Teaching in the AI Era: Understanding Global Trends and Developing Critical Thinkers

December 3, 2025

In partnership with the Council on Foreign Relations and the International Studies Association

Artificial intelligence has the potential to rapidly reshape the global landscape, from transforming economies and labor markets to influencing international relations, governance, and the flow of information. For faculty in higher education, these changes raise urgent questions: How can educators prepare students to understand and engage with AI’s global impact?

This webinar, cohosted by the Council on Foreign Relations and the International Studies Association, examines the intersection of AI, foreign policy, and democratic institutions, and explores how faculty can support students in understanding and navigating these complex dynamics. The conversation offers practical strategies for integrating these topics into the classroom to help students think critically, act ethically, and become informed global citizens in a time of disruption.

Speakers

Kat Duffy, Senior Fellow for Digital and Cyberspace Policy, Council on Foreign Relations

Dessie P. Zagorcheva, Adjunct Associate Professor, LaGuardia Community College, City University of New York; CFR Education Ambassador and ISA Member

Presider

Caroline Netchvolodoff, Vice President, Education, Council on Foreign Relations

Background Resources

Transcript

FASKIANOS: Thank you. Welcome to today’s Higher Education Webinar, which is cosponsored by the Council on Foreign Relations and the International Studies Association. We’re very excited to be partnering with ISA on today’s presentation. I’m Irina Faskianos, vice president of the National Program and Outreach here at CFR.

Today’s discussion is on the record, and the video and transcript will be available on education.CFR.org if you’d like to share the materials with your colleagues or classmates. As always, CFR takes no institutional positions on matters of policy.

We’re delighted to have Kat Duffy and Dessie Zagorcheva with us to discuss teaching in the AI era. And my colleague, Caroline Netchvolodoff, will moderate the discussion.

Kat Duffy is a senior fellow for digital and cyberspace policy at the Council on Foreign Relations. She has over two decades of experience spanning the U.S. State Department, United Nations, civil society, and the private sector, and previously directed the Task Force for a Trustworthy Future Web at the Atlantic Council’s Digital Forensic Research Lab, where she served as a research senior fellow. She has advised governments, companies, and NGOs on aligning emerging technologies with democratic norms and human rights.

Dessie Zagorcheva is an adjunct associate professor at LaGuardia Community College, a CFR Education ambassador, and a member of the International Studies Association. She leads a CUNY-wide initiative on digital disinformation and media literacy. And she has published work in International Security, the Journal of Slavic Military Relations (sic; Studies), the National Interest and the International Lawyer, among others.

And our moderator for this important discussion is my colleague, Caroline Netchvolodoff, who is the vice president of Education at CFR. She joined CFR in 2010 as senior adviser to the president for strategic planning. And in her current position, as vice president of Education, she oversees the development and marketing of CFR’s multimedia learning materials on the fundamentals of international relations and U.S. foreign policy. And I hope all of you are using them in your classroom because they are fantastic materials.

So Kat, Dessie, and Cary, thank you very much for taking the time to speak with us today. I’m going to turn it over to Cary for the discussion. And then we’re going to open it up to all of you for your questions, comments, and to share best practices. So, Cary, over to you.

NETCHVOLODOFF: Great. Thank you so much, Irina. And welcome, everyone. CFR and ISA are excited about today’s conversation. And, of course, we’re delighted to have you with us. As Irina said, I’m Caroline Netchvolodoff, the vice president of Education at the Council, where I lead the educational work, which you’ll hear about as this discussion unfolds. I see that we have, predictably, a lot of educators joining us. And I’d love a show of hands for how many of you are already dealing with the global effects of AI in your classrooms. I suspect nearly every hand is raised.

Well, there’s no denying that, for better or for worse, we have entered a new era. The release of ChatGPT in 2022 marked a clear inflection point in the history of higher education. And the advent of LLMs, large language models, and related technologies like agentic AI, AI tutors, and automated grading tools has upended higher ed, and raised fundamental questions about important things like originality, expertise, cognition, and even the nature and purpose of education itself. So we’re all aware that students are arriving on campus each year knowing how to use AI, but it’s clear that a very small number—at least, what we hear from our ambassadors and our network of educators—is that a very small number arrive knowing much about AI. And the fact that artificial intelligence is reshaping our world and the global flow of information raises some urgent questions.

So today we’re going to take on a big one, which is: How can educators prepare students to engage with AI and understand its global impact? We’re lucky to have with us the two people that Irina introduced, Dessie and Kat. Thank you so much for being here. Kat, I’d love to start with you. Help us set the scene for this unique moment we’re living through. What global trends in AI are most urgent for educators and students to understand?

DUFFY: Thanks, Caroline. And it’s such a pleasure to be with all of you today. I’m really excited for this conversation. I would start by saying it’s imperative to separate out AI as an educational tool and how that is impacting education from AI as a subject matter that is influencing societies, is working as an emerging technology, is impacting geopolitics. And so I’m going to start off really talking more about the geopolitical trends, and then I think later in the conversation we’ll probably delve a little bit more into the impacts of AI as a tool.

