The School of History at Nanjing University has launched what might be its most provocative academic experiment yet: a history essay competition that not only encourages undergraduates to use AI fully and boldly, but requires them to disclose every step of the process — which sources they consulted, what prompts they fed the AI, and what responses they received.
Professor Wang Tao, Vice Dean of the School and the competition's organizer, told reporters that rather than allowing AI to be used covertly, it was better to bring it into the open — to test and revise the rules under real writing conditions. This, he said, may be the most candid conversation the history discipline has had with the AI wave.
From Imagination to Practice: An Experiment That Had to Happen
The competition, initiated by Nanjing University in collaboration with Tsinghua University, Fudan University, Zhejiang University, and several other leading institutions, is open to second-, third-, and fourth-year undergraduates. Participants are required to complete their history papers using a designated platform: the large language model Tongyi Qianwen and the Digital and Intelligent Literature Processing Platform developed independently by Nanjing University. Full process documentation must be submitted alongside the final essay.
In humanities research especially, our understanding of AI still largely exists at the level of imagination, Professor Wang acknowledged. He noted that while debate about AI-assisted writing has been extensive, there are virtually no concrete cases showing just how good — or how bad — it can actually get.That gap is precisely what the competition aims to address. Rather than arguing over whether to restrict or embrace AI, the organizers chose to create a space for practice that is, in Wang's words, open, comprehensive, and reproducible.
We hope that through this hands-on activity, we can work backwards to reconstruct what academic ethics in history — and in the humanities more broadly — should actually look like, Wang said.
Designed Freedom, Designed Constraints
Unlike competitions that simply celebrate AI-generated writing, this one pairs its encouragement of bold experimentation with carefully considered constraints — a combination Wang identifies as the initiative's most original contribution. The first constraint is the use of a single, standardized AI model, Tongyi Qianwen. Beyond practical considerations of technical support, the decision is meant to level the playing field between different models, so that what we are really evaluating is the participant's AI literacy. The second — and more fundamental — constraint is the requirement that all work be conducted through Nanjing University's Digital Humanities Literature Processing Platform, developed by a team led by Jin Bowen, a junior faculty member at the School. The platform mirrors the workflow of traditional historical scholarship: from OCR digitization, through semantic search and the compilation of primary source materials, to analysis and synthesis.
Wang explained that the design is intended to keep the entire writing process anchored, as far as possible, within the academic logic of historical research — rather than simply prompting a large language model to generate text at will. Crucially, the platform automatically logs all user activity, making the entire process transparent and traceable. This transparency is deliberate: it is designed to reinforce the student's role as the decision-maker. Our greatest concern is how human judgment and human values are to find expression, Wang said. The competition's title — I Write History with AI — is itself carefully chosen, placing the I conspicuously before the AI.
The subject is 'I.' AI is only the tool. Through this platform, students must form their own questions, exercise their own judgment, and make their own choices. We want them to understand that they are the core of the research.
Human–AI Co-Review and the Prize for Most Contested Paper
The judging mechanism is equally experimental. Scores are assigned through a human–AI co-review model, with both AI systems and human experts evaluating submissions independently. A special award — the Most Contested Paper Prize — will be given to the entry with the greatest divergence between the two sets of scores.
Wang described the award as a test of our own assumptions. Human experts bring their own scholarly preferences and perspectives; AI systems may encode their own implicit standards of what constitutes a good essay. Does this gap actually exist? Wang asked. And if it does, what does it reveal? That is what we want to find out.
He added that the experiment addresses a phenomenon already well underway in academic publishing: Many peer reviewers at journals are now using AI to review manuscripts. So our competition is trying both things at once — and seeing what new questions emerge.
On the specific impact of AI on historical research, Wang identified its greatest contribution as sheer breadth and efficiency in handling source material. AI's capacity to synthesize and summarize far surpasses that of any human scholar. It can help us see as much of the available material as possible, compensating for the limitations that researchers face due to finite energy, language barriers, and blind spots.
Yet the risks — chief among them AI hallucination — are equally clear. AI can efficiently produce errors that appear entirely plausible, and verifying those errors may cost more time and effort than working from scratch. Wang, however, does not regard this as an intractable new problem. Even without AI, human scholars compiling sources will inevitably harbor biases and miss things. The responsibility, ultimately, rests with the researcher.
Redefining the Core: What Is the Future of the Humanities Scholar?
When asked whether he feared students would grow dependent on AI and neglect fundamental scholarly skills, Wang offered a more searching answer.
The decision to restrict participation to second-year students and above, he explained, was deliberate: organizers wanted participants to have had at least one year to develop a basic grounding in historical method, ethics, and craft before entering.
But his deeper aim is to confront students with an ultimate question: if AI eventually becomes omniscient and infallible, what will be the core competitive advantage of the human scholar?
We don't have that answer yet, Wang said. But through this competition, we are asking students to find out for themselves. They are free to submit AI-generated text without a single edit — that is their own choice. We want them to understand what role they are actually playing in this process. More than anything, we hope they can use this experience to rebuild their own sense of agency as students of history.
Broader Implications for Academic Evaluation
The competition has received the endorsement of nearly ten leading university history departments, including those of Tsinghua, Fudan, Zhejiang University, and Shaanxi Normal University. Wang described the collaboration with characteristic understatement as huddling together for warmth.
The colleagues at the partner institutions have all been thinking and working at the intersection of AI and historical studies, he said. Together, we hope to push the broader scholarly community to take this question seriously.
The ultimate aim, Wang made clear, is not simply to judge essays, but to provoke a fundamental rethinking of how academic work is evaluated. Our greatest hope is that the academic world at large will come to see that the current model — centred almost exclusively on the publication of journal articles — may already be obsolete. In an era when AI can competently produce literature reviews on demand, he argued, scholars must reconsider what academic value actually means. Perhaps the number of CSSCI journal articles published will no longer be the most important measure of a scholar's worth.
The essay submission window runs from April 21 to May 10, 2026, with results to be announced in June. Following the competition, the organizers will host a symposium bringing together prizewinners and judges to discuss the results and data, opening a wider conversation on historical writing and undergraduate education in the age of AI — and offering, from the ground up, one discipline's practical answer to the challenges of the digital era.
What do you think? Share your thoughts in the comments.
*Tranlators: Zhang Junsong, Yang Lingwei
