Reading and Writing Skills in the Age of AI | Naomi Baron, PhD

Studies mentioned in this episode, in order of mention: 

Sun, Y. J., Sahakian, B. J., Langley, C., Yang, A., Jiang, Y., Kang, J., … & Feng, J. (2024). Early-initiated childhood reading for pleasure: associations with better cognitive performance, mental well-being and brain structure in young adolescence. Psychological Medicine, 54(2), 359-373.

Baron, N. S., Calixte, R. M., & Havewala, M. (2017). The persistence of print among university students: An exploratory study. Telematics and Informatics, 34(5), 590-604.

Delgago, P., Vargas, C., Ackerman, R., & Salmerón, L. (2018). Don’t throw away your printed books: A meta-analysis on the effects of reading media on reading comprehension. Educational Research Review, 25, 23-38. 

Baron, N. S. (2015). Words onscreen: The fate of reading in a digital world. Oxford University Press.

Ose Askvik, E., van der Weel, F. R. (R.), & van der Meer, A. L. H. (2020). The importance of cursive handwriting over typewriting for learning in the classroom: A high-density EEG study of 12-year-old children and young adults. Frontiers in Psychology, 11, Article 1810. https://doi.org/10.3389/fpsyg.2020.01810

Baron, N. S. (2023). Who Wrote This? How AI and the Lure of Efficiency Threaten Human Writing. Stanford University Press.

OECD (2015). Students, Computers, and Learning: Making the Connection, PISA. Paris: OECD Publishing.

The Learning Network. (2025, May 8). What students are saying about using A.I. for schoolwork. The New York Times. https://www.nytimes.com/2025/05/08/learning/what-students-are-saying-about-using-ai-for-schoolwork.html

Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In Proceedings of the CHI Conference on Human Factors in Computing Systems.

Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing tasks (arXiv:2506.08872v1). arXiv. https://arxiv.org/pdf/2506.08872v1

Stanford Institute for Human-Centered Artificial Intelligence. (2025, February 25). Annual AI+Education Summit 2025: Human-centered AI for a thriving learning ecosystem. Stanford University. https://hai.stanford.edu/events/human-centered-ai-for-a-thriving-learning-ecosystem

Association of American Colleges and Universities. (2025). Student guide to artificial intelligence. AAC&U. https://www.aacu.org/publication/student-guide-to-artificial-intelligence

[Kris Perry]: Hello and welcome to the Screen Deep podcast, where we go on deep dives with experts in the field to decode young brains and behavior in a digital world. I’m your host, Kris Perry, Executive Director of Children and Screens. Today’s guest is someone who’s helped us better understand the magic of the written word for decades. And now, she’s turning her attention to how generative AI is on the cusp of changing our relationship to reading, writing, and learning. 

Joining us is Dr. Naomi Baron, linguist and Professor Emerita at American University. Naomi is a former Guggenheim and Fulbright fellow and has authored several groundbreaking books, including Who Wrote This? and the forthcoming Reader Bot: What Happens When AI Reads and Why It Matters that explores how digital media is reshaping our relationship with language, cognition, and culture. In this conversation, we’ll talk about what happens when AI reads for us, what we risk by outsourcing our writing, and how parents and educators can help young people build cognitive skills in an AI-saturated world. 

Let’s get started. Naomi, I’m so excited to talk to you today because you are not only one of the world’s foremost experts on how humans learn to read and write, which really gives you incredibly strong insights on how the rise of generative AI is interacting with and impacting those human skills, but you’re also one of my favorite researchers. Before we get into the AI side of things, build us a basic understanding of reading and writing skills. When we’re talking about literacy skills, what should we keep in mind about how this relates to cognition or our general mental and intellectual processes?

[Dr. Naomi Baron]: First, thank you very much for welcoming me to this incredible podcast. I’m glad that you qualified with, “Well, there’s a little more than just the cognitive part when you think about literacy skills,” because I’m going to go even further back and say you have to ask, “Why do we have reading and writing in the first place? How did they evolve?” They’re not all that old. Writing, as best we can tell, goes back to around 3000 BC, which is just a blink in humankind. And each time writing has evolved, because writing has evolved many times in different parts of the world, it’s been for different reasons. Often they were administrative reasons. They were religious reasons. They weren’t to develop our cognitive skills. So when we think about what it is we want from reading and writing today, it’s really important we ask, “What is it that our values are today? What is it that our needs are today?” Because they’re not necessarily the same that people had a hundred years ago, a thousand years ago. 

That said, we know that reading is not something the brain was designed to do. And in fact, when we read—when we learn to read, we rejigger parts of the brain in order to handle this new human invention called reading. And you can see this by, you know, taking adults who were never literate, imaging their brains-–having them develop basic literacy skills, imaging their brains, and their brains have changed. Probably the same can be said for writing, but we can only really measure it easily for reading because usually people learn to read before they learn how to write. 

