From Technocuriosity to Chatty Geeps: Experimenting with GenAI in the History Classroom

Written By Daniel Krutka

Civics of Technology Announcements

Book Clubs:

Book Reviews:

Amy Allen is an associate professor of History and Social Science Education in the Elementary Education program at Virginia Tech. Her research focuses on history education, with particular interests in generative artificial intelligence, religion and education, and signature social studies pedagogies. Her recent work examines how emerging technologies can support disciplinary inquiry.

David Hicks is a professor of History and Social Science Education in the Secondary Education program at Virginia Tech. His research focuses on history education, particularly the intersections of technology, historical thinking, and teaching and learning. His recent work explores the pedagogical possibilities and challenges of emerging technologies, including generative artificial intelligence, for history education.

Last month at the 2026 Civics of Technology conference, we presented a session called “Ghosts in the Archive: The Burden of Representation.” We focused on one use of GenAI that we find especially troubling despite its growing popularity in social studies classrooms: inviting students to “talk” with historical figures. Our basic recommendation was pretty simple: please don’t.

One of the examples we shared was a Character.AI version of Anne Frank who could tell us how many people died in the Holocaust (despite dying in 1945) and also named J. K. Rowling as one of her literary inspirations. There are obvious historical problems here, but there is also a larger representational one. We describe these synthetic historical voices as algorithmic palimpsests. They layer together existing accounts, representations, cultural memories, and all kinds of other things we cannot see into something new that sounds remarkably convincing. They cannot tell us what Anne Frank would have said, but they might tell us something about how Anne Frank has been represented and remembered. Rather than treating the transcript as a historical source, perhaps we can treat it as an artifact of the present.

There are many reasons to be skeptical of GenAI platforms. As seen here, they hallucinate, reproduce biases, and can generate remarkably confident accounts of the past without much concern for evidence. Beyond what appears on the screen are larger questions about privacy, labor, environmental costs, corporate power, and the increasingly prominent role technology companies play in shaping education. And history should make us wary of claims that the newest technology will transform teaching and learning. We have heard versions of that promise before (Cuban, 2003). We take these concerns seriously.

Our “Ghosts in the Archive” presentation grew out of one chapter in our new book, Teaching History with Chatty Geeps: A Technocurious Approach to Generative AI in the Classroom. The book itself grew out of an idea we first wrote about on this blog last fall: technocuriosity. We argued that history educators should neither embrace GenAI uncritically nor sit on the sidelines while edtech companies, administrators, and policymakers decide what these technologies will mean for our classrooms. Over the past few years, we have continued to play with, test, and break Chatty Geeps (our nickname for ChatGPT).

Despite the title, though, this is really a book about teaching history. Our thinking about GenAI came much later. David still describes himself as a Schools Council Project History kid, shaped by learning history through inquiry, source gobbets, and competing accounts. Our later work together with teachers and primary sources through the Library of Congress reinforced that commitment. Students do not automatically know how to source, contextualize, corroborate, reason about causation, or construct historical arguments. Those practices have to be taught, modeled, scaffolded, practiced, and eventually performed independently.

Who is doing the thinking?

Ask Chatty Geeps to analyze a primary source and it will. It can identify perspective, provide context, summarize an argument, and explain significance, usually in a matter of seconds. Sometimes it does this pretty well. But so what?

If the learning goal is for students to learn how to analyze historical sources, a really good AI-generated analysis is actually a really bad and irrelevant artifact. Sourcing, contextualizing, corroborating, and constructing interpretations from evidence are not simply steps toward producing an answer; they are part of the intellectual work through which students learn to think historically (Seixas & Morton, 2013). If GenAI performs that work for students, the quality of the resulting output tells us very little about the quality of the learning.

In contrast, conversations about GenAI in education often focus on the quality of its outputs: Is the response accurate? Is it biased? Did it hallucinate? Those are important questions, especially in history. But an accurate output does not necessarily support learning:— so we have become interested in figuring GenAI differently. What happens if Chatty Geeps is told that it cannot analyze the source for the student? Instead, it asks a question, waits for the student to respond, probes that response, asks for evidence, or pushes the student to reconsider an unsupported claim. Now we potentially have a scaffold rather than a shortcut. Potentially.

Scaffolding and offloading can look a lot alike

Scaffolding is hardly a new idea. It provides support that enables a learner to accomplish aspects of a task that would otherwise be beyond their current capacity. GenAI, however, complicates this relationship because the assistance it provides can be nearly unlimited and immediately available. A student can receive individualized questions, explanations, feedback, suggestions, revisions, and examples in seconds. These capabilities are often presented as educational affordances… and they may well be. But the boundary between scaffolding intellectual work and offloading is quite thin.

That problem is not solved through better prompt engineering. It requires pedagogical judgment about what we actually want students to know and be able to do. When we design educational activities (with or without AI), we continually return to the intellectual work of history: sourcing, contextualizing, corroborating, questioning, interpreting, constructing arguments, and evaluating evidence. Where is that work happening? What is the student doing? What is Chatty Geeps doing? What is the student learning to do without Chatty Geeps? And would the student be better served if Chatty Geeps were not there?

