Mapping and enhancing the acquisition of conceptual knowledge using behavior, neural signals, and natural language processing models
Effective Years: 2022-2027
This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2). The project is also funded from the EHR Core Research (ECR) program with co-funding from the Human Networks and Data Science (HNDS) program in SBE.
The goal of this CAREER project at Dartmouth University is to advance our understanding of how students learn STEM concepts through online course videos in order to improve online education. Continued expansion of the internet backbone and improvements in computing hardware have facilitated improvements in video streaming, enabling videos to be more easily downloaded and shared. This, in turn, raises a number of questions of pressing national concern. For example, what makes for an effective course or training program? Which aspects of teaching might be optimized or automated? How and why do learning needs and goals vary across people? How might we lower barriers to achieving a high quality education? The focus of this project is to understand how to provide learners with automatized instruction that is customized to the needs of each individual. It could have a significant impact on the online learning of STEM concepts and how it can be individualized for members of different communities. Moreover, the investigator’s own teaching and mentoring will bring machine learning, cognitive neuroscience, and instructional design together in a way that will provide development opportunities for the next generation of scientists working on frameworks for characterizing and evaluating real-world teaching and learning. He will create summer workshops, a regular tutorial series, and a series of open online courses.
The project aims to (1) to build a computational framework for tracking individual students’ conceptual learning and understanding; and (2) to test whether and how brain recordings can be used to estimate ongoing conceptual learning and understanding. The project will collect data of what students learn from real online instructional videos to inform, fit, and test models of real-world conceptual learning in such STEM domains as Astronomy and Computer Science programming. Conceptual knowledge is tested by asking participants to solve applied problems that require them to generalize beyond the specific examples presented during training. Applying text embedding-based models run on these data, the investigators will derive semantic maps both of the concepts taught and of what the students learn. The project will also involve collecting a large dataset from participants who engage with a sequence of course videos while undergoing neuroimaging. The experimental data will be used to construct dynamic estimates of each participant’s moment-by-moment conceptual knowledge and their ability to acquire new knowledge. The project will use natural language processing models to quantify concepts and how they relate. This work will provide a foundation for later research and development of automatic adaptive learning systems that can assess individual conceptual knowledge and edit online instructional material so that it is tailored to that individual.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.