It’s also important to know that you can get neurological measures of what reading is doing to the brain, even in young kids. So there was a study done recently with about 10,000 young adolescents imaging the brains of people who read a lot for pleasure and those who don’t. And their brains are different. That is, you can measurably see what a lot of reading, in this case because they wanted to, does to us.

We also know, though, that reading and writing—literacy skills-–are important for things and involve things other than just cognitive development and cognitive skills. They involve embodiment. When we read, we don’t just read with our brains. We read with our hands, we read with our bodies, we read with the noise around us. Similarly, there’s a social component. Much of reading is done in a social context. We talk with people about things that we’re reading. We write for other people. And it’s important in thinking about what it is we want to develop as skills in our children, to understand it’s not just getting higher scores and saying, “Okay, I think you have cognitive abilities that are better than they were because you can read and write now.” But what about social interaction? And what about our feeling of feeling the writing and feeling the reading as we’re doing it? And I bet we’re going to get into more of that in this conversation.

[Kris Perry]: We are. In fact, I really was hoping you might expand a little bit on how to help children develop those reading and writing skills.

[Dr. Naomi Baron]: The very first thing—and it’ll sound as if I’m preaching, but I’m not—is we have to make children believe—and we must believe ourselves—that developing these skills matters. You can’t just preach, “It’s important for you to learn how to read, to do critical thinking, and your reading and writing is going to help make that possible.” We have to give children reasons to believe that it might be to their benefit. So it’s hard to talk about, “You’ll develop your critical thinking,” but it’s easy to talk about, “You know, reading is a lot of fun. You can do it just for the heck of it.” 

Reading is also something that many people, including kids, turn to when they’re stressed. Reading is something that we do when we want to express our imagination by reimagining what is happening in a story. So in some of the research that I’ve done, I’ve asked kids, “What do you like most about reading in print?” And they say—and we’ll get to print versus digital, I know—“It stretches my imagination.” Or I’ve asked university-aged students, “What do you like about writing by hand?” “It makes me creative.” So these are things that people would like to be able to express themselves through, namely reading and writing. For that, developing the skills matters. 

The challenge right now is we don’t have a whole lot of support, either from an awful lot of the adult community—and that goes for parents and it goes for a lot of teachers. If we don’t model in front of our kids, in front of our students, the things we say we believe in, it’s hard for them to take us seriously. 

The other thing is we have a testing situation—and testing is increasingly digital—but even before then, so many of the standardized tests take short passages to read and to analyze, or these days it’s to take two passages and you’re reading them online and then to compare and contrast or whatever, but they’re short. And we’re not setting up a value structure for doing longer reading because we do teach to the test. That has not changed over the decades. 

Similarly, we have a challenge coming from the technology because we know that, for most people, digital reading is done faster than processing those same words in print. And if you survey students about this, they will tell you—you don’t even have to have a stopwatch— they will just tell you, “Yeah, it’s faster to read digitally than it is in print.” Hello? Same number of words. That means you are going faster. 

Also, there’s an increasing replacement in the adult world, but also in that of our children, of text, whether it’s print or digital, with audio. So you go to read an article in the New York Times online, it’ll give you the audio version and that happens in so much of what is available digitally to read. You’re told, “Oh, don’t waste your time, listen to it instead.” So we’re taking away the value proposition, as it were, of reading text. And then there’s AI, which can read and write for us.

[Kris Perry]: Well, thanks for touching on AI because you’ve done such a lovely job of establishing the importance of reading and what it does to our brains and how we weren’t built to do this, and yet we do this and we find joy in it and relief and our imaginations are sparked. It’s so incredibly versatile, reading, and it gives us so many wonderful experiences, but here we are on the cusp of full-scale deployment of generative AI. And you’ve been looking for a long time at the differences between reading from print or hard copy versus digital formats, and I was hoping you could just expand a little bit more on what you found in your research in comparing the two.

[Dr. Naomi Baron]: Sure. I’ll begin with a very quick summary of what colleagues of mine around the world have done with experimental studies, and then I’ll get to what I’ve done in what I call “perceptual studies.” Not every single study that’s experimental says the same thing, but most of them do. A strong confirmation that, if you were reading the same kind of text in print and reading it digitally, the chances are you’re going to comprehend better reading it in print. That particularly goes if you’re reading a longer text as opposed to a short amount. It particularly goes for informational text. Think of a newspaper, think of an article that’s explaining why the earth turns the way it does and the sun does whatever it does, pretty much standing still. And it also goes for anything that involves drawing inferences or dealing with abstractions. So it’s not everything that you read for which we get these kinds of results, but it’s a huge amount of the kind of reading that we do as everyday citizens. Fiction is a slightly different story which we can get into, but it’s a little more complex. 

