you wanting something more discrete, like a solution to a problem? In which case, that is something that you could get maybe a lot of different answers for, but the presentation of it might actually look kind of similar if you’re trying to do this experiment in a vacuum. So I don’t know.

I think we’ll learn a lot about how we ask questions, how we run simulations for hiring or admitting or anything.

Nate McClennen: And I bet that there’s probably some rubric behind the evaluation system of these essays, right? So, well, fascinating. All right. Well, let’s go to our last short take, which has turned into long takes, but they’re really important. This idea of the importance of being an intern, and you and I have talked about jobs and interns.

You want to lead us off, or do you want me to lead off on this one?

Mason Pashia: I mean, I can lead it off. Essentially, they’re just finding that… So USA Today did this report, but we’re seeing this cross-pollinate in a lot of places, and we’re going to shout them all out here. But essentially, there’s so much competition right now for entry-level roles. We know this. We just opened, ran an opportunity for an internship, and there was a flood of people who were interested kind of right out of the gate from all levels of experience. But basically, having had an experience before is the greatest indicator of getting a job, and that can be really any experience. Whether that is a vertical stacked pathway, whether that is sort of periphery but demonstrates some of these more professional skills or more durable skills. It’s really about having experiences, and I think we’ve said that a lot of times. That’s what our experience matters and kind of larger credentialing of experiences work has been about. It’s that you need to have these experiences, you need to recognize these experiences, and you need to be able to interpret from these experiences which things can apply to other ones. And people are having experiences all the time, but really there’s only a handful that currently count as a line item on a resume, and that’s a problem that we’re hoping to solve. But this specific callout is around the importance of internships and around the importance of having experiences in order to then get more experiences.

Nate McClennen: Right. It is the door to early employment, I think, now, having those experiences. Yeah, the first survey… I came through it with a Gap Letter from Achieve Partners, and they write about workforce all the time. The specific data was that report and the survey finding that working during college more than doubles graduates’ odds of landing a job, right?

That’s a pretty significant statistic. And I’m curious. So my daughter’s home from college, and she started her work at a restaurant yesterday and then started her internship today, and I’m going to ask her when she gets home tonight, I’m going to say, “Was there a difference between the 2?

Was there crossover skills?” Right? I actually have a really good study of a sample size of one going on in my house for this summer. So it’ll be fun to see. But it does go back to this pushing idea that experience is our great proxy for competencies. And while we are bullish, and I think the system is bullish, on how do you measure competency and give practices with competencies, these durable skills, there is a great way to do it: set up great experiences for people, whether they’re paid jobs or paid internships or whatever the case may be.

And the big enigma is how do we create a system and a framework, which I know you’re working on for that, and then how is it then reported out via signals? Super interesting. And I think it also comes into this conversation of 2 things. One is that young people are working less during the summer.

And I also saw some specific reports of this summer it being hard for young people to get a job. So there may be this percolating effect of young people now realizing that it’s advantageous to have experiences, so now demand’s increasing, but supply may be down because of the economy or whatever the case may be.

Mason Pashia: Yeah.

Nate McClennen: And I guess finally, as we know, Brightbound did a survey, and this is well cited. It’s amazing how much this statistic has percolated out into the ecosystem. But they did a survey and found that almost 80% of students were interested in trying a work experience, but only 2% completed an actual internship in high school.

I suspect that way more than that actually worked in jobs, and that would be an interesting side survey now that we’re seeing a crossover between internships and regular jobs, is that they may have some duplication in what skills the students learn.

Mason Pashia: Yeah, and I think we’re seeing education systems shift as work-based learning grows, as all these other things grow. I saw, for example, I think it was also from Achieve Partners actually, that MIT just announced a committee to study whether they should offer their own co-op program, kind of modeled after Northeastern’s, which we’ve got some great pieces about on the site from Chris Unger. But if MIT is feeling the need to change to address this, that means it’s a big challenge, and I know that’s near and dear to your heart. So it’s fascinating.

Nate McClennen: Yeah, I saw that as well coming out of MIT, and it is fascinating because students at MIT have a really large number of research opportunities already, working in professors’ labs, and so they’re feeling even that they need more, as in, how can we actually get students out into the working world, not just research labs, to get credit for it?

And I will be very curious to see, especially if liberal arts colleges follow suit in some way. I think there’s this real divide between technical schools, sort of STEM schools, versus the liberal arts. And liberal arts have had experiences where you can go get a job on campus, and they have some internships, but it’s not a big part of their program.