So geopolitically, I will say, you know, I started—my first paid tech policy gig was Y2K. So I’m officially at a quarter-century of working on tech policy. And I have seen every hype cycle known to man or woman. I’ve never seen any technology have the impact politically, capture the imagination and the panic of governments and policymakers around the world, the way that I have seen generative AI tools do. I have a number of theories for that, but the long and short of it is that there is no way now to separate the concept of artificial intelligence from how we think about geopolitical priorities and geopolitical strategies. It is consistently among the top three or four issues being raised in any foreign affairs room. You will hear about artificial intelligence. You’ll often hear about tensions between the U.S. and China. And you will hear about climate. And those three things sort of consistently surface. But artificial intelligence is right there with them.

We are also seeing that the way that different governments around the world, at least, are approaching artificial intelligence looks and feels very different from the way that we saw other technologies evolve. So from sort of the 2000s kind of into 2020, what you really saw was increasing internet connectivity. And then you saw the rise of global digital platforms, many of them American. But those digital platforms hit countries and hit societies that weren’t prepared for them, hadn’t necessarily been built for them. And governments didn’t exactly know what they were getting. And so you have situations where, for example, I used to live in Tunisia. You could really argue that Meta is Tunisia’s information technology infrastructure even more than the telecom companies are in Tunisia, because of how widely it’s used.

And so what you’re seeing now with artificial intelligence, the number of governments that are responding to this, I feel are responding from two different lenses. The first is that they do not want to be taken by surprise again. That can be for good or for bad when you think about things like prioritizing fundamental human rights. But regardless, governments want to be more ahead of the curve and want to understand where the technology is going, and how it could operate within their borders. They also really want to make sure they don’t miss out on the benefits of it. And this is where we’re seeing a big distinction between the highest income countries and the countries that are producing the most AI, and the lower and middle emerging powers.

The highest income countries that have more sort of control over AI’s development are really focused on what governance and risk mitigation would look like. And there have been a lot of different swings in that and various evolutions, where we’re now seeing a move away from a kind of U.S.-China-EU triad, into a broader range of actors, where India is also driving, Singapore, South Korea, ASEAN, Brazil, the African Union. So we’re seeing more governments enter into the picture. But we’re also seeing, in places like in regions where, for example, the African Union is engaging, the focus on governance is much more about equitable access and about the ability to use the technology, and access the technology, and not be left behind, much more than it is about mitigating its risks. And so you’re really seeing very different approaches to the technology based on the local contexts and the local concerns.

I’ll close by saying that in the United States, because we don’t have a lot of federal guidance or regulation on AI, and there are real legal questions in the United States about what would and would not be constitutional in terms of what the federal government can control, you are seeing the same type of fragmentation that we’re seeing at the global level. That same fragmentation is occurring inside the United States. And so we have, I think at this point, well over a thousand different bills that have been proposed across all U.S. stateand territory legislatures that involve AI in some form or fashion. And more than thirty states, including Puerto Rico, have now issued their own official AI guidance for K-12 schools. Forty-seven states have reported AI as a leading ed tech priority in 2025, surpassing cybersecurity. But only 6 percent of states—and I can provide sources for all of these things. But as of last count, only about 6 percent of states have sustainable funding plans to support any of this. And at least twenty different states in the U.S. introduced AI education bills just in 2025.

So if you are feeling confused, overwhelmed, like you can’t get your arms around it, that is just plain common sense. That is not you not knowing what’s going on. That is just a rational, logical response to a really hectic, fast-moving emerging technology. I’ll stop there.

NETCHVOLODOFF: Wow. I mean, incredibly, first of all, sobering, fascinating, and helpful context. Let’s build on that and talk about the role that educators play in this transformation. Dessie, I’ll throw this next question out to you first. What knowledge, skills, and perspective can educators help students develop to navigate a connected and AI-driven world?

ZAGORCHEVA: Thank you. Let’s start with placing AI literacy within global literacy. As educators, especially in political science, international relations, we have always tried to develop global literacy skills, and knowledge, and perspective. But in this case we should also emphasize AI literacy. And I don’t think that academia is doing enough. I think one of the reasons that educators are feeling so overwhelmed is precisely because there have not been sufficient resources and time spent on AI literacy, both for educators and for students. And, again, this complicates our task. We have a generation of young people that is not trusting any source—almost any source of knowledge, traditional media, not institutions. But they’re trusting chatbots. So I think one of the first thing that we should start doing is let them know a little more about the chatbots, because they’re overly trusting them—sometimes naively trusting them.

And part of the problems come from that they—the overreliance on these chatbots comes precisely from this. So in terms of AI literacy, maybe we should start with things how AI chatbots hallucinate. Just telling them that they make mistakes. They come up with all kinds of sources, or legal cases, or evidence that just does not exist. So they have to be careful about that. Talking to them about AI biases. Sometimes they trust AI chatbots because they believe that AI is objective. The chatbots are giving them objective information, unlike the rest of us humans who have political, ideological, and all kinds of other preferences. So talking about AI and how AI knows things. Obviously, AI is trained on vast data of internet. And it has the same cultural, gender, racial biases that we may have. So alerting students to some of these issues will actually help them.