Interestingly, if you ask students—and this has been done with, sort of, seven- and eight-year-olds, and it’s been done with older middle school and then high school and university students—if you ask them, “Do you think you’re going to comprehend better if you’re reading in print or if you’re reading digitally?” These days they will tell you, statistically, they will tell you, “I’m going to do better reading digitally.” And they don’t. So there’s a perception in, we’ll call it the “current generation,” but actually it spans about a generation or two, that, “Well, I do all this stuff digitally, I must be pretty good at it, and therefore it must be better than doing it in print.” Statistically, no. 

My own research—I, along with my colleagues, have worked with university students in five different countries. We’ve worked with middle and high school students in two different international schools. And we find very much the same kinds of things, regardless of the age. The first finding when you ask, “Do you think you comprehend,” or, “do you think you learn?” We’ve asked the question differently at different times. Better reading in print versus digitally. Overwhelmingly, the answer is, “I believe I comprehend better if I’m reading print.” “Do you think you multitask more if you’re reading in print or digitally?” No surprise here: “I multitask more when I’m working digitally.” And then we asked various kinds of open-ended questions about, “What is the one thing you like most? What’s the one thing you like least about reading in print?” Similarly, about reading digitally. And I’d like to let the students speak for themselves because they have fascinating things to say. 

So, regarding, “What is the one thing you like most about reading in print?” The cognitive issues. “Easier to concentrate,” “Easier to imagine what the author is saying,” “Makes me feel calm and relaxed.” But they also talked about the embodiment of reading in print. So things like, “The smell of the paper stimulates something in my brain to concentrate,” or, “The feeling of turning each page and anticipating what’s next,” or, “You feel like you’re actually reading.” The number of students in different countries at different ages who said, “Reading in print is real”—nobody said that about digital reading. It was astounding. These were all open-ended, voluntary answers. 

Okay, “What do you like least about reading in print?” Okay, because not everybody was sold on it. “I get discouraged because the book looks so big.” And this is important for us to understand as educators and as parents that kids can get intimidated by something that looks like too big a hill to climb. So, maybe we need to think about giving fascicles, giving it in smaller parts. And then when they’re ready, they can read the whole thing together. 

Or concentration. This is one of my favorites. “What do you like least about reading in print?” “You have to use more of your concentration.” There’s an awareness that we tend to concentrate more when we read in print, or it takes you longer to read. Well, depends on what speed you go at. 

“What do you like most about reading digitally?” Well, of course, there’s, “You can access the internet,” or for the students at the International School, if English was not their native language, they could use the dictionary, and there all these conveniences, change size of the font and convenience and so forth. But one said, “I can read faster. That’s what I like about reading digitally.” 

“What did you like least about reading digitally?” One of my favorites about cognition: “My attention span is shorter and I tend to want to skim.” So this is a lesson that I think we all need to learn, and we’ll come back to this in our conversations with other topics. Students are aware of an awful lot of the advantages and disadvantages of reading in different media, and the same is true by the way of using AI and not using AI, but we don’t stop and ask them, “What do you think? What’s your take?” So, we’ll get to some specific suggestions on how to go about doing this.

[Kris Perry]: Well, one of the things you didn’t touch on, but when you speak about embodiment and reading, for me, is—my earliest memories of reading were with my parents and being with them and physically, you know, being on their lap or sitting next to them on the couch. And I remember doing that with my own kids and just how much reading was part of a physical relationship with parents and kids. And I see that in high-quality early learning centers and even in libraries in high schools where there’s a hangout corner. And I just think there’s something, also, that you can do more together with the physical book or with physical reading material that’s really exciting, and I think it creates an opportunity for bonding, as well. 

Now I wanna pivot to writing because we’ve talked about embodiment, we’ve talked about reading and writing, of course, is something you can do by hand or you can do it with a keyboard. And I thought maybe you could share a little bit about the pros and cons that students shared with you regarding typing versus writing.

[Dr. Naomi Baron]: Sure. Let me begin with studies that other people have done. They’re imaging studies, largely using MRIs, which I do not use, but I’m glad that some of those colleagues do. Working with young kids in lower school or middle school, or even when kids are just learning how to hold a stylus, how to hold a crayon or a pencil, you can indeed see differences in the images, whether they are writing something that is print as opposed to touching a keyboard. That is, the brain is registering differences in what kind of writing you’re doing, whether it’s by hand and doing something on a keyboard. 

And then there’s a fascinating study that was done—actually a couple of studies done in Norway—with older adolescents and with young adults doing actual imaging of people who were writing by hand. They’d happen to be cursive—of course, Norwegians do a lot of cursive writing—as opposed to using a keyboard. And it’s really clear that there were different kinds of brain connectivity happening if you’re using your hand as opposed to using a keyboard. And one of the authors of the study said in an interview was this: “A lot of senses are activated by pressing the pen on the paper, seeing the letters you’re writing and hearing the sound you make while writing.” And I mention those because in the research that I did with university-age students, both in the United States and in Europe, some of them said exactly the same sorts of things about what they perceived as being important about writing by hand. So you guessed that some of my favorite questions are, “What do you like most, what do you like least about…” In this case, it was, “What do you like most, what do you like least about writing by hand as opposed to writing with a keyboard?” 

