Cognitive Learning in Humans and Machines: Insights from Psychology and AI

Speaker 1: Maria Ioanna Magkouta

How do humans learn?

For centuries, learning was confined to classrooms that looked much like this one; rows of students passively listening to a teacher and attempting to absorb knowledge. In a lot of cases, this is still the norm. But is this the most effective way to learn? With advances in neuroscience, cognitive science, and artificial intelligence, our understanding of learning has evolved dramatically.

In this talk, we will explore what it truly means to learn, exploring the cognitive mechanisms that shape knowledge: memory, attention, pattern recognition, and metacognition. Pushing beyond, we will also answer pressing questions: How do we transfer knowledge across domains? How do we learn to learn and what motivates us to persist in problem-solving?

Contrasting traditional education with more active, dynamic, adaptive approaches, we will discuss how learning can be enhanced through productive failure, embodied activities, and what role technology should play.

In an age where information is vast and easily accessible, should education still focus on knowledge acquisition, or should we rethink what the focus of education should be? To prepare students for the future, we need to rethink how we design learning experiences.

Join us as we explore this urgent question: What is the future of learning, and where should our focus be in an era of intelligent machines?

Marianna Magkouta is a doctoral student at the Professorship for Learning Sciences and Higher Education. She holds a master’s degree in Learning Sciences from Université Paris Cité, where her thesis focused on optimizing the integration of computer simulations into the learning process for understanding physics concepts. Marianna is passionate about advancing the understanding of how embodied experiences can enhance learning, aiming to bridge theory and practice in innovative ways.

Speaker 2: Mirlan Karimov

How do machines learn?

In this talk, we will learn about the underlying mechanisms of two machine learning paradigms, discriminative and generative modelling. We will explore what these models learn and whether they learn a converging representation of the underlying reality. Finally, we will identify the current gaps and challenges in creating truly intelligent systems.

Mirlan Karimov is currently pursuing a PhD in Computer Science at ETH Zurich. His research focuses on world models and video generation, which he applies to the development of data-driven simulators for training and testing autonomous driving systems at Mercedes-Benz.