Generative AI in higher education: How AI-trained students will shape genomic medicine
Source PublicationScientific Publication
Primary AuthorsRojas, Guerra, Frez et al.
"It is like cooking a meal where you have the ingredients but your friend has the recipe. Neither of you can make dinner alone, so you have to talk constantly, check each other's work, and adjust the heat together to avoid burning the food."
For decades, unlocking the full potential of genomic medicine has presented immense challenges. Scientists struggle to model the complex biology of human genetics. Progress can be terribly slow when dealing with such vast datasets. We desperately need a better way to train the next generation of researchers to solve these massive, global problems.
These results were observed under controlled laboratory conditions, so real-world performance may differ.
This brings us to a fascinating new study about Generative AI in higher education. Researchers wanted to see how university students actually work with artificial intelligence when solving complex problems. In a small, controlled lab study, they asked thirty undergraduates to complete advanced calculus tasks alongside an AI peer. To make teamwork necessary, the researchers used a 'jigsaw' structure. This meant the student had part of the information, and the AI had the rest. They had to talk to each other to find the answer.
The right way to use Generative AI in higher education
The study measured the exact ways students interacted with the machine. It turns out that human-AI teamwork is not a single, uniform behaviour. The researchers identified eight distinct roles. Some students acted as 'Low-Effort Guessers' or 'Early Disengagers'. Others acted as 'Co-Regulators'.
The results were clear. Students who actively shared their thought process and negotiated intermediate steps with the AI were highly successful. In the high-interaction group, 71 percent solved the tasks. In the low-interaction group, only 19 percent succeeded. Those who simply delegated the work failed. Those who built a shared understanding thrived.
From calculus to genomic medicine programmes
What does a maths test have to do with the future of healthcare? Everything. The way we teach students to collaborate with AI today shapes how they will tackle real-world biological crises tomorrow. This research suggests that if we build education platforms requiring deep, structural teamwork with AI, students learn how to direct machine intelligence effectively.
We could soon apply this exact collaborative modelling to training programmes in genomic medicine. Imagine a future researcher and an AI working in a similar jigsaw structure to decode complex genetic disorders. The human provides the biological context, ethical boundaries, and clinical limits. The AI processes millions of genomic sequences and predicts molecular interactions.
They cannot work in isolation. The human must interrogate the AI's logic, just like the successful calculus students did. By externalising their reasoning and co-regulating the research process, future scientists may speed up the development of precision therapies. While this foundational study is limited to undergraduate mathematics, the pedagogical approach suggests a hopeful trajectory. By mastering AI collaboration in university, tomorrow's researchers will be ready to tackle the genomic complexities we struggle with today.