Understanding Human Learning – with AI

Marcel Binz has launched a new research group at the Max Planck Institute for Biological Cybernetics

• New research group leader: Cognitive scientist Marcel Binz has been building the “Core Cognition” group since October 2026.
• AI as a research tool: Language models serve as a testbed for theories of human thinking and learning.
• Science as a team sport: an international team and a focus on selected, ambitious projects

With Marcel Binz, the Max Planck Institute for Biological Cybernetics has gained a new research group leader as of October 2026. Binz is fascinated by how people learn, and by how machine learning methods can help model and explain that process. To pursue this, he is building the “Core Cognition” group, which uses computational models to work toward a unified theory of human thinking.

Binz’s background is broad: he started out in mathematics, but his interest in applications led him to switch to cognitive science. He appreciated the great breadth of this field – from its mathematical and computational foundations to their application to human thinking – as well as the freedom to set his own priorities. It was there that he first came into contact with machine learning and wanted to understand the technical foundations behind it. Since only a few places in Europe offered a master’s program specializing in it at the time, he moved to Stockholm in 2015.

His subsequent move to a PhD in psychology at the University of Marburg may look like a change of direction, but to him it was a natural next step: “However different my stations look on paper, my goal has remained constant,” he says. “I’ve always developed models meant to capture human behavior and, above all, human learning.”

ChatGPT and its consequences

In Marburg, Binz took up meta-learning as a way to model human learning in realistic everyday situations. In this approach, AI models are trained to understand new tasks on their own and adapt to them. In other words, they learn how to learn and can adjust quickly to new situations, much like people do.

Binz’s research received a decisive boost during his first postdoc in Eric Schulz’s group at the MPI for Biological Cybernetics: with newly released language models such as ChatGPT, there were suddenly systems available that could handle a wide range of different tasks. “It was mind-blowing,” he recalls. “You could ask the models anything, and they’d respond with interesting behavior.” At first, he had the AIs solve the same tasks as human participants and compared their responses. He then tried to make them more human-like by training them on large datasets from behavioral experiments.

Binz is now taking the next step: using the models not only to predict human behavior but also to test theories of thinking and learning in a targeted way. “In the end, we want to gain insights that are useful for the behavioral sciences,” he says.

 

Science as a team sport

It is the first time Binz is leading his own research group at the Max Planck Institute, but the task doesn’t daunt him: independent projects and the supervision of PhD and master’s theses were already part of his day-to-day work in his previous role at the Institute for Human-Centered AI at Helmholtz Munich.

Binz also puts great emphasis on collaboration. “Science is a team sport,” he stresses. His team currently comprises four members from four different countries, bringing together expertise in neuroscience, computational linguistics, and machine learning. This interdisciplinary mix is designed to foster a diversity of perspectives and approaches.

The local research environment is a major draw for Binz. “Tübingen packs one of Europe’s hotspots for machine learning and cognitive neuroscience into a small area,” he says. The city also offers him plenty of opportunities to unwind: Binz enjoys hiking, climbing, and board games with friends. Between that and his research, he adds with a smile, his calendar is already pretty full.