Prior to coming to EPFL, Tanja Käser was a senior data scientist at the Swiss Data Science Center of ETH Zurich. As she explains, her interdisciplinary research combines computer and education science. She uses artificial intelligence (AI), machine learning, algorithms, and data mining to model and predict human learning and behavior. This facilitates, for example, the customization of learning tools.
IC: Can you tell us briefly about your research background? What inspired you pursue a combination of education and computer science?
TK: I completed my master’s and PhD degrees in computer science at ETH Zurich. My research uses machine learning to understand and improve human learning. I am particularly interested in creating accurate models of human behavior and learning.
I chose this research direction because I am very fascinated by how humans learn, and also because of my desire to have an impact on society with my research; i.e., through providing high-quality education to everyone.
IC: What is your mission for the Digital Vocation, Education, and Training Laboratory?
TK: My vision is to use technology to support the vocational education of students, to help them become better learners. Increasing digitization means that knowledge circles are becoming shorter, which in turn necessitates the adaptation of knowledge and skills.
Digital environments, such as interactive simulations, have the potential to teach students new content, while at the same time allowing them to practice important skills for learning. The use of AI to personalize these learning environments will also enable students to learn content more efficiently, and to develop more effective learning strategies.
IC: In your opinion, what are some of the most interesting problems to be pursued in the field of digital education today?
TK: Gaining a full understanding of student learning, so that we can accurately model it and hence facilitate more effective and efficient teaching. Education is both about students learning material effectively, as well as about preparing them for continuing to learn on their own. Until now, a lot of digital and online education has been focused more on the first part. However, computers also make it easier to observe how students learn by gathering data on their interactions and learning processes, and to therefore gain new insights.
I am also interested in the explainability of model decisions. Today, we have powerful AI models, but they are often a black box. It can be hard to explain to teachers and students how models make decisions. I think this is important for education, so that that people using these models understand how decisions are made.