So, some of the responses are what I like most about writing by hand related to cognitive issues. “The words seem to have more meaning,” or, “It adds to a high level of creativity.” I hinted at this earlier. “It leaves tracks in my mind.” I could chisel that in stone, it is so precious. Or, “I can see what I’m thinking.” But people also talked about embodiment of saying, being able to physically form the letters made writing be more personal to them. It made it theirs as opposed to somebody else’s. 

What did they not like about writing by hand? It takes longer and, for particularly a lot of Americans, but it’s true of people in other countries as well, they don’t do that much handwriting. If you’re writing an awful lot—I remember my college days when we used blue books and we wrote for three hours, your hand hurts, and yes, I still have the little bump on my finger from that pen pressing into it. 

Okay, what did they like most about digitally writing? They had nothing to say about cognition. There was nothing cognitive they liked about—that the 200 or more people expressed. But they said, you know, “It’s great having internet access while you’re writing.” “You can check your spelling.” “You can do corrections.” “The product comes out looking neater.” That’s good. 

What did they not like about writing digitally? Cognitive issues—less retention, less focus compared to handwriting. Or one of my favorites, “It’s aloof.” And then there was the embodiment issue of, “I feel detached from what I’m writing, since it looks just like every other piece of text that people write online.” That is, when people write by hand, there’s a real feeling of, “It’s mine, it’s personal.” And I will stress that personalization issue because it’s the same thing we’re going to see when we say, “Okay ChatGPT, you write this for me, I’m sort of out of time,” or, “I’ll just do a little editing afterwards.” You interview people, or you hear them give comments and surveys, and they will tell you, “The personal part is gone, it’s not mine anymore. I don’t feel ownership.” So there’s a connect.

[Kris Perry]: Yeah. I love these real examples. You also mentioned earlier the prevalent perception that reading content digitally is more engaging and convenient. So I’m thinking about parents and wondering whether they have the same perception and that there’s something that they could be doing to nurture reading in young readers. Becoming aware of the difference, what can they do to help emerging readers?

[Dr. Naomi Baron]: There are two conflicting things that we hear from both parents and educators of young children. The first is it’s really important to read print books with kids. “Laps, not apps,” as it’s often said. And we really feel we need to do some of that. And it doesn’t matter if you’re a computer scientist, if you’re an AI person, if you spend your whole life on a digital screen for your job, there’s this belief it’s important.

But at the same time, there’s a concern—there’s a fear, if you will—that if we’re not introducing digital stuff to kids early on, we’re somehow depriving them. And that continues into lower school and to middle school and to high school. And it’s a far more complex story. 

So, what is it that we know from the research? The first thing we know is that an awful lot of the well-meaning, well-intended books designed particularly for very young kids before they’re starting to read at all, before they’re entering even kindergarten—a lot of those books have bells and whistles and sounds and colors and videos that detract from the story. And what becomes very clear is the kids get understandably entranced by all that stuff that’s going on, but the purpose of the book was often to follow a story, to get engaged with a character, to care about a character. There are some people who are now working on designing books that are much smarter for young kids, but that’s a real effort and you need to get publishers to support it and there’s not a lot of support for things that aren’t flashy. So, that’s a challenge that we have for a number of books for the younger kids. 

Sometimes, that notion of being more attractive if you’re digital can work in our favor. So there’s studies that have been done, particularly in the UK, but elsewhere I’m sure as well, of particularly young boys – so we’re talking about eight, nine, ten-year-olds – who really want to have nothing to do with reading. You know, “Boring, girls do that.” Right? Or, you know, “Why should I care?” And there have been experiments forming book clubs with e-books with kids who didn’t want to read, and son of a gun, you can often get them to get interested in reading and then to get interested in print. The same thing is happening with audiobooks. Audiobooks are not necessarily a turn-off from reading print. Sometimes they’re an alternative avenue to get people, kids I’m talking about in particular, to want to read print. So what becomes important is understanding this variation in where our children or our students are in terms of what’s intriguing to them, what’s not intriguing. And the goal is to get them interested, to get them hooked in the best sense of the term. And sometimes finding an alternative avenue of doing that using technology can turn out to be quite helpful. 

Interestingly, when I’ve surveyed students and said, you know, “What do you like most about the convenience factors of reading digitally or reading in print,” you’ll have two students who are the same age—I’m thinking of middle schoolers and high schoolers now—who will say the exact opposite. So one will say, you know, “What I really like most about reading digitally is you don’t have to carry books around.” And then you’ll have someone else who might be sitting next to them in class, says, “What I really like about print is it’s more portable.” Or, “Digitally, ease of access, all you have to do is open the device and read.” And then another kid says, “What I like about print is its ease of access. You can start reading so quickly.” 