So I will be curious to see in higher ed if they follow suit with what’s happening in the technical schools. All right, for deep dive…

Mason Pashia: I’m ready. I mean, we’ve been doing it, so let’s just keep going.

Nate McClennen: Let’s just keep diving. We’re going into the deep end, so we’ve been in the shallow end underwater for a long time, so we’re just going to keep swimming to the deep end.

Mason Pashia: Keep.

Nate McClennen: All right.

Shorts Content

Deep Dive: Agency & AI’s Role in Education

Nate McClennen: We have 2 big topics today. So the first one is on agency, and the second is on how AI is affecting teachers’ preparation and support.

So number one around agency. So let’s just do a quick background. Agency. So self-determination theory from psychology. I think you and I have talked a little bit about this, but this idea of students are more motivated when they feel connected, so they feel like they belong.

It’s a safe space for them. When they can demonstrate mastery, self-efficacy, they actually understand that they learned something, and that they have autonomy and agency, that they have control, and that they have purpose, some sort of relevance. So all 4 of those things are connected to the self-determination theory.

And so there’s really good research that when those factors are in place, motivation increases, academic outcomes increase, mental health, learning, etc. So I think there’s a solid body of evidence that agency as a part of self-determination theory matters. However, here are the data points.

In a survey of Iowa middle school students, just under 50% feel that they’re able to learn at their own speed, and this drops to just 35% among high school students from a Gallup poll in 2025. So we know that as the student gets older, their ability to choose goes down.

Yet 90% of adults feel that they have agency around all sorts of parts of human flourishing. So control over their economic well-being, their social well-being, purpose in life, etc. So when they get to be adults, they may have more agency, but when they’re in school, they have less agency.

So Antonio Perez writes a Substack. He’s from Walnut Hill Ventures, and he prompted me thinking about this, and so kudos to him, just about talking about how more and more schools are saying we need to develop agency in young people, but there are dangers there if we’re just saying, “Oh, agency is this.”

And the concerns are… I’m going to read his quote directly because I think it’s really well done. He said, “We are drifting towards the same mistake we made with reading. A real and important human capacity is being discussed,” that is agency, “in a way that gets fuzzy fast on how it is actually built.

Agency is not what happens when adults disappear. Agency is what happens when adults create the conditions, model the moves, support the struggle and then release responsibility over time.” And I think that’s super important, right? Really profound about this idea that if we just say we’re going to give students more agency and suddenly you’re giving them 100% agency, that’s not going to work.

And if we confuse agency with adults stepping away and not doing anything, we’re in real trouble. So I’m going to pause there for a second and make sure we’re on the same page about agency, and let’s talk and riff a little bit about what we think schools might be thinking about in this area.

Mason Pashia: I really like that framing. You know, I tend to… I kind of bristle at the word agency sometimes, which we may have time for me to dig into a little bit, even…

Nate McClennen: Bristle away, Mason. Come on, bristle away.

Mason Pashia: I’m turning into the “yeah, but” guy. I don’t need to be the “yeah, but” guy. I’m curious, though. One question I have is this 90% of adults feel like they have agency around human flourishing.

Nate McClennen: Right.

Mason Pashia: Does that mean that we are successful in teaching agency? That’s just an interesting point because I think the way that we think about so many skills and dispositions is that they are a progression of learning, and then at a certain point in your life you have mastery of it in some way.

And 90% is a high number, and I’m surprised by that. I know a lot of people who would say that they don’t feel like they have control over whether or not they can flourish within their context or setting. But I don’t know. Do you…? I guess first question.

Nate McClennen: Yeah. I mean, are we being successful? Right? Maybe we have to do nothing, because if that data… And that’s one survey. It was 2,000 people from Archbridge Institute. And so we’ll put that in the show notes. And again, surveys are surveys, and they were broad sets of questions.

And it was about… they used the word control. Do you feel like you have control over X, Y and Z? So I think there’s really probably a pause to say what do we really mean, and how many adults really feel like they have agency? However, almost by default, when you become an adult, you probably feel like you have more agency because you actually are forced into it.

You’ve got to do things that are required. And so the question is probably not whether or not humans develop agency, but whether or not agency is a tool we can use earlier to help increase outcomes, right, in school. I guess there are 2 different things going on there.