In addition to that, we could talk about the privacy and the security risks that the technology poses, and also issues of academic integrity. These, again, we do not have—we have not had much time to spend discussing with our students. Some of the growing academic concerns have been about overreliance on AI and how this has affected critical thinking and communication skills of students. So thinking about what kind of skills, in addition to knowledge, we should develop is precisely critical thinking, media literacy or information literacy. In addition to that, communication, collaboration skills, students being able to work in a team, students being able to clearly communicate, persuasively describe their arguments. And also creativity and problem solving. In this current situation, obviously, being innovative will make a big difference in the future. So thinking about their future careers, creativity and problem solving should be central stage.

NETCHVOLODOFF: That’s great. I’ll sort of tag on to what Dessie has said and focus a little bit—AI literacy is just essential. And I think we all know that. And I look forward to seeing more courses. I saw in the chat that Nigeria is offering AI literacy across all universities. I mean, let’s get more of that happening in this country. But, you know, AI has the potential to transform research, teaching, administration, admissions, publishing, and so much more. But even in a world in which AI becomes ever-more powerful and widespread, I think we all know that basic skills—many of which Dessie has mentioned—like clear thinking and strong writing, will remain essential.

So I’m going to take a little step backwards and say that at the Council on Foreign Relations, we define global affairs literacy as possessing the knowledge, the skills, and the perspective to effectively engage as informed citizens while navigating our connected world. And you can go to the CFR Education website to read more about this, where we go into more of these, the knowledge, the skills, and perspective. But, in a nutshell, basically by “knowledge” we mean a foundational understanding of the issues, the forces, and the actors that shape today’s world. The skills, Dessie’s done an incredible job of listing them, but I’ll just reinforce a couple. The skills refer to an ability to comprehend, communicate, and address complex international challenges collaboratively. There are a lot of Cs. Sort of the four Cs of twenty-first century learning.

And perspective involves—and this is a really important one, I think we can all agree. Perspective involves a disposition to understand, and most important, respect that differing—that people arrive with differing viewpoints and priorities to any conversation, to any discussion. So when students step away from the internet and from AI into a professional space or a public square, they learn about the world. They learn about other people. And they learn about themselves. And so they practice—when they’re not using the internet and AI—they practice contextual agility, I guess is the best way to put it, critical thinking, empathy, social skills, all of which ultimately strengthens their agency.

Which gets to my sort of final point, which is that I think it’s really essential that—an essential goal for educators is to help students master AI, while simultaneously strengthening their agency. So rather than courses that simply allow for the use of AI, and the ethical use of it, and how do we avoid cheating and plagiarism, courses are needed in which students read, talk, and write about AI. And I was recently reading something, I think it was in the Chronicle of Higher Ed, Jane Rosenzweig, who is the director of the Harvard College Writing Center, teaches a course that helps students think critically about AI itself, so that they’re—and this is, I think, really super important—so they’re able to recognize when using AI solves a problem versus when it creates a new one. So my hope, I think—and I think it’s beginning to happen—but my hope is that there will be a proliferation of such courses across campuses in the near future.

Kat, anything to add to that?

DUFFY: No. I mean, I think there are—I guess on AI literacy, maybe I would say I think that there—as someone who uses AI, multiple AI systems, day in and day out, and has for years, I think there are some approaches to AI literacy from an educational perspective that might be counterintuitive but worth considering. So the first is that AI is actually a very good teacher on AI. (Laughter.) So you can really work with students, or just with yourself, on, you know, whichever sort of model you choose. So you might be working with Anthropic’s Claude, or with OpenAI’s ChatGPT, or with Google’s Gemini, or with Meta’s Llama, or with Perplexity, which taps into all of those systems, or an open-weighted model. Whatever you’re using, you can actually query the AI to say: I’m trying to figure this out. What’s the best way to work with you on it? And the AI will actually come back and help you with what we call prompt injection, which are essentially queries, right?

So it feels counterintuitive, but part of how you develop AI literacy is by understanding that you are still talking to a computer, to a machine, and there are very specific ways that you would talk to a machine, in the same way that I talked to a toddler differently than I talked to the president of the Council on Foreign Relations. And I talk to a Spanish speaker from Cuba differently than I talk to a Spanish speaker from Spain, in many respects. The AI systems are sort of the same. I use Claude differently than I use GPT. Not everyone needs to get to that level of literacy, but the more that you engage with whatever tool it is that you’re using, the better you will understand where it’s strong, where it’s weak, where it could serve your purposes, where it might not, and how it can help you. So I would say that’s one thing people tend to underestimate, is how much the AI can be a teacher for you in terms of using it as a tool.

