Jiho Yoon has just become an official PI in IMPRS

A mentor, not a dictionary: Jiho Yoon on mentoring PhD students in the age of AI

Jiho Yoon has just become an official principal investigator (PI) in IMPRS. Two years ago he was finishing his own PhD in the same group. He now leads projects inside Stuart Parkin's department, closely aligned with him. And with the role came the first PhD student he works with day to day: Naye Kwon, who first arrived on an internship, worked hard, and asked to stay.

The role is not new to him. The mentoring has been going on for years. Since 2022 the group has hosted Korean undergraduates every year, one or two at a time, six months each, and Jiho has looked after them the whole time.

Kwon was one of those interns, and she is direct about why she stayed. "The reason why I decided to start the PhD here is that Jiho's leading," she says.

"Education is not simply what you can learn from the classroom," he says. "You can really learn how to conduct research in a proper manner." That part isn't lectured. It's picked up at the bench, beside someone who already does it, one problem at a time. The research he and Kwon work on, atomic layer engineering for next-generation 3D memory devices, has no answer in a textbook. "There is no single answer for that," he says. "Only the good idea could come during the very close interactions with other scientists."

The thing a doctoral student most needs to learn is the one they cannot learn alone.

 

"Don't use your PI as a dictionary"

His first rule for a new student: don't treat me as a dictionary.

"Using your PI as a dictionary or the encyclopedia doesn't work," he says. "You should not really expect someone knows the answer. But at least during the conversation, during the active discussion, you can somehow get insight, and you can even criticize someone's ideas." A mentor, in his telling, is the partner you argue an idea into shape with, and he clearly enjoys that part. "I'm very interested in being a good mentor," he says, "to be the partner for the next generation of scientists."

He keeps "leading me to ask a lot of questions," says Kwon, and "whenever I'm confused, or I have some questions, then he's always open to have a discussions."

He also cares about what a student asks. "The question should be much longer than simply 'why,'" he says. His preferred shape: first say what you saw and measured, then what you understood and expected, and only then ask why. A question that arrives one second after the data, he notes, is not the same as one that arrives after a day of thinking, and he can hear the difference.

 

The pocket dictionary

A mentor was never the one with all the answers, and that is easier to see now that everyone carries a machine that hands you any answer you ask for, instantly and confidently, whether or not the question was any good.

He isn't dismissive of the tools. His worry is more precise. "Even if you ask any question, they're going to give you very nice-looking answers," he says. "But if you don't exactly know what you're asking for, then you will really mess up with this AI." He ran a small test on himself. He asked ChatGPT and Gemini about one of his own experiments. "They always deliver you a very nice answer. But none of them was really matching with what I actually saw in my work."

For him, that mismatch is the whole point of the job. "No one knows the result. We are building up this knowledge," he says. "You cannot write the future. You can only see the future when you do something and make the right questions." A machine can hand you an answer. It can't hand you the question worth asking, or notice when a mistake "drives your results in a very interesting direction," which he calls the most charming part of his work.

 

Quality over hours

The habit he wants students to keep is about time. "No one can run the marathon at a very intense pace for two or three years," he says. "That's human nature." He won't push long days. Eight to nine hours of genuine critical thinking, he reckons, already leaves you exhausted. Quantity is the one thing AI can produce endlessly, and the value a person adds is quality, which doesn't come from a twenty-hour day.

"Many brilliant ideas come out when you are doing something else," he says, "like walking, or taking a shower, or swimming. When you are doing science, you cannot simply turn off and turn on your brain." Everyone has the same twenty-four hours; the skill is using them well.

"He always said that it depends on the personality, and he recommend me to not to overwork in the lab. So I try not to stay like over six at the clean room," Naye Kwon says.

 

The IMPRS

None of this rests on one mentor. It's the reason a doctoral school like IMPRS matters, and Jiho is candid that he came up without it. "I did my PhD in the same group here, but I was not in the IMPRS. When I first joined, we didn't have it yet."

He sees plainly what it adds. A Max Planck Institute is not a university and cannot award a degree itself. IMPRS gives students visibility that they did their PhD and research here, "and I think this can make them more proud." Around that sits the support he values most for the daily work: an IMPRS seminar most weeks, a retreat, coordinators who look after the students. "All this kind of environment makes the PhD students become more like one spirit," he says. "It not only binds them to each other, it also brings a good synergy on their research."

It also helps that they're doing it somewhere liveable. Halle is affordable, "everything is really affordable for young people," with good nature nearby and Berlin an hour away by train. The labs, he says, are ones "you cannot find anywhere in the world," where you can pull "very tough-level scientists" into an unconditional discussion. "This is kind of really a paradise for research."

He is patient with hard problems, "I have no problem to solve these problems," and what he wants to pass on is a way of working: ask the longer question, argue it out with people who'll push back, and do not try to do it alone.