Mason Pashia: And I think Antonio’s getting at some of this, but that feels to me like those 90% of adults feel like somebody pulled out the rug from under them and now they’re free-falling, and that is a version of agency. It’s like, “No one’s really making the decisions for me. I kind of have to do it.” There’s nothing in that that is implicitly, “I’m loving this.”

I think that if you introduce a metric around flourishing and thriving and loneliness and all these other things that we’ve seen, you’d be like, “Oh, actually, a lot of people are really struggling,” which I think to me is representative of the fact that agency might not be embodied.

Even if people feel like they have agency, it is not a skill that people know how to flex the muscle or use in a way that serves them. And so that is a really hard thing to measure. But I think that’s ultimately what we’re talking about, is that bridge between feeling able, feeling like you have permission to do something, and then actually bridging that to the thing that makes you feel good or arrived or whatever.

So I think that’s interesting. Agency for me, right now, is being… if you could fund a word, I feel like agency is a very funded word in the world right now. It is like everyone in Silicon Valley is talking agency. Everyone who has put out an AI product is just like, “Oh yeah, we’re building agents, but you just need to be more agentic.”

Nate McClennen: Yeah.

Mason Pashia: There’s this weird thing going on where we’re kind of using the word to mean a thing and its opposite, which is like the person needs to have the skill to exist alongside agents. And so I don’t know. It’s a word that I’ve always felt a tension with because it is so party of one, and I think that there’s actually something probably more at this intersection, this interconnected… Like, what is agency in a group, right?

Nate McClennen: Yeah.

Mason Pashia: We talk all the time about, and I know I’ve probably even said this on Catching Up, but there’s this adage on LinkedIn right now that drives me insane, which is basically, “It’s never been easier to build something.” They chalk that up to AI. And to me, it’s actually never been harder to build something in the world. Like, if you try to build a road, you’re screwed for 20 years.

Nate McClennen: Right, right, right.

Mason Pashia: And there’s all this kind of bureaucracy, and just teamwork is really hard, and coordination’s really hard, and I know we’ve been thinking about that with Sangeet Paul Choudary’s work on how AI can empower that. Anyway, I’m going on a rant, but basically that thing is… agency in a group, I think, is actually something that implies values in a way that I’m more receptive to. And I don’t know if I see that many people talking about that. I think it’s beyond collaboration. It’s kind of collaboration, courage, agency together. So I don’t know, but that’s my initial reaction to this thing you’ve presented to me.

Nate McClennen: Yeah, no, it’s really interesting. So I think relevant to our audience, right? So if you’re in a learning space or you’re an intermediary or whomever, and you’re talking about we want to build student agency, I think it’s worthwhile to actually… and Antonio’s essay talks about this, right?

He’s asking us to say, “All right, let’s take a step back and actually ask what are we trying to do here? What are we trying to achieve?” Because I do think it’s being overused, and I’m guilty of that because I think there’s some power there. I also think it’s interesting how you talk about this as this… it may be connected to this individual versus common-good purpose of education that we’ve been talking about, is that individual agency, I have choice over my own learning.

I can choose this part of the playlist, or the teacher gives me these 2 prompts and I get to choose them, or I get to write my own prompt, or I get to go figure out my own real-world learning experience, is very different from agency. I feel that along with a group of my peers, I can go make a difference in the world through a collaborative, agentic process where I can make a lot of choices, etc.

And that is more of a common-good type endpoint, whereas… I mean, agency’s neutral. The word is neither positive nor negative. You can have agency to do something terrible, and you could have agency to do something really good, but I hadn’t really thought about the group versus individual.

So I think the takeaway here from this deep dive for me is there are a few schools out there that are thinking about this really, really deeply. So like Red Bridge out in California, who we’ve profiled before, they have progressions around agency that are separate from their performance.

And so you could have someone who is high-performing in, say, math, but they have low skills on the agency progression, which then gives them a different learning environment because of that. And so I think there’s some sense of how can we be more specific about it? How can we think clearly about how does it affect the individual versus the group?

I think there are some real takeaways here, but we’ll put the Substack in our show notes. I just think it’s worth looking at. He has some provocative questions in the end about it.

Mason Pashia: And I actually really like thinking of agency as one’s ability to control their attention. We’ve talked a lot about attention, but I think if you’re talking agency, in a moment where you have TikTok on one hand or hanging out with a friend on another, that is a skill that we need to just keep building as a civilization, to just choose this one path that is more generative to you.

