Toward the Ideal Tutor: Active Learning, Motivation, and the Promise (and Limits) of AI in Education
This exciting TeLEd event will feature Louis Deslauriers, who is Director of Science Teaching and Learning and Senior Preceptor in Physics at Harvard University (USA). Louis is the lead author of some of the best known and respected articles on active learning in science education. He is also an atomic physicist and all-around great guy. We have had some amazing speakers for our TeLEd events, including some of the biggest names in our field, and Louis ranks up there among the biggest. He was here in Bergen a few weeks ago and gave a wonderful talk. If you were lucky enough to see that one, this one will complement it nicely. If you weren’t, you still have an opportunity of rectifying that.
How close can we get to the learning gains of expert one‑to‑one coaching in university courses? One-to-one coaching — combining deliberate practice, instant feedback, and affective support — is widely recognized as the gold standard for learning. Drawing on our Science 2011 study, I will illustrate what high‑learning‑density active learning (many decisions and feedback cycles per unit time) looks like and why it is so effective. I will then discuss findings from our PNAS 2019paper, on the illusion of learning: when instruction is more effortful and thus, perceived as disfluent, students often feel they learn less even as their performance improves—an instance of fluency bias that favors polished lectures over productive struggle and can negatively distort student course evaluations. I will also share new results from two‑stage exams—now common in North America—where students complete an individual exam and immediately retake it in groups for partial credit (e.g., 20%). In a randomized comparison of post-exam feedback conditions, students who received group exam feedback scored ~10% higher on later isomorphic questions than those who received instructor feedback, even though students in the group-exam condition spent much of the time debating errors (roughly 40% of the time)—showing the cognitive value of explicit error engagement, a hallmark of deliberate practice that is rarely taught in higher education. Finally, I will examine the role of AI in education, highlighting both promising applications that enhance engagement and the critical limitations that prevent current systems from approximating true tutors, given their limited capacity to capture the nuances of student thinking.