Distinguished digital humanists from all around the world, welcome to Korea!
It is an overwhelming joy for me to stand before you today as ADHO holds its 36th annual conference here in Korea. Eight years ago, in Mexico City at DH2018, I had the privilege of presenting our bid to host this global conference on behalf of KADH, the Korean Association of Digital Humanities. What KADH envisioned and prepared for over the last eight years has finally come to life today in Daejeon. I would like to extend my deepest gratitude to ADHO and the DH2026 Local Organizing Committee for their selfless dedication in making this historic event possible.
To be honest, as I prepared for this keynote on such a historic day, I pondered deeply about what message I could share with the leading minds of our global community. I know many of you in this room possess far deeper technical knowledge and richer inspiration regarding cutting-edge innovations than I do.
After much reflection, I realized that the most meaningful contribution I can offer is a candid reflection on my own four-decade journey—a personal quest to bridge traditional humanities with digital technology. In short, this is the story of a classical humanist who has spent a lifetime striving to find a path through a rapidly changing digital landscape.
I am a classical humanist at heart, passionate about reading and interpreting Classical Chinese texts from East Asia. Back in my twenties, my daily life revolved entirely around deciphering these ancient writings. But early on, I was struck by an overwhelming sense of frustration: the sheer volume of East Asian classical texts was staggering, while the number of scholars available to study them was far too small.
In those moments of despair, a faint hope emerged: Could computers become a partner in classical scholarship?
Right after completing my doctoral coursework in Korean Philosophy, I joined KAIST—the Korea Advanced Institute of Science and Technology, which was then the nerve center of Korean computer science research. There, I immersed myself in training and hands-on development involving databases, text archives, and information retrieval technology.
Then, in 1995, I introduced The Annals of the Joseon Dynasty to the world on CD-ROM. At the time, CD-ROM was the absolute cutting edge of digital media. The Annals is a monumental historical archive—a national treasure that virtually every Korean knows and takes pride in. With that project, Korean society began to realize for the very first time that computers could actually serve academic and educational work in the humanities.
In the early 2000s, under the banner of "Developing into a Digitally Based Knowledge Powerhouse," the Korean government poured massive financial resources into building data archives. Thanks to the precedent set by The Annals, classical humanities became one of the major beneficiaries of this policy.
As a result, vast digital archives became the new libraries for classical scholars. Yet, having played a central role in planning these projects from the very beginning, I began to see a new problem emerging.
While accessibility to classical materials improved beyond measure, the actual time required to read and interpret those texts did not decrease at all. Furthermore, we discovered that classical scholars were actually utilizing only a tiny fraction of the entire archived collection. Recognizing this gap led me to a skeptical question: “Was building these foundational data archives really worth the massive effort?”
Yet, this doubt and disappointment gave rise to a fresh perspective on the future of classical humanities archives:
Returning to my original field of classical humanities as a professor at the Academy of Korean Studies in 2004, I seriously sought to prepare for the future of the humanities data archive.
Every year, as my students and I finalized a new data curation project and drafted our final report, I made sure to include one concluding statement:
"The primary consumer of the semantic data in our archives will be AI, rather than humans. And AI will use this data to help human audiences access the world of humanities far more easily."
Since ChatGPT arrived at the end of 2022, AI's ability to interpret classical Chinese has developed to a degree I can only call astonishing. For over 18 years my team and I had built XML archives of the classics and ontologies of Korea's traditional culture — and now the AI-based environment we had prepared for seemed finally within reach. We were overjoyed.
But this brought another frustration. Many scholars of the classics fear that an AI fluent in classical Chinese could endanger the very existence of their field — and honestly, I cannot fully shake that worry myself. Their skepticism does not come from AI falling short of our expectations. It comes from a fear that AI may extinguish something important that humans have long done. And the danger is not simply that AI takes over the work — it is that an “abundance of easily accessible knowledge” brought by AI might wither human self-directed intellectual inquiry. And, in the end, neither human nor AI truly does it any longer. That is what extinction would mean. However strong the results of AI-based research I might show, as long as my colleagues see them as the work of a machine rather than of a human, they will count them not as the renewal of the humanities but as its disappearance — and work the academy takes no interest in cannot be sustained.