And obviously not every time. It’s not always the right move. There are instances where that’s not the case. But I really like agency as thinking of you almost as, like, a spotlight, and the agency is to put it on certain things. But I think it becomes so much… it becomes inflated when people talk about it in this way in which you’re… I don’t know. It becomes this kind of thing that I don’t believe that it is when it gets ballooned this way.

Nate McClennen: Right, right, right. Well, we’ll see. I mean, it is showing up everywhere, and everyone’s talking about how do you build student agency. And certainly, going back to the very beginning of this deep dive, there is good evidence in psychology that motivation increases with some sense of autonomy, right?

The research would indicate is somewhat analogous to agency, that you have choice and have some control. So I think there’s a real legitimate… we can rest on that as a foundation. I think the question that you and I are wrestling with is control of what? Like, what are we actually controlling and making decisions around?

And whether that’s group versus individual, or is it being able to put a spotlight on the things that are most relevant in this attention world that’s constantly seeking to divert our attention. All right, well, interesting good things to think about. If any listeners have a school or district or division or organization that uses the word agency, it’s time to take a look at that and make sure you all understand what that means.

I think that’s my takeaway.

Mason Pashia: Yep. Agreed.

Nate McClennen: All right. We’re going to dive into our second deep dive. We’ll probably go a little bit quicker on this. 2 things here. So we are doing a lot of work around what do educators’ roles and responsibilities look like in the age of AI, right?

And so a couple different research projects. One is, does AI save time for teachers? So a really interesting project. This was run in England, and they surveyed a bunch of schools and 259 science teachers. They broke them into 2 groups. One was given a set of responsibilities and able to use ChatGPT in this case.

The other was the same set of responsibilities, not able to use ChatGPT. And what they found was that those who were using ChatGPT spent 69% of the time the control group spent, so 2-thirds of the time that it took the control group to do the same work. More importantly, the quality, when randomized and looked at by judges, was the same between their outputs.

This was like a lesson-design type experience. And so the conclusion, one, was that it can save you time on some tasks that a teacher typically does. And then the second conclusion was that time got reallocated to things that are more uniquely human, whether that was building relationships with students and things like that.

This makes sense, but we’re starting to get research out that supports that.

Mason Pashia: Yeah.

Nate McClennen: And then I think that there’s another piece of evidence around this, or research around taking a number of teacher conversations from AI, so a totally different experiment, about teacher use of a chatbot or some sort of large language model.

And they looked at over 13,000 different conversations, and the predominant uses for AI were around lesson planning, differentiating instruction, assessing student learning, like we talked about before, and reflecting on practice. So I kind of wonder those 2 things. One, does it allow us to reallocate our time?

And 2, these are the things that at least the initial uses of text-based bots by teachers are showing… here’s what they’re doing. Does that jibe with some of the research that we’ve been doing with Ed3 in terms of the future role? How does that sync in your mind?

Mason Pashia: Yeah, I mean, it’s super aligned. I think in general, this reallocation idea and making it more of a time-saving rather than a transformative full overhaul of the system, that’s kind of right where we’re sitting. So if anybody hasn’t listened to Nate’s conversation with Viridi Saraf, they had a great conversation about this report that came out about a month ago.

We’ll drop a link in the show notes, but we’re working with them to think about a portrait of a teacher in the age of AI, and it’s just been really interesting. I mean, more and more educators are saying that they use AI. Over half use it weekly or more often. But at the same time, and as anybody knows who’s used AI that’s listening, it’s pretty hard to make it a transformative technology right now.

It is something where you can come up with a real time-saving solution. You can be like, “Oh, wow, I didn’t know it could do that.” But actually embedding it as a transformative system takes a lot of work, it takes a lot of trust, and I’m just not 100% sure that we’re there yet. But I’m really glad that we’re getting some research out there because we’ve been kind of saying this on a lark for a few years, just that it’s possible that it’ll free up time for teachers to do more and better with different things.

And I think that the more reports, the better, to suggest where AI can enter the classroom and the teacher workflow.

Nate McClennen: And maybe it connects back to the first short piece that we talked about, where the most challenging part of transformation happens, and that long-term implementation tail that we talked about. And if educators had more time in that phase, I think they could…

If they have more time anyway, whether it’s in the early R&D phase, the pilot phase or this long-term implementation phase, that could help with this transformation work, with the caveat that the system they’re in or they are equipped to do the transformation work, because the reverse can also be true.