So, we’ve got to be really mindful of these differences between people in terms of what their predilections are. We know from the so-called PISA exams for 15-year-olds done internationally by OECD that the people who are the biggest readers by choice tend to read in both modalities, in print and digitally. And I’ll bet they listen to audiobooks these days, as well, too. So it’s not as if we want to say no to one format, it’s rather what works for that person at this point in his or her life.

[Kris Perry]: Yeah, the fact that we’re using all of these different modalities is really interesting. But the biggest elephant in the room these days when it comes to reading and writing skill development is generative AI, large language models. And I want to talk to you about how these models like ChatGPT and Claude are redefining our reading, writing, and even thinking skills. 

In your research, you also talk about AI as “reader.” Would you explain what it means to say that AI reads, and how does that differ from what humans do when we read?

[Naomi Baron]: A little bit of context. Ever since ChatGPT came out at the end of November 2022, people have been so excited about AI writing things. You know, are students going to have their essays written for them? And then we realized, it can do computer coding. And by the way, it can also draw pictures if you tweak it and so forth. What people didn’t talk about is the fact that if you want to have something written, you have to have something that is first “read.” Now “read” is in scare quotes because AI doesn’t read the same way we do, but it has to have access to, not just a text, but a gazillion texts, as it were. And then it pulls out from what it has “read,” and I’ll put that in scare quotes, a written response. 

So what humans do when they read is they may just read. In fact, historically, many people learn to read and never learn to write. So you could be a reader and never be a writer, okay? Or humans might read something and then hopefully think about it. Or humans might read, think perhaps, and then write. There’s a—there was a semiotician by the name of Robert Scholes, who also was a teacher of English and literature and so forth, who once wrote, “The only way we know our students have done the reading is by the writing they do.” And in much the same way, the way that we know that AI has “read” is by the writing that it does.

[Kris Perry]: Incredible. I love that connection. You know, at the Institute, we hear from educators and parents—and frankly, even our own children—about how common it is for adolescents at school, as well as even adults at work, to use tools like ChatGPT for writing assignments or writing products. What is the research telling us about what happens cognitively when we use ChatGPT as a writer?

[Naomi Baron]: We’re just beginning to get answers to that question. So, I’ll give you three versions of answers. One version is when you just interview students and you say, “So what do you think is happening when you’re using Chat GPT?” The New York Times had, back in May 2025, surveyed a group of high school students just asking, you know, “What do you think about using, particularly, Chat GPT, but it could be any other large language model?” Here’s some of what they said.

“Use of AI removes the students’ need,” though I’m speaking very formally,  “removes the students’ need, which is to think, which is very dangerous. If students grow up without using their brains, then they definitely won’t be able to use them later.” Or another one said, “If I do use AI for ideas, it doesn’t give me the chance to try and come up with my own things.” So again, I’ll come back to the point I made earlier. If you bother to ask students, “How do you feel about whatever the technology is?” They know a lot more. They’ve thought a lot more about it than perhaps we gave them credit for. 

Okay, similarly, you ask these kinds of questions of college students and you look at studies that have been done and they worry about losing their thinking skills, their analysis skills, their writing skills. But then there are some studies that have been done that were based on surveys of a type where you give a bunch of questions to people and ask, “Did you do a lot of thinking in order to produce what you were supposed to produce as a result of this exercise?” So what do I mean? One of those studies was done largely by a group from Microsoft Research, six out of the seven researchers. So Microsoft, obviously, is thinking about the issue. What they had done is survey a number of so-called “knowledge workers,” that is, people who in their jobs regularly had to use an AI tool. And they asked the people in the survey, “When you’re using the tool, how much, basically, of your own brain power are you using?” And what they reported was precious little—that their use of their critical thinking skills, because that was the term that was used, went way down when they were using one of these tools. And what the authors of the study said in their conclusion is without regular practice cognitive abilities can deteriorate over time.  There was another study similar to this just done in Switzerland. 

But I want to come to a third type of study, because it’s really interesting because it used imaging of the brain. This is a study that just came out in June 2025. And what it did is it had people of various ages write an essay using ChatGPT, or using search—let’s say old-fashioned Google search, or using nothing other than their own brain. But while the subject participants were doing the writing of this essay, they were wearing this interesting kind of cap, which is using electroencephalography. So you have these electrodes that are stuck in different parts of your head, and it can map which parts of your brain are active at different times when you’re doing this writing. Lo and behold, we now have tangible evidence that the students who were writing using their own brains had more electrical activity in their brains at the time of writing, than did the people who used a search tool, than did the people—namely, at the bottom of the totem pole—who were using AI. Also, when the people were interviewed about, “So, how did you feel about the writing,” people who were doing their own writing without any of these tools felt more ownership over what they had written than people who were using search. And the least sense of ownership was when AI was doing the writing. So again, students are perceiving that there’s a difference in the way you write—plus, we now have the pictures to prove it.