So this third frustration set my third direction, the one that shapes everything I will say today. In building an AI-based environment for the classical humanities, the first priority must be to ensure that human scholars hold the initiative in the work.
The core of the "humanities crisis" so heavily discussed in the AI era is not simply whether AI will replace human scholars, or whether AI will erase the humanities altogether. Over time, certain tasks inevitably fade as eras change. We must ask whether classical humanities falls into that category.
The real question we ought to ask is this: “In the age of AI, do humans still need to learn, explore, and study the classics?” Finding the answer to this question is where the future path of classical humanities must begin.
To start, I want to examine why people view AI and humanistic inquiry as opposing forces. What do we truly mean by "humanities"? Here, rather than treating humanities merely as a modern academic discipline, I prefer to understand it through the ancient concepts of Xue (學 - Learning) and Xi (習 - Practice).
Many of you may be familiar with the very first line of Confucius’s Analects:
“學而時習之, 不亦說乎?”
“To learn, and in time to practice what one has learned — is this not a joy?”
— The Analects, Book 1 (『論語』 學而)
This short sentence holds three essential elements that define the true nature of humanistic study: Xue (學 - Acquisition of Knowledge), Xi (習 - Practical Application), and Yue (悅 - Joy of Fulfillment).
First, Xue (學) is learning and acquiring knowledge—understanding new facts, grasping new perspectives, and learning what was previously unknown. Modern research and education have mostly concentrated on this single dimension.
Second, Xi (習) is putting that knowledge into practice in real life. It is the process of testing what you learned through your own actions and experience, applying it, and making it truly your own.
Third, Yue (悅) is the joy derived from that practice. It is not simple entertainment, but a profound sense of achievement, self-worth, and personal growth.
According to Confucius, scholarship does not end with merely "knowing things". Learning must lead to practice, and practice must lead to joy. That joy then becomes the driving fuel for further learning. In other words, true scholarship is a continuous, escalating cycle of Knowledge, Practice, and Fulfillment.
If scholarship were defined solely as accumulating vast knowledge, AI would already be a far superior competitor. Humans can never match AI in remembering massive datasets, searching across countless archives instantly, or translating and summarizing texts in seconds. If we view humanities strictly as information storage and processing, human researchers are bound to feel powerless before AI.
However, if the essence of scholarship goes beyond mere knowledge acquisition, the story changes completely.
In the realm of Xi (Practice), AI cannot act on behalf of humans. And in the realm of Yue (Joy and Meaning), that experience belongs uniquely to humans. AI can generate information, but it cannot practice knowledge in its own life, nor can it feel fulfillment or growth through that process. That is something only humans can do.
From this perspective, the anxiety scholars feel before AI is not entirely caused by AI itself. Perhaps we should reflect on whether we failed to pass down the original spirit of the humanities for too long, confining scholarship strictly to the bounds of knowledge acquisition. AI is not threatening us; rather, its arrival is compelling us to remember the forgotten essence of the humanities.
Regarding this, a Korean Confucian scholar Jeong Yak-yong (1762 - 1836) offered a profoundly important insight when interpreting this passage from the Analects:
“學所以知也, 習所以行也, 學而時習者, 知行兼進也. 後世之學. 學而不習. 所以無可悅也.”
“Learning is for knowing; practice is for doing. To learn and in time to practice is to let knowing and doing advance together. The learning of later ages has no joy, because it learns but never practices.”
— Jeong Yak-yong, Noneo gogeumju (丁若鏞, 『論語古今註』)
This interpretation offers a deep resonance for our AI age. Learning is essential—without it, there is nothing to practice and no joy to gain. But learning must not stop there. When learning connects to practice, it gains meaning in life, and the joy born from practice fuels even deeper learning. That is how scholarship becomes a living, vital activity.
Therefore, sustaining the value of classical humanities in the age of AI is not about knowing more facts or producing more precise translations than AI. That is what AI already does well, or will soon do better. Instead, our mission is to move beyond knowledge-centric research and education, to restore a practice-based humanities—one where we actively do something, and in doing so, discover fulfillment, self-worth, and meaning in life.
