AI Literacy Across Disciplines: History

Insights and best practices for teaching AI literacy to history students

Posted on August 17, 2026 in Blog Posts

At a glance

Course	History, China and the World
Audience	Undergraduates across majors; no prerequisites
Instruction format	Credit course, fully online, taught in an intensive winter session
Primary AI tools featured	ChatGPT and AI Music Generator
Core AI literacy themes	Critical, computational, and creative AI literacy

Setting the stage

Like many universities, my own institution has adopted no single rule for generative AI at the time of writing University at Buffalo instead addresses the technology’s presence through its existing academic-integrity policy and leaves the particulars to individual instructors, who are expected to set and disclose their own approach. Working within that latitude, I integrated an AI-powered component into a history course, which was taught fully online during an intensive winter session. In my classroom, the critical evaluation of AI-generated knowledge constitutes an assessed component of the work. From the first session, my course policy requires students to take the following four concerns: they must use these tools transparently, verify what the tools produce, remain attentive to what the tools omit, and take responsibility for the analysis they submit.

Because the course fulfills a general education requirement, its students come from many majors and bring different experiences with the technology. Many have already used ChatGPT to summarize a reading or draft an essay, yet few have examined how these systems decide what to say. Students of history feel this tension most acutely, since ChatGPT can generate a fluent historical narrative in seconds, diluting the value of their own abilities in doing the same. Yet, they also anticipate that AI will figure into their careers. As a result,  they seek guidance that neither prohibits it nor surrenders to it completely.

The course is taught online, where students most frequently encounter machine-generated history. I treat, here, artificial intelligence as both an instrument of historical inquiry and an object of critical reflection. Each session accordingly presents a single machine output for examination rather than an open-ended invitation to experiment.

---

Do you teach AI literacy?

We want to hear from you! Our “AI Literacy Across Disciplines” series seeks to highlight how librarians and educators are engaging their academic audiences, whether undergraduates or advanced researchers, about AI in their fields. We’re seeking to curate insight, develop best practices, and figure out this new landscape together. Read the call for papers.

Learn more

---

Teaching AI Literacy

My guiding principle is that the AI-generated content must be examined rather than simply used without scrutiny. I organize each session so that a machine output serves as a specimen for analysis, while the student’s evidence-based critique constitutes the graded work. The title of the midterm captures this principle: “Generative AI Is Splendid, but Historical Literacies Guard Me.” These tools can accelerate historical work, but they cannot replace the source criticism, contextual judgment, and ethical responsibility at the center of the discipline. My learning outcomes follow accordingly. Students should be able to weigh a machine’s claims against primary sources and relevant scholarship. They should recognize hallucination, fabrication, bias, and omission, and explain how training data and prompting shape what a system asserts.

To this end, I organize the curriculum around three interlocking competencies: Critical, Computational, and Creative. Each is embodied in an assignment that pairs a machine output with a written reflection.

Critical AI literacy

The first competency, critical AI literacy, extends a practice that history has always taught. In one assignment, students select a figure from the transnational history of Chinese sport, among them the physical educator Charles H. McCloy and the table-tennis champion Rong Guotuan, read the assigned scholarship, and then compare that scholarship with ChatGPT’s account of the same figure. The discrepancies are instructive. The model dates McCloy’s arrival in China to the 1920s, although he had served since 1907, and it reports Rong’s birthplace as Hunan when he was in fact born in Hong Kong. Students record these errors in a short reflective essay and consider why a fluent, confident answer so readily conceals them.

A second assignment addresses bias and the politics of knowledge. Students ask ChatGPT to assess paired male and female figures from Chinese history, such as Song Qingling and her husband, Sun Yat-sen, and then contrast the results with the judgments of gender historians. The pattern is consistent: ChatGPT introduces Song primarily as Sun’s wife and passes over her advocacy for gender equality. Students post their analyses to Padlet and respond to one another, recognizing that AI-generated history is neither simply right nor simply wrong. It reflects the unevenness of the archive and the bias embedded in training data.

Computational literacy

The second competency regards computational literacy. Artificial intelligence lowers a technical barrier without relieving students of the labor of thought, especially when adding computational components . Because most of my students have no background in programming, I have them ask ChatGPT to generate the code for a method, such as a word cloud, a topic model, or a word-embedding analysis. They then run that code on a primary source in a notebook environment. A word cloud of Zhou Enlai’s 1955 speech at the Bandung Conference, for example, foregrounds “China,” “Asia,” and “common ground,” which opens a discussion of China’s postcolonial self-positioning that might not be obvious from the generated text alone.

For a larger primary source, such as a thousand-page volume of 1947 US diplomatic records on China, a word-embedding analysis ranks “military” and “ammunition” among the terms most closely associated with “Chinese,” pointing to the centrality of the US government’s military aid to the Nationalists during the Chinese Civil War. Students read the machine’s results against their own close reading, ask what it foregrounds and what it suppresses, and treat the analysis as an interpretive artifact, not an objective fact.

Creativity

The third competency concerns creativity. It asks students to translate a historical interpretation into creative work. The midterm draws on the tradition of the “unessay,” which Jessamyn Neuhaus and Christopher Gates describe as an assignment that teaches history through a medium other than the conventional paper. A student might use an AI Music Generator to compose a song on themes such as Mao’s diplomacy and then submit 200 words of critical reflection.

While the generative tools make such production accessible to students without training in musical composition, the disciplinary requirement remains firm for creative assignments. Every project must advance a historical argument: what it claims, what evidence supports it, whom it addresses, and what its chosen form clarifies or distorts. I assess the reflection alongside the quality of the creative work.

These three competencies are not separate units but mutually reinforcing dimensions of a single pedagogy. Computational and creative work earn their place only when subjected to critical judgment. Critical literacy acquires its force only when students apply it to a source they have analyzed or a song they have made. The pedagogy aims to ask, of any AI output, how its claims might be verified, what it omits, and which perspective it encodes.

While adding AI-powered components into the classroom, I still reserve room for students with a preference for traditional assignment formats. Each task is offered alongside an alternative: one that examines a machine output, and the other following a conventional format, such as a primary-source essay on the same figure or theme. The parallel track keeps the graded work centered on historical judgment rather than on the tool, and it preserves each student’s choice about how far the technology enters their own learning.

Reflecting on Impact

The clearest sign of success for the course is that students cease to treat machine-generated prose as authoritative once they have audited several outputs against evidence. Students often conflate plausibility with evidence until an assignment obliges them to test a claim and watch it collapse. My advice to colleagues and librarians is that students should be required to read the evidence independently before they generate anything, document their prompts, and have a no-AI option.

Additional resources

• Shu Wan, “The Jack in the Black Box: Teaching College Students to Use ChatGPT Critically,” Information Technology and Libraries 43, no. 3 (September 2024). https://doi.org/10.5860/ital.v43i3.17234
• American Historical Association, Guiding Principles for Artificial Intelligence in History Education (2025): https://www.historians.org/resource/guiding-principles-for-artificial-intelligence-in-history-education/
• “Artificial Intelligence and the Practice of History: A Forum,” American Historical Review 128, no. 3 (September 2023): 1345–89.
• Jessamyn Neuhaus, “Introduction to the Special Issue on Teaching History with the Unessay,” Teaching History: A Journal of Methods 47, no. 1 (2022): 2–5.
• Benjamin M. Schmidt, “Words Alone: Dismantling Topic Models in the Humanities,” Journal of Digital Humanities 2, no. 1 (2012).

---

📣 Do you teach AI literacy? We want to hear from you. Check out our call for papers, “AI Literacy Across Disciplines.”

Related Posts