The second thing is that I think people tend to—and students in particular—are looking to AI to give them very fast answers to big, hard questions. And this is the concern that we all have in terms of its impact on critical thinking skills. And you can feel this. The more that you start to rely on the systems, those of us who grew up not with that, I—personally, I can feel my brain start to atrophy a little bit. It gets a little muddier. It gets a little messier, because really all I am ingesting is a lot of syntheses of a lot of other people’s thoughts. But that is pulling away from my ability to have my own thoughts, because it’s pulling back from the amount of time where I’m thinking critically myself. So I have struck—I really work to strike a balance.

But in terms of AI literacy, if what you’re trying to teach is critical thinking in terms of whether the systems are trustworthy, it’s actually better to have people start closest to home, closest to lived experience. Where do they have the most granular knowledge? That may be a local news event that occurred last week and that occurred two years ago, where all of your students know it relatively well. Or it might be that a student is, like, let’s say, a member of the Somali diaspora in Minnesota, and is really looking at news coverage or understanding of Somali diaspora issues through the different models, and then asking for that information with different sort of primary language sources.

But where you will find the most effectiveness is engaging with the AI on those subject matter areas where you have the strongest internal capacity to vet whether or not you buy the answers you’re getting. And if you really know the subject—it could be chocolate chip cookies. Like, it doesn’t matter. If you really know the subject matter well, you will start to get a sense of, like, OK, well, that’s right, but, no. Hard pass. Like, I don’t—I am not an unsalted butter person in my chocolate chip cookies. I like to do salted butter. I have strong feelings about that. The AIs will mostly tell me unsalted butter, because that’s the norm. So this is another thing in terms of AI literacy to consider, is really starting—allowing each student to start as small and granular and close to home as possible. And as an educator, you can do the same. And from there, you build the literacy in such a way that you have a stronger spidey-sense for whether or not you can trust what you’re getting back from the different systems.

NETCHVOLODOFF: That is really a great insight, I think invaluable. I just learned something very important for our work at CFR Education. Thank you, Kat.

I think this is a good time for us to move to a discussion of the practical, because I think that most of our audience is quite eager for some suggestions and guidance. So, Dessie, I’m going to throw this to you first. What practical strategies or classroom examples have you used, or seen others use, to foster critical thinking, something that we’ve really hit hard on here, project-based learning, and active engagement in this AI era?

ZAGORCHEVA: Thank you. Some of the practical strategies and the resources that I have used come from CFR Education resources. And we’re so lucky to have those resources. They’re really helpful. My students really get excited, especially about one of the resources that I will speak later on in more detail, the simulations. So, first of all, CFR has a lot of resources related to AI. And they’re very flexible. It will depend on you how much you would like to introduce AI in your classes, and in what way. There are resources that are a wide range of resources in terms of the effect of AI on foreign policy, on geopolitics, on economics, on national security. And we have resources that include expert analysis, articles, videos, podcasts, backgrounders, explainers.

And depending on how much educators want, how much time educators want to spend, they can use a lot of these resources. Some of the questions that they can use to integrate into their own teaching, depending on the subjects that they’re teaching. For example, there are resources on how AI affects governments and how governments adopt and adjust to AI, and what governments do with these new technologies. For example, if you’re teaching authoritarian governments, there are resources about how China uses AI for various purposes, how China uses AI for surveillance, for repression of its own citizens, how China exports surveillance technologies. So if you’re teaching about that you could use those resources.

In addition to that, we have resources about AI’s influence on conflict and how the military has integrated AI. There’s resources about the laws, the lethal automated weapon systems. There are resources about AI affecting climate change, how AI affects global competition, let’s say the relationships between China and the United States. So there’s this wide range of resources. And, again, they are nonpartisan expert resources that are extremely useful.

Some of the strategies that are important in terms of making things more lively in the classroom. I use AI simulations. In the case of simulations for AI purposes, the Council has a simulation on how to regulate AI. And it discusses the different ways of regulating AI, including in the United States, the European Union. But the simulation is basically on students advising the president what options to choose in terms of regulating AI. Should the United States choose the option to regulate on the national level? Should we try to have a global level regulation, let’s say, through the United Nations or other institutions? Or should there be no regulation whatsoever?

But before we get into the simulation itself, first of all why to use simulations? How are simulations helpful in our education? My students find simulations are some of the most exciting experience they’ve had. They’re very motivated. They’re very enthusiastic anytime I mentioned simulations. There’s no other way to motivate them to learn. So this is a very great resource. It has all kinds of benefits over lecturing and some other methods. In addition to this increased engagement, simulations are an immersive learning experience. In the case of the AI simulation and AI regulation, for example, our students are simulating the National Security Council of the United States.

Our students get into the shoes of decision-makers—let’s say the president, the vice president, the Cabinet secretaries. This allows them to hear different perspectives to try to simulate the decision-making process as it is in real life. And what does it mean to make decisions under pressure? What does it mean to make decisions with incomplete information? What does it mean to take numerous different arguments into account? And, again, this cannot be learned just through lecturing. It is so much better for them to get—to experience the process themselves.

