Tongmyong University's Greg Kang argues that higher education must move beyond producing good employees and teach students how to think, act and adapt.
By Eugenia Lim
19 August 2026
A former colleague of mine, a professor of literature who had taught for 30 years, told me last year that she had started feeling useless. Her students could get better literary analysis from an AI chatbot than from her lectures, she said. She’s not wrong about the analysis. But she is wrong about what makes her valuable, and that confusion is at the heart of the challenge facing universities today.
The question “what should universities teach in the age of AI?” is being asked everywhere in higher education right now, and mostly being answered badly. The most common answers oscillate between two failure modes: either doubling down on exactly what universities have always done, on the grounds that the fundamentals haven’t changed, or pivoting wholesale toward AI literacy and technical skills, on the grounds that everything has changed. Both are wrong. Neither grapples seriously with what AI actually changes about the relationship between knowledge, skill and human development.
Let me try to be more precise. AI systems are already better than most humans at a specific and expanding range of tasks: retrieving and synthesising information, pattern recognition across large datasets, generating plausible text and images, translating between languages, performing certain kinds of logical analysis. Within the next few years, this range will expand considerably. The question is not whether universities should teach the content that AI can reproduce. It is whether universities are adding value in ways that AI cannot replicate.
The honest answer is: sometimes yes, often no.
What AI cannot do, at least not yet and perhaps not ever in the same way, is engage in genuine relationship. It cannot provide the experience of being challenged by a person who knows you, who sees your potential and who holds you accountable to it. It cannot replicate the intellectual community of people grappling with difficult ideas together in real time. It cannot provide the embodied experience of being in a place dedicated to learning, with all the social and cognitive effects that produces. These are the things that justify a university education, not the transmission of information.
This means that universities need to rethink what their lectures, seminars, tutorials and assessments are actually for. If a lecture is primarily a mechanism for transmitting information, it is now in direct competition with AI, and it will lose. But if a lecture is a performance of intellectual engagement by a scholar who has spent decades developing a perspective on a problem, who can model how an expert thinks and feels and why it matters, then it remains irreplaceable. The question is whether lecturers are doing the first thing or the second, and whether universities are creating the conditions for the second to happen.
Assessments are the most urgent problem. Most university assessments still measure the things that AI does well: information retrieval, synthesis, structured argument. An essay question that asks a student to summarise the key debates in a field and argue for a position is something that a competent AI can answer better than a competent undergraduate. Universities that continue to set these assessments are not only failing to develop their students, they are also failing to learn anything useful about whether their students are developing.
Assessment reform needs to focus on what only humans can demonstrate: judgment in context, the ability to respond to unexpected situations, the capacity to integrate knowledge with personal experience and ethical commitment. This points toward more oral examination, more live performance, more project-based assessment with real-world stakeholders, more portfolio approaches that document development over time. It also points toward being more honest about what grades actually measure and whether the things they measure are the things that matter.
The curriculum question is harder. I am sceptical of the rush to integrate AI literacy as a discrete subject, as though teaching students to use ChatGPT is the equivalent of teaching them to read. AI tools will change faster than any curriculum can track, and the skill of using them effectively is not really a skill that needs teaching so much as a disposition that needs cultivating: curiosity, experimentation, critical evaluation, awareness of limitations. These are the dispositions that a good liberal education has always tried to develop.
What does need explicit attention in the curriculum is the ethics and governance of AI, not as a technology topic but as a humanities topic. Students in every discipline need to understand how AI systems are built and what values they embed, how they can be used to concentrate or distribute power, and what responsibilities come with deploying them in consequential contexts. This is not a module to be bolted on to existing programmes. It is a perspective that needs to be woven through the curriculum in the same way that environmental or gender perspectives have been woven through in previous generations.
I keep coming back to my former colleague, the literature professor. What she does that AI cannot do is not literary analysis. It is the cultivation of a relationship with a text, developed over decades of reading and living, that she then uses to challenge her students to develop their own. The question “what should universities teach in the age of AI?” has a simple answer and a hard one. The simple answer is: the things AI cannot teach. The hard answer is: figuring out what those things are and reorganising everything around them.
Eugenia Lim is a writer with over 10 years of experience in Singapore's broadcast industry. She is also a producer for Channel NewsAsia and is based in Seoul, South Korea.
.jpg)
