Deep human learning requires meaningful effort and engagement with a topic. However, AI-powered tools are often designed for user retention over meaningful engagement, at the risk of impairing learning and critical thinking abilities. In this project, we explore how friction can be integrated in these tools to support deeper learning.
Deep human learning requires meaningful effort and engagement with a topic. Typically, this happens in two steps: First, a moment of difficulty or friction, then a moment of resolution. In contrast, current AI-powered tools are often designed for user retention over meaningful engagement. Thus, they offer a friction-less, sometimes overly flattering experience, unfortunately only supporting superficial human learning and impairing critical thinking abilities.
In this project, we explore what kinds of friction can be integrated to create effective learning experiences using Generative AI. Looking at this issue in a holistic manner, we consider both adjustments to the LLM behavior itself, but also the interface in to interact with this LLM. As such, we explore, implement, and evaluate AI-powered learning tools including different types of friction to support deep learning of mathematics while sustaining motivation, critical thinking abilities and improving learners’ learning processes.