You just do better and better at the things that are traditionally done in schools, and that may not be beneficial for young people.

Mason Pashia: Totally. No, I agree. And I just want to round out this kind of final AI segment for today. Unfortunately, another very AI-heavy one. Sorry, listeners who are tired of hearing us wax about this. But I was poking around and saw this kind of opinion piece, but it’s from the Joan Ganz Cooney Center, which is the Sesame Workshop that is of Sesame Street fame, and they wrote a piece kind of reframing these aspects of AI design, and I just found them to be really succinct and effective. So I just wanted to run through those real quick. I think this also informs some of the ways that educators can think about using AI, but what this is really describing is if we designed AI systems to support human flourishing. So the first one is shifting from interactive to relational. So that is something that really takes the learner, takes the student and then puts them back into a web of people and belonging rather than just having them interface. I think that’s sort of obvious, and we’ve talked about it, but I actually think that it’s a pretty stark interrogation of the word interactive, which many people are prone to just throw around willy-nilly, and that’s kind of true of all 4 of these actually. The next one is to shift from personalized to pluralistic. So this gets back to what we were talking about at the beginning today a little bit, but this is saying that AI systems tend to just mirror back you. So maybe this is the bias feedback coming in. Maybe this is sort of this way that you’re writing a college admissions essay.

But instead, we actually need AI to create a pluralistic approach that expands the person’s world rather than keeps it so honed in, which we’ve seen over the last few years algorithms do time and time again. So that’s kind of interrogating the word personalized, changing it to pluralistic.

Nate McClennen: Yeah, yeah.

Mason Pashia: The next is from adaptive to developmental.

I think we’ve been guilty of using adaptive. I feel like a lot of the systems… AI was adaptive, or we used adaptive to kind of encompass machine learning for many years. But this is really talking about the thing that we may talk more about on a future episode, but like how when you go into these new language models, you can suggest basically, “Keep challenging me,” like, “Don’t give me the answer.”

Nate McClennen: Hmm.

Mason Pashia: And so this is really about helping you scaffold learning rather than providing answers. And I think that’s a really smart design technique to bake into at least some of these models. And then this last one, I don’t 100% know how this works yet. This was the most fuzzy to me.

But it’s a shift from intimate to equity-oriented. And I don’t know if I actually like that framing as much, but I think what it gets at that I do like is that it’s asking, “How do we make these models not surveillance, and how do we make the inner life of these people that are using it get more than just mere data out of this,” right?

Like, how do you fundamentally think about lived experience as something that serves the person who is using it, not somebody on the other side of the equation? And I think these are 4 really smart design principles that we just need to keep engaging with and maybe call into question some of these words, sort of like agency.

Nate McClennen: You know what’s so interesting? So I love these. Here’s what’s interesting, is that it is possible that the introduction of this generative technology that’s sort of changing and shifting the landscape in significant ways really just becomes, first, a mirror to ourselves, like you talked about before, and second, an impetus to get us to a place that’s better, a place that is emphasizing relations, a place that’s pluralistic, a place that’s developmental when we think about learning and then equity-oriented, right?

So if you take those 4 things, and I think there are others as well that we could add to this list, but maybe the opportunity here that this particular center is talking about is an opportunity that has actually nothing to do with AI, but it’s an opportunity that we can use AI to say, “Oh, by the way, we need to spend more time building relationships human to human.” I think we needed to spend more time building relationships human to human before AI, even before technology. I think this is a critical human thing that makes the world a better place. A pluralistic approach that introduces new perspectives and diverse ways of thinking? Absolutely. And that was before AI, so maybe this is just amplifying the need for these things, which maybe will give it a higher chance of becoming successful.

Mason Pashia: Yep. I think you’re right on.

Human Expression & Closing

Nate McClennen: All right, Mason, we covered a lot of ground. It is okay that we’re talking a lot about AI because I think we’re using it as a mirror for ourselves and for others that are listening as well.

Mason Pashia: We are.

Nate McClennen: But let’s end with a couple things here. We’re going to end with our human expression and then go into our song for this week.

So I’m going to start with human expression. So you know me, I am not the most poem-philic person in the world. I don’t love poetry. However, I’ve been trying to read a little bit of poetry, and Mary Oliver is in a book beside my bed because I find her poetry really easy to read. And I found one that was really, really short, and I wrote it up.

