A model for estimating the visual complexity of a molecule using graph theory metrics: an educational perspective

Abstract

Introduction. Research on understanding the visual complexity of molecules is critical to enhancing chemistry education, improving teaching methods, and creating educational resources in STEM education. One effective strategy for improving student achievement involves assessing and incorporating the complexity of learning materials into lesson planning. In chemistry education, various tools have been proposed for this purpose, such as analyzing illustrative molecules using graph theory or expert assessments by chemists. However, these tools remain under-researched, which limits a complete understanding of chemistry teaching methodology in school and university. The aim of this study was to investigate the visual perception of molecular complexity of high school and undergraduate students (ages 16 to 19 years), and to find out how it correlates with graph theory-based estimates of molecular complexity. In the experiment, learners rated the visual complexity of a set of molecules randomly selected from school textbooks. Participants and research methods. Fifty-six learners divided into two groups (for training and testing the ML model), 42 and 14, participated in the study. The groups included first-year biotechnology students from ITMO University (St. Petersburg, Russian Federation) and students from St. Petersburg high schools (grades 9-11). Participants' ages ranged from 16 to 19 years (M = 17.5; SD = 2.96). Methods: source analysis, design, pedagogical experiment, self-assessment of pupils and students, specialized ML methods such as random forest method and linear regression, graph-theoretical methods, statistical. Results of the study. The collected data showed that the visual complexity rated by students was positively correlated with the complexity score obtained using graph theory (r = [0.59-0.84]). Based on these data and an ad hoc survey of students, a machine learning (ML) model was built to produce a complexity estimate for any set of molecules. Practical significance. The resulting model is an open-source toolkit that can be used in chemistry classes to tailor individual assignments or to develop adaptive methods for testing chemistry knowledge. This model can be useful for trainees, beginning teachers and teacher trainees.