Research
We have the goal of understanding the principles of intelligence and decision-making in artificial and natural agents. We study human learning and decision-making in complex and realistic environments to extract the formal principles of behavior and come up with strategies to improve decision-making. We study and improve machine learning algorithms using insights from cognitive science. We particularly focus on three areas of research:
Understanding and assisting human behavior
One of our main research areas is computational cognitive science. We study how people think, learn, and decide using methods from psychology, data science, machine learning, and neuroscience. Our goal is to formally understand the building blocks of human intelligence. Here, we focus primarily on how humans solve complex tasks, e.g., in video games or naturalistic experimental scenarios (e.g., ordering food or making other health-related decisions), in an attempt to distill the principles of intelligence into formal algorithms that can help us understand human behavior.
Understanding and improving artificial agents
With the advent of powerful artificial agents such as GPT comes the urgency to understand how they learn and make decisions. These foundation models are not only incredibly large, but they are also extremely opaque. We, therefore, study these algorithms using tools from cognitive science in the hope to understand better how they behave. Moreover, we use insights from psychology and neuroscience to improve current deep learning algorithms in general, for example, by making them more sample-efficient.
Computational mental health
In the third area of research, we use computational models to understand and improve people’s mental health. The nascent field of “computational psychiatry” considers the interactions between individuals and evolutionary, developmental, and current environments that collectively define mental illness. It then attempts to use these to provide insights into the nosology, prognoses, and possible cures for some of the most common mental disorders.