In addition to that, simulations provide deeper understanding. My students are saying that thanks to the simulation they appreciate much better the complexities of the decision-making process. They also appreciate much better how complex the decisions on AI, in particular, is. Let’s say they would say things like, I had no idea that the European Union, the United States, and China have so much diverging views about AI, how to regulate AI, how to use AI. So a lot of this is very beneficial. And it is much easier to achieve through a simulation rather than through lecturing.

And a very important thing as well, just like before that we talked, in addition to the knowledge we also need skills. And simulations are one of the best ways to achieve the skills that we just discussed, namely critical thinking, communication skills, collaboration, creativity, problem solving, the ability to listen to arguments that you disagree with, and how to deal with such arguments without dismissing the person. So all of this is very important. And I think simulations are great resource in order to help our students adjust to the new environment in order to be able to also address different perspectives.

And in my classroom, it’s also very interesting because I always have students from various different parts of the world, international students. So they bring their own perspectives. They bring perspectives about what their country is doing, how the United States’ position is affecting their country, and all of these global political relations come into play. And again, I can go into more detail about that—

NETCHVOLODOFF: No. No, no, I mean, yeah, no, this is really wonderful. And, you know, I would say that the—you know, as much as the Council offers a vast array of resources, all of which are valuable in their own right, our simulation—our library of simulations, I think there are sixty now. And some are National Security Council simulations, others are UN Security Council simulations. But for all of the reasons that Dessie has cited, and then some, they are a really solid vehicle for building global affairs literacy because they tackle the three pillars—the knowledge, the skills, and the perspective. And perspective building, as I said earlier, I think is just incredibly important at this sort of moment in time. And so we are going to be churning out more simulations in the coming year. And happy to answer any questions about those.

But these are incredibly helpful strategies. I think we’re—I’m mindful of the time. So I think we need to turn to another central piece of this puzzle, which is helping students stay grounded and optimistic as the world is shifting around them. And so the final question is, you know, an observation and a question. AI can feel overwhelming. I know it is for most people. Not Kat. (Laughs.) Maybe not Dessie. But it is for me, and I think—

DUFFY: It’s still overwhelming for me.

NETCHVOLODOFF: (Laughs.) And for many of you on this call. But how can we help students in particular stay hopeful, engaged, and see themselves as agents of meaningful change? Kat, you want to take that? I have a couple of thoughts on it, but I’ll let you guys dig in first.

DUFFY: Sure. I mean, would say the first thing is agency, agency, agency, right? Just because there is an artificial intelligence involved doesn’t mean that their intelligence shouldn’t enter the equation. And so what I would really urge all educators to do is just embrace the reality that all of your students are using these tools. We call this shadow AI, but it’s a real problem because when use is limited to shadow AI, you can’t actually have good, honest, open discussions or agreements around how it should be used, the way to use it, the way to engage with your professor, the way to talk about it. You really—this is not a tool that you want flying under the radar where your students can’t be forthcoming with you about how they’re using it and the questions they have.

So I would say, especially—it’s a little different, I think, on the sort of K-8 side at least, but high school and above, just assume this is on everybody’s phones, it’s on everybody’s computers. They’re all playing around with it in different ways. There is no way of avoiding that. I wouldn’t try. And I think the more that the adults and the educators in a young person’s life right now try to avoid it, the more that they indicate to our young people that they don’t care about their future, and they don’t care about preparing them for that. They mostly care about protecting themselves from uncertainty and change. And I don’t think that’s the message any educator wants to give to any student.

NETCHVOLODOFF: Wholeheartedly agree.

ZAGORCHEVA: Yeah.

DUFFY: Agency is so key. And, again, this is where, you know, AI is only reflective of that—well, currently because we’re talking about language models. That’s the type of artificial intelligence we’re mostly talking about these days. That may change. But at the moment, the AI that people are relying on for generative AI tools is coming from language models. And language models come overwhelmingly from that information which was digitized. That which can be digitized is the knowledge that is reflected. And there is a whole world of knowledge and experience out there that hasn’t been digitized. So thinking about what the cultural influences are between different countries’ different approaches, or how different AI systems were built. Getting your students to think through things like their data privacy. Are they using the product or is the product using them? That’s always a really good question. And most people don’t like to be a sucker, right?

NETCHVOLODOFF: They do not. (Laughs.)

DUFFY: So this is—I would say, like, that’s one thing. And then the final thing I would say is where does this allow your students to shine? And where does it allow you to shine in ways you couldn’t necessarily before? So where you used to spend two or three hours building your decks for a curriculum or a presentation or a lecture, that’s the type of thing that can now truly be done probably in about a half an hour, max, once you sort of get the hang of it. And that gives you so much more time to put your brain where you really want it to go, and not, like, moving a box around a PowerPoint slide, right?

But for your students as well, if you have, for example, a student who’s a—like, who speaks English as a second language in the United States. If part of what you’re grading them on is their ability to read and write and speak in sort of professional English, that’s one thing. But if that is not the key skill set that you are teaching for them, then perhaps what they could do is provide a copy of what they’ve written on their own, and then also run it through an AI system that acted as an editor for them and that they worked with in terms of turning something into professional English, and allow them to turn in both. You can have a conversation about what they learned from that, what the differences were, but you may also find that you are more able to accurately analyze where that student is, how they’re thinking, and where their growth is, because you’re not dealing with linguistic barriers that are blocking your ability to see their thinking process.