If the crisis of humanities stems from reducing it to mere knowledge acquisition, we must ask the next question:
What is the "practical humanities" that we must restore in the age of AI, and what should it look like?
I believe humanistic practice in the AI era must meet at least two key conditions:
First, it must bring genuine joy to the practitioner.
As Confucius and Jeong Yak-yong emphasized, practice (習) must lead to joy (悅). And what brings us joy? A sense of achievement. The satisfaction of accomplishing something through one's own efforts, and the self-worth created by making something new to oneself. Ultimately, humanistic practice is an experience where human initiative and creative effort are rewarded with joy.
Second, practical action in the AI era must take place within the AI environment itself.
Today, AI is already our environment for research, education, and daily life. Living or researching while completely excluding AI is no longer realistic. Thus, the key question is not whether to use AI, but how to master it according to our own intentions and gain fulfillment through that mastery.
This is what the classical humanities must become if they are to remain sustainable. The spirit is nothing new — it has been pursued since the time of Confucius — but to enact it on the new stage of AI, we need training before we can take the first step. So I designed a training program to help traditional scholars practice within an AI environment, and I gave the activity it aims at a name: "AI Classical Translation Studies." Let me introduce it to you.
AI Classical Translation Studies goes beyond rendering the classics into modern language. It is a practical discipline that reorganizes classical knowledge into a structure AI can understand, so that, through the collaboration of human and AI, the interpretive tradition of the classics can be carried forward and expanded.
Traditional translation produced a result for a human reader. But the age of AI brings a new task: now AI, too, must be able to understand the classics — and for that, a plain translation is not enough. The structure of a text, the meaning of its terms, the relationships among concepts, the context of the document — all of this must be organized into a form AI can process. This is why the discipline places such importance on structured knowledge: XML, semantic data, and ontologies. Through such work, human interpretation and AI interpretation can be compared and can verify one another. In other words, this is at once the work of conveying human interpretation to AI, and the process by which humans review and reflect on AI's interpretation in return.
To experiment with this approach in actual educational and research settings, we implemented a hands-on platform called CCTI (Classical Chinese Text Interpreter).
CCTI utilizes LLM APIs to support the step-by-step interpretation of Classical Chinese texts. Crucially, this system is not a tool designed to generate automated translation results. Rather, it is crafted as an educational and research environment designed to learn, critique, and review the entire interpretative process.
Furthermore, by integrating a Wiki-based collaborative workspace, we extended the traditional classroom reading culture (gangdok) into the digital realm. Users share their interpretation results, compare differing readings, and jointly build better interpretations. While classical education in the past relied heavily on face-to-face learning between master and student, it can now evolve into a broader community learning process online.
In the spring of 2026, after a year of R&D and pilot runs, the Institute of Traditional Culture—a specialized institution for translating Classical Chinese—launched two courses: "Introduction to AI Classical Translation Studies" and "Translation and Curation of Inscriptions on East Asian Painting."
The students in these courses included classical researchers, graduate students, and university professors. Every class was co-taught by a professor of classical humanities (specializing in Classical Chinese literature or art history) alongside a professor of digital humanities.
The coursework combined practical translation with training on how to structure texts into AI-readable data.
In the translation practice sessions, students repeatedly executed a 5-step workflow:
Within CCTI, all data exchanged between users and the AI was strictly formatted in XML. Parallel to the translation process, students received instruction from the digital humanities professor on designing XML schemas tailored to text characteristics and communicating precisely with AI using XML formats.
Additionally, by introducing DBMS operations, database-backed AI APIs, and Wiki platforms, we deepened students' understanding of "data-driven AI collaboration." We firmly believed that teaching data technologies in isolation offers little value to humanists; thus, all technical instruction was integrated directly into the process of interacting with AI using the exact texts researchers were working on.
Through this curriculum, we confirmed several positive outcomes:
The experiment also raised, quite forcefully, a task we have yet to solve. At first the students found the technical side daunting — the AI tools, the XML markup, the Wiki syntax — but as the weeks went on they found it was something "you can do once you learn it." What weighed on them far more was the environment itself: the servers and databases behind data-based collaboration. In our final discussion they asked:
“When this course ends, where can I go on doing this work? Must I research, study, and pay for such an environment entirely by myself? Is that even possible?”