We have a piece of poster paper on our wall, and this one’s up on our wall. So it’s called “Instructions for Living a Life.” I know you know this poem, so you can laugh when I say it, but: pay attention, be astonished and tell about it, right? It’s the simplest poem in the world, and it made me think about this poem because when I was back at my parents’ house over this last week for this wedding and graduation stuff, they live in a suburb outside of Boston, and across the lawn from them, we watched a fox carrying a dead rabbit into what we think is the den for that… where the fox’s den was because it just seemed like it was actually underneath the barn, etc. And there’s a longer story here, but the point was, there were 4 of us, and we were just sitting out on the suburban porch, which is facing the street.

We’re looking across the road at someone else’s yard, and there was this fox just carrying this large rabbit. And we were paying attention. We were astonished, and now I’m telling about it. So I’m just living Mary Oliver poetry right now. So that’s what I have for you today.

Mason Pashia: I love it. Yeah. Do this infinite times a day. I think that’s wonderful.

Nate McClennen: Yeah.

Mason Pashia: And that is a great… no matter how known or short that poem is, it is fantastic. On a similar kind of human nature and wildlife theme, I saw a data point the other day that last year, 2025, over 100 dams came down across 30 states, which has now reconnected 5,000 miles of waterways for fish and cleaner rivers, etc. And I think this is important because something implicit in that is a human learning, right? That is… doing that is humans owning up to the mistakes or the things that we’ve made before. And maybe in certain places, that was the right solution for a time, but it’s important to undo those things. And so I think part of human expression is acknowledging you’re wrong or changing your mind, and there’s no more visible or beautiful way to change your mind than to remove a dam. So, pretty cool.

Nate McClennen: Yeah, that power of humility, right? Recognizing when we headed the wrong direction, even though it felt like the right direction at the time, I think is critical. So yeah, thanks for sharing that. So all right, let’s jump to music. We were supposed to have a special guest today, so this music is for Rebecca Middles, who was supposed to jump on today, but she couldn’t show up.

So I had made the music thinking that she was going to come on, so this is for her. So cue it up.

Mason Pashia: Rebecca, this one’s for Rebecca.

Nate McClennen: This one’s for Rebecca.

What’s That Song?

Song: Got your message in the morning rain
Said the school moved fast again
Tablets out and the old board’s gone
New tricks now, but we move on

They map the kids by what they need
Not one-size shoes, not the same old feet
I’m late, but I’m here tonight
Tell me what’s been coming right
I’ve been…

Mason Pashia: Oh, that’s pretty cool. There’s a…

Nate McClennen: The bass beat, the first part’s pretty good, I think.

Mason Pashia: Yeah, that’s maybe the most interesting backing track I’ve heard from one of these AIs. Like, very spacious. The guitar has a nice kind of sub, like a countermelody to the voice. Very cool. I’m getting something… It sounds very 2005, like a kind of alternative soul world. I don’t know the… I don’t know what prompt you used. I feel like Rebecca told us her favorite band, but how did you get around the…

Nate McClennen: Oh, I just put in, “in the genre of,” and I cut and pasted her bands: Ry Cooder, Michael Kiwanuka, R.L. Burnside, Citizen Cope, and I just said, “Make some version of that that’s similar.” So I thought they did a pretty good job. And the words are pretty funny because, you know, “the door swings wide and the teacher has new life,” and then “bring me up to speed,” and then they talk about students having small minds.

I mean, it’s just a funny set of lyrics. But “bring me up to speed,” Catching Up, there’s our song for today. And I actually think it’s one of our best ones yet in terms of the music.

Mason Pashia: That’s pretty cool. That’s a fun one. Yeah. Good job. And I wonder if Rebecca could tell that that is from the genome of Rebecca’s musical taste, so we’ll have to share it with her.

Nate McClennen: Yes. We will share it with her for sure. All right, Mason. Thanks. That was awesome. Great to catch up. Good long session today. A lot of learning. Audience, hope you enjoyed it. Give us your feedback. Send us things you want us to talk about.

Mason Pashia: Right back at you, Nate. Always a pleasure. Have a good day.

---

Links

• Watch the full video here
• Can Student Influencers Help This District Rebuild Enrollment?
• Personalized Bias in AI Writing Feedback
• Percentage of High School and College Students Ages 16 Through 24 Who Were Employed
• Shut Out: High School Students Learn About Careers but Can’t Try One That Pays
• Student Agency Boosts Engagement and Readiness
• Agency in America
• We’re About to Get Agency Wrong

Guest Bio