And so I think there’s lots of creative and beautiful ways it can be used. And it really does—I think Dessie is so right. It really comes back to that question of the fundamentals. Just returning to the fundamentals. What could you unlearn? What could you relearn?

NETCHVOLODOFF: Yeah. Good. Well, we’re eager—my favorite part of all of these webinars is the Q&A. I’m eager to open it up. I’m just going to close by saying that students need to understand that offloading their creativity to a machine is not just cheating themselves out of real thinking, but it’s actually boring. It’s not interesting. It’s not that—to Kat’s point and to Dessie’s point, creativity is more exciting. And so I think educators have an enormous role to play in helping students understand this. So I’m now going to turn things over to my colleagues Deanna and Julia, who will help us open up the conversation for questions.

OPERATOR: Wonderful. We will take the first written question from Katie Laatikainen at Adelphi University, who says she’s using a simulation right now. She uses Model Diplomacy simulations in nearly all of her courses. But I have a question about cultivating knowledge while addressing AI literacy. It is a huge challenge for faculty to embrace AI in their pedagogy when students don’t have a foundation to be critical about the results AI provides. Unless you have that granular knowledge, it is hard to be critical about AI in what we are teaching and knowledge we are trying to establish. How do we thread that needle?

DUFFY: I mean, Dessie—

NETCHVOLODOFF: Dessie, Kat? Yeah.

DUFFY: You’ll have better tips on this than I will, certainly. Again, I would say part of it goes back to the suggestion I made around there is—there is using AI to learn, and then there’s also them learning what AI is and can do. The best way for them to develop critical thinking skills about AI’s limitations and capacities is starting where they already have critical thinking skills, which is someplace very close to home. And that’s probably very personal to each student.

So maybe part of that is actually something that doesn’t have to do with the broader subject matter of the course curriculum, but instead has to do with, like, a first week assignment of choose the thing you think you know best in the world—is that Taylor Swift, is it—like, it could be whatever. And have them do a project on that to get a sense of what worked and what didn’t. And I don’t know that that answers your question in its entirety, but it can at least help with tempering students’ reliance on it, and their understanding of the value of those building blocks, and why they matter, and why you might need to build them in ways that don’t have to do with AI.

NETCHVOLODOFF: Great.

Dessie, anything to add?

DUFFY: But, Dessie, you’re—like, you do this all the time.

NETCHVOLODOFF: You’re in the classroom. So eager to hear from you.

ZAGORCHEVA: I just wanted to add just a small point that in my classroom sometimes it helps to question the output of the AI by using different chatbots. Let’s say they’re using ChatGPT. We just asked them, OK, could you please compare this with some other chatbots? And then they started thinking about the differences. Are there some biases? Is it factual? Why is it that one responds in one way, the other one in the other way? So this can always help. And it also increases their critical thinking, because it just, again, makes them to question arguments, to see different evidence from different perspectives.

DUFFY: Yeah. Sources, as well, Dessie, I think you’re so right. You can really use the chatbots against each other in terms of verification. But you can also push them for sources. And there is a way of sort of starting big with whatever the AI produces, and then working with the students to dig down to say well, OK, you got these four competing analyzes. And they use this range of sources. Now, go research the sources. Which sources are reliable? Which are not? Which ones would you want to take and why? What are the languages they come from?

And so again, it’s—that’s not a subject matter specific thing, but that reverse engineering a little bit starts to help them understand the fundamentals of, like, what are the balanced sources? Am I only looking at, you know, one language? Should I be looking at multiple languages? Do I have the right balance of perspectives? So that’s another thing you can do, is kind of start big and then dig and dig and dig and dig into what the results were, and make the students just keep vetting it, and keep vetting it, and keep vetting it. Reverse engineering.

ZAGORCHEVA: And sometimes just by looking what sources the AI has suggested, they see that some sources just do not exist. So they were just made up.

DUFFY: Mmm hmm. Or it’s Wikipedia, Wikipedia, Wikipedia, Wikipedia, Wikipedia.

OPERATOR: We will take the next raised hand from Clemente Abrokwaa, who is an associate teaching professor at Pennsylvania State University.

Q: Hi. Thank you very much. Very, very good insights.

And one thing that I think got to mention earlier about students wanting fast answers, that’s something we see around here. It’s very, very common. They wait until the last minute when the assignment is due, and then the next day they give you a very polished English. And then you wonder.

But my question for you is—I was glad you mentioned living in Tunisia. Yeah, that was very interesting for me. I’m originally from Ghana. So my question is, how is—or, how are the African countries, especially you mentioned Tunisia, funding, you know, this project? And what are they getting the experts from?

DUFFY: It’s a great question. I would say the African Union has been doing some really interesting work in the past year trying to come up with regional approaches. You will certainly see some divides between sort of North Africa and primarily Arab—like, Arabic language countries, and then non-Arabic speaking countries. I think you’ll probably also see distinctions between Anglophile—like, or Anglophone approaches and Francophone approaches as well on the continent.

