RI: Small: CompCog: Pique: A Cognitive Model of Curiosity for Personalizing Sequences of Learning Resources

One of the most significant challenges in education is to simultaneously provide personalization and scale. How can each learner in an online class of hundreds or thousands be provided with knowledge and challenges that suit them personally? This project will develop an AI system for personalized learning that is inspired by cognitive models of curiosity, creativity and intrinsic motivation. Pique (short for the 'Personalized Curiosity Engine') is based on understanding what makes an individual learner curious, and then recommending resources that will stimulate their curiosity. Pique's cognitive model of its learner uses natural language processing techniques to figure out what sequences of resources will be familiar enough to be accessible, but sufficiently new to not be boring.

Pique is a novel cognitive system drawing on technologies from intelligent tutoring systems, computational creativity, and natural language processing. Its key contribution is combining a cognitive model of curiosity with educational recommender systems. We will evaluate the effectiveness of Pique first with simulations and then with students at a large comprehensive public university. Evaluation will take the form of a comparison between the full Pique system and a modified version with its cognitive model of curiosity disabled. This will enable us to determine whether recommending resources that are simultaneously curiosity-stimulating and fit to the task is more effective than recommending resources that are just fit to the task. Given the interdisciplinary nature of this research we will disseminate our results broadly, including to the educational technology, cognitive systems, information retrieval, and computational creativity communities.