[Kris Perry]: You know, this is maybe a little bit off course, but one thing that keeps coming to mind in addition to this connection between reading and writing is just free-associating verbally or responding verbally to a cue in a classroom, for example, or in the workplace at a meeting where people are brainstorming or they’re drawing upon previous experience or they’re trying to solve a hard problem. And people then are drawing upon their experience—what they’ve read, what they’ve heard, what they’ve seen. And I imagine—well, let me not get ahead of you. Has that been studied? Is there also a third dimension or domain of learning literally versus, you know, having it machine learning, you know, sort of embodied?

[Naomi Baron]: In order to write down something that you’re brainstorming, you first have to be able to think about it. And there’s no reason you couldn’t verbalize it as opposed to write it down. So, you’re right—whether it’s in the workplace or whether you’re sitting and brainstorming together in a classroom or in a team of people preparing a project for a subsequent class—if you’re not able to have things come out of your own head, but need to rely on what you’re finding electronically, you’re in trouble. 

But there are more things that are also in trouble—and we can get to these later, perhaps—and that is the kind of communication that we have with each other, particularly written communication that is driven by AI—so, an email that AI writes as opposed to you writing it yourself. So you take what happens when I start writing an email with Microsoft Outlook. “I hope,” and then it fills in, “you are well.” Well, I might be hoping something else. I hope the heat in Washington would stop, or I hope the dog would stop barking in the background or whatever it is. But it predicts, because it’s using generative AI, what I’m likely to say next. And then there were earlier tools. There was Gmail’s Smart Reply and Smart Compose that predicted, on the basis of what you have written before, what you’re likely to have written, wanted to write yourself if you bothered to do your own writing. 

Interestingly, there are studies that show that if you ask a person—so, I’ve written you an email. And now a little birdy tells you that I didn’t actually write the email. That email was generated by AI. And then I ask you, “What do you think of me?” And if indeed the recipient believes that the sender didn’t actually write the email himself or herself, the recipient tends to think that the sender is arrogant and self-centered and not interested in you, the recipient, which may or may not be true. 

So there are all kinds of other things about interaction, some of which is brainstorming face-to-face, some is sending an innocent text or email, because there’s predictive texting, right? I mean, predictive texting is no different from what’s happening when you write in an email and—I mean, it’s slightly different technologically, but otherwise it’s the same principle of you’re not writing it yourself. You’re not caring enough, to put it bluntly, about the other person to say, “What do I write that’s personal for this person I care about?”

[Kris Perry]: Are there particular concerns for children’s cognitive development versus adults when using generative AI for reading and writing tasks that were previously human only? And when you think about those concerns, of those, which are you most concerned about?

[Naomi Baron]: Okay. In some ways it’s too soon to give what I’d love to be able to give the audience for this podcast an answer of, you know, “Preschool children do this and then for lower school…” We don’t know enough now. We don’t know enough about where these tools are going to go. And the other thing that happens is these tools are so buried in what seems like plain, vanilla working with any digital tool that you don’t always know that AI is back there. 

Okay. So, what is it that I would give as a message for everybody? Children in particular, but adults don’t get let off either, and that is to establish a skill-based base, something you can come back to. Even if at some point you’re using a lot of AI, when you need to use your own brain, that you’ll be able to do so. I used to say to my students, these were university students, “What do you know when the internet is down? What do you know?” Because we used to get a lot of blackouts in Washington and people couldn’t look things up. But now I ask, “What can you read, what can you write if ChatGPT is down?” Or Claude, or name your favorite—or Llama, or name your favorite large language model. So the whole issue becomes do you have a foundation that you can call upon? And that’s where building that foundation early on, but then continuing to reinforce it matters. It’s one thing to say, “Well, we’re going to really work with lower school to say there’s going to be no AI and then middle school everybody does AI.” Well, whatever you try to inculcate early on is going to fade the same way our foreign language skills tend to fade. And the other thing is, there’s certain kinds of analytical skills you don’t develop until you get older, until you get to be a teenager, until you get to be a college-aged student and then, heaven help us, even older. So I don’t, at this point, have any time recommendations in terms of by what age, by what grade, to do what. But building and then practicing that foundation is incredibly important. 

But I also want to add that it’s not just cognitive skills we’re trying to build. We’re trying to build the notion that when you read for yourself, when you write for yourself, you’re developing who you are. You’re developing your own voice. You’re expressing what you want to express as opposed to what some algorithm thought would be a great collocation. And I think, however important it is to talk about cognitive components, we shouldn’t leave out the others.

[Kris Perry]: Yeah. You know, we are sort of as a culture obsessed with physical fitness, but I think what you’re describing here is mental fitness—making sure you build skills for yourself that you keep, that belong to you, that are a part of you, that are uniquely you. 