One of the things that we’ve seen is that the United Nations is actually moving much more of its work to Kenya. So Nairobi is going to become a much more significant hub than it has been on questions of international governance and digital governance. I would expect Nigeria to be a sort of leading voice in some of these areas, because Nigeria was—Nigeria and Kenya are both pretty heavy-hitters in terms of things like open data. But Senegal actually has done super—has a really interesting hub as well of folks sitting in Dakar. And Suriname has done really interesting work. And Ghana has done fantastic, interesting work.

In terms of where the money is coming from, there are funds that are being developed at the international level. I’m a little unclear on where those disbursements are right now, in part because they’re tied into what is a broader kind of weakening of multilateral funding spaces and multilateral orders, driven in part by how much the United States has pulled back from so many multilateral institutions and from UN funding. And so I think everyone, not just the African nations, are scrambling a little bit—with global health money having declined, the EU is spending a lot more money on its defense budgets than it ever did before, with pullback on NATO, you know, that U.S. has pulled out of the World Health Organization, has—so I think the question of money, and where the money is coming from, is wrapped up, honestly, in much bigger questions around how other key priorities are also getting funded, and what that looks like in terms of international aid and development.

I will say, though, just one of the things that is so exciting and interesting about what’s happening in Africa, there’s really great work coming out of South Africa. You know, there’s a two to—between 2,000 and 3,000 different languages in Africa. And I think if you were to do a survey right now, you would find that at best thirty to thirty-two of those are represented in the current language models that we have. You know, that is an enormous erasure of culture, of knowledge, of learning, of human intelligence.

And so some really fascinating work is coming out of Africa right now in thinking about how you do deeply localized work around inclusion of minority languages, of indigenous cultures, of tribal languages, of tribal expertise. What it would look like to build that into AI systems and what it would also look like to articulate that that knowledge is not reflected, and that we should not think of artificial intelligence as a general reflection of all knowledge because some knowledge is simply not going to be captured there. And we should be able to name that, and point it out, and then protect that knowledge in other ways.

NETCHVOLODOFF: I think we have time for a couple more questions here.

OPERATOR: We will take the next written question from Marcelo Rodriguez, who is a professor and librarian at the University of Arizona College of Law.

He asks: My latest conversations with students regarding AI have turned into employability. Now that the AI can pass the bar and employers seem to be using AI tools to replace entry-level positions, this has been a useful point to engage students in these conversations. Any thoughts?

ZAGORCHEVA: I think this is a useful way to engage students in conversations also about their future careers and how AI is used. Some of them think that they should use AI in their work applications. And they think that—they’re sure that employers are using AI in order to read their applications. So I think this is also a good idea to introduce some of the topics that we have been discussing, about the bias that is inherent in AI outputs, and about some of the biases that these technologies may cause. There have been cases about gender, cultural, racial biases, and whether employers will have such biases because of using AI.

So all of these big discussion of the use of AI, and ways that—the risks about using AI for such purposes, we can have with these students. And I think their interest in getting employment and how to go about this can be used for enlightening them about those things. One of my students wanted to ask, for example, how to know whether, when they apply for a job, the employer is discriminating against them because they’re using some AI and they’re not familiar with the algorithm. So we talked about transparency of algorithms, what employers could do about this, ethical uses of AI in such situations. So it is a very, really good question to start discussing all of these topics around that question.

DUFFY: I would also say the statistics might—that we have available may be a bit misleading right now. So keep your eye on them. If you are looking at statistics, let’s say American companies. And if you’re looking at statistics around increasing AI use across American employers, you are going to see staggering statistics. Ninety-five percent of companies that have over $100 million in revenue have reported adopting AI tools. Overwhelmingly that adoption is pilots. And it is a reflection of a real pressure between 2023 and 2025 from corporate boards, and from C-suites, and from investors, and from the market to be able to show that you were going to be able to achieve all of these efficiencies with AI, and produce better returns, essentially. That is not really proving to be the case.

So, so far, of that 95 percent that have adopted some sort of tool—I can’t remember the exact statistic. It’s around—maybe 10 percent, maybe 5 percent feel that they have reached adoption at scale in some area. And only 1 percent can report a positive return in terms of their profit and liability. So where we are is very much in the world of, like, throwing spaghetti against the wall. And there’s a little bit of performative attempts at AI use right now. I think we’re probably three to five years out from truly widespread AI adoption across large enterprise systems, at least. So just to temper the panic a little bit.

I do think we’re going to see large waves of layoffs because of the investment that is going into AI. So something, like, 80 percent of the capital flow in the United States since ChatGPT launched has gone into AI investments. We call that CAPEX, right? So those enormous capital flows now have to get recovered somewhere else. And one of the things that we’re all really watching is that people won’t be firing employees because AI has gotten more productive and they don’t need the employees. They’ll be firing employees because in order to offset those capital expenditures they’re going to have to have fewer expenditures on personnel, right, and on employees. And so you’re going to see cuts out of economic externalities, not actually because of increases in AI’s productivity. We are seeing some real changes in computer science and coding. So that is a—that is a really big area of distinction. We’re also seeing increases in needs for things like skilled and trained engineers, because there is such a boom in the United States around building things like data centers. So you’re seeing some areas of jobs actually really explode, especially in more vocational areas. And then you’re going to see some other areas dip.

