The Pedagogy of Talking to the Dead: Inquiry, Empathy, and Historical Thinking
The phrase “talking to the dead” sounds glib in a marketing deck and serious in a history classroom. What an AI historical-figure chat actually does for students is not impersonation; it is invitation. The student is invited to ask a question of someone who, by definition, will never answer in real time again. That invitation makes a different demand than a textbook does. The demand is the foundation of what Sam Wineburg has spent thirty years calling historical thinking, and the implications for K-12 classrooms are unfinished.
This piece is a thought-leadership essay for department chairs, instructional coaches, and curriculum directors thinking about where AI fits in social studies for the rest of the decade. The frame is theoretical but the implication is concrete. If the discipline is teaching students to ask better questions, AI historical conversations are one of the few new tools that change what questioning looks like inside a 45-minute period.
The Wineburg argument and why it still matters
In Historical Thinking and Other Unnatural Acts (Temple University Press, 2001; winner of the 2002 Frederick W. Ness Award for the most important contribution to understanding the liberal arts), Wineburg argues that historical thinking is “neither a natural process nor something that springs automatically from psychological development. Its achievement goes against the grain of how we ordinarily think.” Wineburg is a Stanford professor of both history and education and trained as a psychologist, which is what gave the argument its analytic teeth: the book was not philosophical advocacy but a cognitive account of what historians actually do that students do not yet do. That account shaped most of what the Stanford History Education Group / Digital Inquiry Group has produced in the two decades since. The discipline of history, in this reading, is not the accumulation of facts but the cultivation of a counter-natural cognitive habit, which students do not naturally exercise. The classroom teaches them to source, contextualize, corroborate, and read documents closely.
The upstream version of the same point sits in Dimension 1 of the C3 Framework: “Developing Questions and Planning Inquiries.” Before sourcing, contextualization, or corroboration, the student needs a question worth asking, and the framework distinguishes carefully between compelling questions (suggestive, provocative, worth further exploration, typically drawing on multiple disciplines and focused on real social problems) and supporting questions (discipline-specific scaffolding that contributes the knowledge needed to answer the compelling question). The C3 authors are direct about the alternative. They write that when social studies is reduced to “reading textbooks to answer end-of-chapter questions and taking multiple-choice tests,” students “quickly become disengaged.” The disengagement is not a motivation problem; it is a question problem.
A textbook does not let a student ask. An AI historical-figure conversation does.
What “talking to the dead” actually demands of a student
The cognitive work of asking a historical figure a question is more demanding than asking the same question of a search engine, for three reasons that compound on each other.
The first is that the figure has a vantage point. A student asking Frederick Douglass why he wrote the 1852 “What to the Slave is the Fourth of July?” speech has to engage with Douglass as a person inside a specific political and rhetorical situation in 1852, not as a Wikipedia entry. The conversation requires the student to model the figure’s perspective, which is the move historians call historical empathy, and which the SHEG / Digital Inquiry Group curriculum has been teaching variants of for two decades. The AI conversation operationalizes that move for every student in the class simultaneously, rather than serially through the teacher.
The second is that the figure has a record. A source-grounded AI conversation is, in effect, a guided rereading of the documentary record the figure left behind. A student asking the AI Lincoln what was happening the week he drafted the Gettysburg Address is engaging with Lincoln’s actual correspondence, his earlier speeches, and the contemporaneous press coverage preserved in the LOC Lincoln Papers . The conversation is a rehearsal of the close-reading work the assessment will measure.
The third reason is that the figure has limits, and a serious platform makes those limits visible. A figure can only speak from the record they actually left, so when a student asks a question the documentary record cannot support, the figure should say so explicitly rather than improvise. That epistemological humility is exactly the move the discipline wants students to internalize. The historian who knows what the record cannot tell them is the historian who reads the record honestly.
The empathy question, handled carefully
“Empathy” is a word the discipline uses carefully and with reason. Historical empathy is not the same as sympathy, and it does not require endorsement, identification, or the suspension of moral judgment. Wineburg himself has been explicit on this point : the discipline asks students to model how a historical figure understood their situation while keeping intact the present-day capacity to judge that figure’s choices. A 14-year-old who can engage with Robert E. Lee’s strategic reasoning at Gettysburg without losing the capacity to recognize the Confederacy’s defense of slavery is doing the discipline’s hardest cognitive work.
