Project Background

When learners open up a chemistry textbook or exam filled with paragraphs, data tables, graphs, and diagrams of different colors and spatial orientations, an eye-tracking system can detect visual features engaged with this viewing data, creating an opportunity to support viewing strategies through visual feedback. Analyzing where and how study participants look at a page to tackle a chemistry problem illuminates how individuals more or less effectively approach and persevere through challenging content. Since the Fall of 2023, the RevChem Collaborative Research project led by Professor Sarah Hansen from Columbia University and Professors Katherine Havanki and Matthew Jacobs from The Catholic University of America has delved into the ways eye-tracking data can influence personalized educational methods to enhance problem-solving capabilities. 

Problem solving depends on the intake of information before the mental work of processing and calculating begins. However, if an individual’s gaze skips a critical piece of information prior to the processing stage, their calculating effort will be in vain. The RevChem team is specifically collecting data to inform an Artificially Intelligent tutor that will cue problem solvers about how to focus on specific elements of a challenging question upon first exposure.

Methodology

This summer, I will be supporting RevChem in the final phase of the research project by developing a script for researchers and a formulaic approach for walking participants through the human-subject research. I will also be conducting the research using a monitor-based eye-tracking system while curating and evaluating prompts to promote problem-solving in chemistry learners that will inform a Machine Learning (ML) tutor. 

Research Questions and Objectives

(1) How does how people believe they solve problems differ from how they actually solve them? 

Much of the time, as research from the former phases of the project has uncovered, there is a dissonance between how a participant believes they approach a difficult chemistry problem and the way their eye actually moves across the page. By revealing this distinction to the participant, they can adjust their method and regain agency in guiding their own problem solving.  

(2) How do web-based and monitor-based eye tracking differ, and how can the different types of data collected inform Machine Learning?

An ML tutor may cue a participant's focus when approaching a chemistry problem, but it is important to ensure that the training data collected from the monitor-based system will support students using more accessible web-based systems to the same extent. This analysis ensures that the scope and sustainability of the research are maximized.