And the last thing I would say on this is there is an amazing virtue right now to being someone who doesn’t have a lot to unlearn. So what students bring to the equation right now is a world of tools at their disposal that give them all these capacities that they just didn’t have before. They can just learn how to do things faster and more creatively. And so there are also, I think, some really interesting career paths that our young people will be able to build for themselves because they can just build more things, and do more things, and try more things out, and express it in different ways. And so I also wouldn’t sleep on all the possibilities for whole new types of research, or work, or tools, or services to be developed by the young people who haven’t internalized, well, this is the way we do it, and now we should throw AI on top of it, but are instead just looking at AI and saying, well, this thing annoys me. I wonder if I could use AI to fix it, and go off and build on their own. So I think that’s a really exciting, interesting area as well where I’m very intrigued to see where our young people go. I’m too old. I have way too much to unlearn.

NETCHVOLODOFF: (Laughs.) I’m looking to Deanna to inquire whether we have time for another question. I find all of this fascinating. I’m sure our panelists—I’m sure our participants do too. So can we squeeze in one more question, or are we coming to the end here?

OPERATOR: Absolutely.

NETCHVOLODOFF: Great.

OPERATOR: We should have time to squeeze in another. We have another question from Lady Yartey-Ajayi from Utah State University, who asks: How can educators ensure marginalized voices are not erased or misrepresented in AI-mediated learning environments?

ZAGORCHEVA: This was one of the things some of my students discussed in one of the previous classes that we discussed AI bias, and how AI is trained on a lot of material that is Western-centric. And a lot of the voices in this way have been marginalized because of lack of languages or because of their materials not being present in the datasets on which AI is trained. So they wanted really a more active position on that. And they wanted some ways to get involved. And I think this is also a way how to—related to some of our previous discussion on how to keep them engaged and how to keep them involved, is to find their—really to find their passion, to find their voice, and to help them say that they have a unique contribution. And because when they believe in that, they will be more active and they will be more involved. And I think educators should certainly do more about that, including those voices, those experiences in the community that so far have been neglected or missed, and that are not really well represented.

DUFFY: I would just—there is—there is—no technology is neutral, because no human is neutral, because data is not neutral. Because that which AI was built on is not neutral. And that is not specific to the United States, for example, right? If you think about the use of AI in India, 90 percent of the Indian internet is in Hindi. When you think about how systems have been trained, it’s probably likely that you’re going to see better results on questions or prompts that involve things associated with Hindi learning, data, knowledge, or that language, than you would with Tamil, for example, or Malayalam. Same thing with medical queries. We know, for example, in the United States that women, in particular, have faced systemic bias in terms of medical research vis-a-vis men. And we know that within women, Black women in the United States have faced systemic bias in terms of being underrepresented in broader research.

And so if you are—if that is a known problem, there are some ways that companies can try to correct for that through weighting. And you can kind of dig in to seeing if they’ve tried to address it, or if different systems have tried to address it. But if you know—if you know that those biases exist in the real world, it’s pretty easy to dig into the sort of AI world and unearth them. I will say a lot of the leading American systems, at least, are getting better on this. So if the last time that you really worked at this was, you know, twelve months or even nine months ago, with every model update I’m actually seeing improvements, if not necessarily in inherent bias, in the model’s ability to explain where bias might exist or how it could have been addressed. And so I think this will remain a problem. I don’t think it will remain the size of—and the specificity of the problem that it was two or so years ago. I hope I’m right on that. Very possible I could be wrong. (Laughs.) But that’s probably the way—that’s the way I think about it now.

NETCHVOLODOFF: Great. Great. Well, now we are—I do know the answer to this question. I think we are out of time.

And I just want to thank everyone for joining us today. And a huge thank-you to Kat and Dessie for this fascinating conversation. I’ve been seeing in the chat how, you know, excited, and stimulated, and thinking about all these issues the participants are. So I would encourage you to explore the resources, CFR Education resources, but also resources across the entire Council website, CFR.org. And also just wish you the best holidays. And encourage you to please participate in future ISA and Council events. We love having you here and are eager to hear your thoughts on this. And I know a lot of them are in the chat. We’ll go through them and we’ll respond accordingly. But, again, thank you. And thank you, Kat and Dessie, so much.

DUFFY: Well, and can I just say thank you, educators? (Laughs.)

NETCHVOLODOFF: Yeah. I mean, you all are on the front lines of protecting our democracy, of building all of this.

DUFFY: God’s work. You’re doing God’s work. Thank you, all of you.

ZAGORCHEVA: Thank you so much.

NETCHVOLODOFF: Thanks, everybody.

(END)