Sometimes experimentation gives us better reasons not to use AI

Technocuriosity has led us to find some uses of GenAI that we think are worth exploring, but it has also made us more skeptical of others. Technocuriosity is not a pathway toward eventual AI adoption. We do not begin with the assumption that every history lesson has a GenAI-shaped hole waiting to be filled, nor do we assume that because a technology can accomplish something efficiently, it should be used to accomplish it. Technological change always involves tradeoffs, and those tradeoffs are unevenly distributed (Postman, 1985).

This tension is where we see technocuriosity in conversation with a growing body of critical and technoskeptical scholarship. Pleasants, Krutka, and Nichols (2023), for example, ask educators to interrogate the relationships with technology that schools normalize and to consider what technologies require, privilege, and displace. Clark and van Kessel's (2025) AI in Social Studies Education extends related questions into social studies education through an emphasis on thoughtful practice, disciplinary inquiry, and human judgment. Budhai and Heath's (2025) Critical AI in K–12 Classrooms places questions of justice, power, surveillance, and agency at the center of decisions about AI in schools.

We share many of these concerns but emphasize what becomes visible when critique is paired with situated experimentation. What do we learn by actually working with these systems, pushing against their boundaries, documenting their failures, and sometimes finding possibilities we had not anticipated? For us, then, experimentation is not an alternative to critique. Sometimes working with these systems gives us a much clearer sense of what the problems actually are.

One reason we keep returning to this work is that history education seems like a particularly useful place from which to investigate GenAI. History teachers already ask students to question authority, interrogate sources, corroborate claims, attend to context, identify absences, evaluate evidence, and recognize that a coherent narrative is not necessarily defensible. These are useful habits when encountering a technology designed to generate remarkably coherent language. They are also reasons history educators should have some say in what happens next. If we do not experiment with GenAI and develop informed arguments about where it supports learning, where it interferes with learning, and where it does not belong at all, those decisions will increasingly be made by people with considerably less knowledge of pedagogy, historical thinking, and the particular students sitting in our classrooms.

What does this look like in practice?

Much of Teaching History with Chatty Geeps is devoted to this practical question. We start with the kinds of intellectual work already happening in history classrooms and ask where, if anywhere, GenAI might support that work.

Across the book, we play with GenAI and primary sources, historical questioning, self-regulated learning, chronology, causation, counterfactual reasoning, assessment, historical representation, critical AI literacy, and digital wellness. Some of these experiments worked, but many also provided reasons not to use AI for that purpose at all. Importantly, none began with the question, “How can we add AI to this lesson?” They began with historical learning; then we asked whether there is a defensible role for GenAI within it.

We also made the book something of a “Choose Your Own Adventure.” Read it cover to cover or skip the chapters that don’t interest you. Try one thing, try five things, or try nothing.

We should probably also acknowledge that the book contains rather more Alice in Wonderland, Victorian history, strange historical detours, footnotes, and unusual punctuation than the title might suggest. Chatty Geeps has also contributed indirectly to David’s campaign to revive the dog’s bollocks (:—), partly because the em dash has become evidence that everything was written by AI. (As an aside: during copyediting, he discovered that he had been using the wrong version of the dog’s bollocks throughout the manuscript. There is probably a lesson about technological expertise in there somewhere…).

Our work on technocuriosity begins from the recognition that technologies are not neutral tools; they embody choices, values, and particular ideas about what they are for. Our contribution, we hope, is to bring those questions more directly into the disciplinary work of history. History educators should be part of figuring out what learning with GenAI becomes, particularly while there is still so much left to figure out. Ultimately, that is what Teaching History with Chatty Geeps is about. It is not an argument that history teachers need more (or any!) GenAI in their classrooms. It is an invitation for history educators to help figure out, with disciplinary and pedagogical care, where these technologies might belong and where they do not.

If GenAI leaves you cold, come for Alice. Stay for the primary sources, spirit photography, historical thinking, questionable footnotes, and the dog’s bollocks.

And because we really do want teachers to be part of that conversation, the book is free. Teaching History with Chatty Geeps: A Technocurious Approach to Generative AI in the Classroom is available open access from Virginia Tech Press.

References

Allen, A., & Hicks, D. (2026). Teaching history with Chatty Geeps: A technocurious approach to generative AI in the classroom. Virginia Tech Press. https://doi.org/10.21061/chattygeeps

Budhai, S. S., & Heath, M. K. (2025). Critical AI in K–12 classrooms: A practical guide for cultivating justice and joy. Harvard Education Press.

Clark, C. H., & van Kessel, C. (Eds.). (2025). AI in social studies education: Tools for thoughtful practice with generative artificial intelligence. Teachers College Press.

Cuban, L. (2003). Oversold and underused. Harvard University Press.

Pleasants, J., Krutka, D. G., & Nichols, T. P. (2023). What relationships do we want with technology? Toward technoskepticism in schools. Harvard Educational Review, 93(4), 486–516. https://doi.org/10.17763/1943-5045-93.4.486

Postman, N. (1985). Amusing ourselves to death: Public discourse in the age of show business. Penguin.

Seixas, P., & Morton, T. (2013). The big six historical thinking concepts. Nelson Education.

Daniel Krutka
Next

Disability, AI, and the Limits of Accessibility