[Naomi Baron]: Right.

[Kris Perry]: And that you do that through reading and writing and taking information in and digesting it, metabolizing it, and then using it again when you want to. And I really love that you pointed out that so many of the AI tools that have been developed are already embedded, and not always clearly so, in products. Even in the voice controlled-products, like Alexa and Siri, that children might be using during the day at home. So, we know it’s surrounding us. So I love taking a moment and thinking about building that skillset for yourself that you can take with you that’s not tied to a computer. What are the other ways in which you think GenAI can be positively, and hopefully effectively, used in classrooms for learning?

[Naomi Baron]: We’re just figuring out, meaning thousands and across the world, hundreds of thousands of teachers experimenting—seeing what works, what doesn’t work—as to how to best use these tools within a classroom setting. I’ll give just a couple of examples of places that I would guide listeners and viewers to turn. One is things going on at Stanford University. There’s something called the “Stanford AI Plus Education Summit” for K-12 educators and they have all kinds of good information there. There’s also a project at Stanford called “Stanford CRAFT.” That’s capital C, capital R-A-F-T. That is a program for high school students which has teaching tools designed for high school students, and I’m sure lots of schools are developing their own tool kits. But I want to talk conceptually first before I talk about specific examples. The Khan Academy is also doing lots of interesting experiments using its tool, Khanmigo. And finally, there are some good guides coming out as to how to effectively use these tools, and when not to overuse them designed particularly for college students—but they will apply, I’m sure, equally well to high school students. The American Association of Colleges and Universities, in partnership with Elon University, has produced a really helpful student guide to artificial intelligence. And some of that is learning how AI works and what it does and what it doesn’t do. And some of it is giving suggestions on, “So how would you want to use it?” 

All right, now let’s get to a couple of experiments, and I’ll just give my own, because I had a lot of fun with them. And I’m not a K-12 teacher, but I am a person who experiments with AI. So one of the things you can do is put in alternative versions of a prompt and see what kinds of answers you get. Or give the same prompt to multiple large language models and see what kinds of answers you get. And you might have an agenda for what you’re trying to accomplish. So if, for example, you want to teach different levels of vocabulary—let’s say you want to teach really sophisticated vocabulary to a group that hasn’t been challenged enough in learning new vocabulary words now that the SAT doesn’t give these big vocabulary words anymore. You also might want to see what kind of humor you can get out of a machine. So here was my experiment. I took Claude, one of the large language models, and I said, “Claude, here’s your assignment: Please write a 200-word summary of Shakespeare’s King Lear and do it with really sophisticated vocabulary.” Okay, here’s what it wrote, the beginning of what it wrote: “King Lear, a monarch of considerable sagacity, yet afflicted by hubris, orchestrates a fateful partition of his dominion among his three daughters, Goneril, Regan, and Cordelia. The allocation is to be predicated upon the respective declarations of filial devotion.” You want to teach vocabulary? This could be fun. 

But now I asked, “Well, what about for a 12-year-old? How about a 200-word summary?” Beginning of it: “King Lear, an aging king, decides to split his kingdom among three daughters, Goneril, Regan, and Cordelia. He asks each of them to tell them how much they love him.” And I’ll mention that one of the tools that an awful lot of people are taking advantage of—teachers are taking advantage of—is taking very complex texts and making them more accessible to young kids. Sometimes there’s, you know, the children’s version of The Odyssey and sometimes there isn’t, but now you can create your own. So this can be a very helpful tool in the classroom. 

But then I also said, “Well, what if I wanted something humorous? Let’s see what AI could do.” So I asked for a 200-word summary of King Lear that was humorous. Here’s the beginning: “Picture this. A cranky old king decides to play a game of who-loves-Daddy-most with his three daughters. The prize? Just a casual third of the kingdom.” 

So, you can have fun with this, and it can be educational at the same time. This is not a question of, “Oh, are you going to use this for cheating?” It’s not a question of, “Oh, I have to write an essay, and then I’ll get AI to edit it,” which is commonly what people do, or, “AI will write the essay and then we’ll critique it together.” And that can be helpful, but you could also do things that are a lot of fun, pushing the abilities of AI to engage with you.

[Kris Perry]: That was a really fun example. Are there any AI practices you think that we should avoid, or that might interfere with cognitive development, or even a sense of ownership of one’s perspective? In other words, taking it in the other direction. Is there anything you think is really worrisome at this point?

