Collaborative Research: Using Machine Learning to Improve Visual Problem-Solving in Chemistry Education

This project aims to serve the national interest by increasing student success and confidence in solving chemistry problems, which is of vital importance to encouraging students to pursue STEM degrees. Chemistry learning materials are filled with a variety of visual information (chemical symbols, mathematical representations, graphical information and text) that students must understand and process in order to solve problems. Knowing where and how to look at this visual information is critical to forming new ideas, recalling required knowledge, and performing the steps necessary for problem solving. Machine learning will be utilized to investigate the link between where a student looks and their resulting success in solving a problem by tracking their viewing behavior with a novel eye-tracking system. The resulting software will support students during the learning process by providing real-time feedback if they are not viewing relevant features of a problem. The project team aims to expand to additional STEM disciplines once the benefits in chemistry education has been demonstrated. As a result, this project has the potential to provide a more accessible and high-tech method for carrying-out STEM education research while also having a substantial impact on student learning.This cross-institution collaboration aims to engage in foundational work to create an intelligent tutoring system that uses webcam eye-tracking data and artificial intelligence to provide chemistry learners with real-time feedback during problem solving activities. The study will be executed in three phases: 1) traditional screen-based eye tracking methods will be compared to webcam-based eye tracking methods, 2) data collected in phase one will be used to train and evaluate a machine learning model to predict student outcomes, and 3) the team will investigate the ability of early cues to change the viewing patterns of students actively engaged in problem solving. Through the systematic development of a machine learning model that uses eye tracking to predict achievement outcomes of problem solving in chemistry, the development of a feedback model for early intervention is possible. This individualized feedback has the potential to support students to engage with problem solving in more productive ways, build confidence, and provide valuable insights into how students solve problems. By providing immediate, individualized feedback, this research has the potential to support a large number of students in a chemistry problem solving context, including those who may not exhibit traditional help-seeking behavior. Further, comparing web-cam eye tracking to traditional monitor-based approaches has the potential to provide the chemistry education community with a valuable new tool for engaging in research. Findings from this project will be shared through conferences and publications for undergraduate STEM educators as well as to specific communities interested in eye-tracking methods. The NSF IUSE: EDU Program supports research and development projects to improve the effectiveness of STEM education for all students. Through the Engaged Student Learning track, the program supports the creation, exploration, and implementation of promising practices and tools.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.