AI historical conversations support this carefully or carelessly depending on the platform. The careful version anchors the figure to the documentary record, restricts the topic scope where appropriate, and surfaces the figure’s perspective as one perspective among many that the student is responsible for weighing. The careless version flattens the figure into a generic chatbot voice, removes the documentary anchor, and produces what amounts to historical fiction. The two design choices produce different classroom outcomes, and they should not be conflated in procurement conversations.
For sensitive history (the Holocaust, slavery, colonialism, residential schools, civil rights atrocities), the empathy frame requires additional teacher framing. UNESCO’s report on AI and Holocaust education , produced with the World Jewish Congress, makes the case directly: historical empathy with survivors and witnesses is the pedagogical goal; freeform impersonation of perpetrators is not. The line matters more here than in other periods because the stakes of getting it wrong are higher.
What changes in a classroom that does this well
When a department adopts AI historical conversations as a structured pedagogy rather than as a novelty, three things change in student behavior, and they are the same three changes Wineburg’s research and decades of SHEG / DIG follow-up work have associated with strong primary-source-based instruction.
The first is that students start asking sharper questions in non-AI contexts. A student who has spent a semester practicing inquiry with a source-grounded AI figure tends to ask sharper questions of their teacher, their textbook, and their classmates as well. The chat is the rehearsal; the broader discipline is the muscle it builds.
Primary-source citation is the second change. Students who have spent more time inside the documents asking specific questions of specific passages tend to cite them more accurately in their writing afterward, which is the small-but-visible shift teachers tend to report first in survey responses.
The third change is harder to measure but matters most: students start engaging with disagreement as a feature of the discipline rather than a bug. Two figures with opposing vantage points produce a richer student response than a single figure does, and the corroboration move SHEG identified two decades ago becomes practiced rather than aspirational.
The implications for department-level decisions
A department chair thinking about AI for the 2026–27 school year and beyond should be asking three questions that follow from this pedagogy rather than from the marketing.
The first concerns disciplinary moves: which moves does our current pedagogy actually teach, and where are the gaps that AI conversations could close? For most departments, the gap is in scaffolded inquiry practice. Students do not get enough at-bats with sourcing, contextualization, and corroboration before the assessment asks them to demonstrate it, and a structured chat sequence is one of the few mechanisms that can multiply the at-bats without multiplying teacher hours.
The second is about historical empathy as the department understands it. A platform whose figures are source-grounded supports empathy as the discipline defines it (modeling perspective without endorsing it). A platform whose figures are freeform impersonations does not; it produces something closer to historical fiction, and the difference shows up in student work.
The third is the procurement-defense question. Can the choice we are about to make stand up in front of a school board, a parent, or a journalist who has just read a headline about another AI Hitler chatbot? A platform whose design choices on sensitive history, source grounding, and teacher controls are public and durable is a different proposition than one whose choices live only in product configuration and shifting terms of service.
What this essay is trying to argue
The pedagogy of talking to the dead is not a metaphor for impersonation. It is a literal description of what historical thinking has always asked students to do: read the document, ask the figure why they wrote it, reread the document with the answer in mind, disagree with the figure where the evidence warrants it, form a defensible position, and write the position down with the evidence attached. The sequence has been the discipline’s spine since Carl Becker called every man his own historian in 1931, and Wineburg’s cognitive account in 2001 explained why it remains so hard for students to perform without scaffolding.
What AI historical-figure conversations change is the scale at which the scaffolding can run. Every student in a 28-person class can now do the asking simultaneously, with the teacher’s framing and review running over the top, and the classroom can finally support the inquiry-arc work the C3 Framework has been describing since 2013 in a way the textbook alone never could.
If you are a department chair or curriculum director thinking about where AI fits in your social studies program over the next three years, the Humy newsletter is where we publish what we are learning from the K-12 departments doing this work in real classrooms. Subscribe to the Humy newsletter and the next update on inquiry-based AI pedagogy will arrive when we publish it.