[Naomi Baron]: Well, the first thing that’s worrisome—and I have absolute empathy for teachers K through graduate school—is how do you talk about what’s okay to use and what’s not okay to use with AI? And we end up often avoiding the conversation. There are all these studies that are being—surveys that are done of students now who say, “Please, tell us what you want. Tell us what’s okay. Tell us what’s not okay.” And I don’t blame the teachers for not having the answers, because what might be okay today might not be tomorrow. Or, you know what, AI is in spell check. Is spell check okay to use? AI is in Grammarly, which is one of the most commonly used grammar check—it should be “grammar teaching,” but often it’s “grammar check” tools. It’s totally AI-driven. Is that okay? Is that not okay? It’s hard to know how to approach this. But even to have the honest conversations about it, I think, is a very important first step. Even if you as teachers say, “I don’t have the answers.” Every university I know of now has a committee trying to figure out, “What should our policies be for AI? Should it be university-wide or college-wide? Should it be up to the individual teacher?” We don’t have the answers, but we need to be honest with ourselves that we don’t have the answers. 

Okay. But in terms of what to do in the classroom to help boost cognitive development and sometimes social development, it’s important to say, “In our classroom, we’re sometimes going to actually write by hand.” And there’s a movement in a number of schools and countries to say, “We’re bringing back handwriting.” And it’s not just because we don’t want kids to cheat. It’s because we want them to experience cognitively what writing by hand can do.

Okay. Similarly, reading in class. There are classes now that begin by reading aloud the assignment because we know that students are—this has gone on long before gen AI came along, but it’s only going to get worse. Students are not doing the assignments that you’re giving them to read. So, what we need to do is some reading together.

Similarly, annotate either by hand and turn in your annotations so it’s clear you read it, or there’s another tool—and it’s digital but it has a social component and it has a motivate-to-read component—and that’s to do digital annotation with tools like Perusall, which was a digital annotation tool created at Harvard University, largely by the physicist Eric Mazur, who found his students were getting A’s on their exams, and then he would see them three weeks later, and they didn’t remember any physics. He said, “Something is wrong with this picture.” They were Harvard students, right? And there’s another tool, Hypothesis, and a lot of high school and college teachers already know about these tools. But what happens is, you annotate online, and then other students in the class who are also needing to do the reading, because those annotations are showing up with your name on it. And then you can comment, having read, comment on, having read what the other person has written, and you get discussion going on. And it happens to be digital discussion, but it’s discussion, and reading that was necessary to have the discussion, which is a whole lot better than what we’ve had in the past.

[Kris Perry]: If you could wave a magic wand…

[Naomi Baron]: Yeah?

[Kris Perry]: …what do you most wish people really understood better about generative AI and its implications for human skill development?

[Naomi Baron]: May I do two waves rather than one?

[Kris Perry]: Yes, you may do two.

[Naomi Baron]: Okay, thank you, but only two. The first is that we need to remember we created this technology. We did it, and we get to control—if we choose to—how we use the technology. The problem right now is that big tech is driving us to want to use, and to feel we need to use the technology more and more. So there are many large businesses that are saying to their employees, “You must use AI, you must use these tools.” In fact, Amazon has recently announced, the CEO of Amazon announced to its employees that, “You’re going to be using AI tools because if you’re not, you’re not going to have a job soon.” And that was pretty blunt. Lots of companies are saying the same. What schools are saying over and again is, “OMG, if we don’t get our students to know how to use these tools and they don’t get practice using them, they won’t get jobs. Therefore, we need to use the tools more and more. We need to train students on what these tools are, what they’re good for, what they’re not so good for.” 

But my second waving of the wand is to say it’s up to us to balance, to say, “What is it that we know from not just the research, but our own personal experience, our own personal brains? What do we know that relying on our brains does for us—does for us intellectually, does for us socially, does for us as building our own persona of who we are, what we think, what we believe, how we evaluate what other people think in a simple democratic sense of knowing to whom you should vote for?” You should vote. 

So, I think there are things we can do. There are countries—the United States is not one of them—that are taking seriously the question of how much we want to use these technologies. I’ll just give one simple example. About a week or so after ChatGPT came out—that is, in early December 2022 after ChatGPT had just come out—there was a group of teachers in Norway, teachers of Norwegian language and literature, this kind of union or organization of them, who made a plea to the Norwegian government: “Don’t let our students be using ChatGPT to do the writing and thinking for them, because our responsibility as teachers is to teach them how to think, to teach them the process.” And they made a big point that it is the process of writing, not the product, that matters most. It’s the process of rewriting, rethinking, getting critiqued, and going at it again. That is what education and an educated person becomes.

[Kris Perry]: Unfortunately, we’re getting short on time, even though it feels like we just started. A huge thank you to Dr. Naomi Baron for her remarkable expertise on reading, writing, and the fast moving frontier of generative AI. Her insights challenge us to think critically about how we use and don’t use technology as we support children’s learning and development. 

If you found this conversation sparked new questions or ideas, don’t stop here. Subscribe to Screen Deep on your favorite podcast platform, leave us a review, and share this episode with educators, parents, and anyone curious about the future of literacy in today’s digital age. You’ll find more about Dr. Baron’s work and related research in our episode notes at www.childrenandscreens.org. I’m Kris Perry. Thanks for listening, and we’ll see you next time.