Humanities scholars are used to working alone. But building an AI-collaboration environment cannot be a one-person endeavor. Such an environment must be provided as shared infrastructure — much like a university library. Just as they use CCTI, scholars should be able, with a single login, to reach a server, run a database, call on AI, visualize data, and create Wiki documents.
Establishing such environments is the responsibility of universities and public institutions supporting humanities research.
Today, I am reflecting deeply on how we can construct a Public AI Research Environment for the Humanities.
At the heart of this course is the work of organizing the interpretive tradition of the classics into data, and structuring it into a form AI can understand. When human scholars build data grounded in their own judgment, and use it to steer AI's interpretation toward their own view, I call that act "a humanities that teaches AI."
When you examine AI's interpretation, it looks convincing at first. But the deeper you go, the more doubtful passages appear — and then, suddenly, a judgment arises: "I see this differently." From that moment the scholar seeks out more sources, thinks more deeply, and refines their own reading. And the fruit of that effort remains as data — data that raises the quality of AI's knowledge in turn.
Naturally, building specialized Vertical AI for specific disciplines requires sustained collaboration among teams of scholars, coordinated by institutional leadership. Yet even so, because this quality improvement originates from individual scholarly contributions, data-driven collaboration retains deep meaning as "Humanities that teaches AI."
There is another truth here. In an age when AI has become the very environment of our lives, this work is also training — training on how a human finds and carries out work of their own, neither tossed about by AI nor turning away from it. So humanistic scholarship in collaboration with AI is at once a process of educating AI and a process of educating the human who must live in the age of AI. The old classics of the East captured this mutual growth in a single phrase:
敎學相長 (교학상장, Jiao Xue Xiang Zhang)
"Teaching and learning foster each other."
— The Book of Rites (『禮記』 「學記」)
Education is an interaction, in which teacher and student both grow. And so "a humanities that teaches AI" is not simply the making of data for a machine. It is a process in which we reconstruct our own cultural tradition into a language that can speak to AI, discover the meaning of that culture anew, and train our own capacity for practice. In other words, to teach AI is, in truth, also to train ourselves.
As AI takes over more and more of what we do, our lives will grow more convenient — but we may also lose the will to decide and to act for ourselves. In a world where everything is provided automatically, there is a real danger that a person goes on living, yet can no longer take charge of their own life.
Here a vital question arises for a future humanities: In an age when AI performs so much of human activity, how will the human remain human? I believe the role of a future humanities is the ongoing reinterpretation of what it means to exist as a human being within an AI civilization.
But we must not let this stay at the level of abstract theory — for if it does, it too will be diluted into the very talk that AI itself can generate. To remain human, one must act, in the real world; for human autonomy is not sustained by thought alone, but formed in the experience of deciding and practicing for oneself. And since our reality is already inseparable from AI, that practice must take place in coexistence with AI. A future humanities is neither a discipline that rejects AI nor one that submits to it. It explores how to keep human self-direction alive while walking side by side with AI.
Even when AI can search and generate information in every field, we must still choose for ourselves what to hold important, what to remember, and how to live. And the classics are proven content for training exactly this — our self-reflection, our sense of direction, our spirit of self-direction. The reason the study of the classics must continue in the age of AI is not that it preserves the past, but that it is training for the future: training for humans to remain human.
The reason I shared the case of AI Classical Translation Studies today is not because the experiment is technically novel. It is a proposal to move our concerns about values of classics from abstract theory to practical experimentation, offering a modest first step.
Allow me to conclude my address with a request to the global digital humanities community gathered here today.
It is my earnest hope that digital humanities will move closer to the foundational texts and traditions of traditional humanities, beyond technological innovation alone.
This does not mean all digital humanists must become traditional philologists. Rather, it means that opening pathways for traditional research and education to enter digital environments naturally is another core mission of digital humanities.
Digital humanists are those among humanists who have traveled closest to the digital frontier.
I hope that digital humanities will take an active interest in this role:
I deeply hope that all of you will bring your wisdom and leadership to this vital role as mediators of transformation, connecting traditional humanities with the AI world.
Thank you very much.