Impacts of Symbolic Regression-Integrated Physics Classes on Students' Perceptions of Using Machine Learning in Science

This study explores the design and implementation of symbolic regression (SR)-integrated physics classes (SRPC) to encourage students to actively engage in scientific discovery using real-world data. The classes, involving 18 second-year high school students in South Korea, were conducted over six hours across two weeks. Students completed a Technology Acceptance Model (TAM)-based survey before and after the classes, and semi-structured interviews were conducted with five students. Analysis of both survey and pre-interview data revealed that, despite a limited understanding of machine learning (ML) principles, students displayed a positive attitude towards its use in scientific research and expressed a behavioral intention to use it. However, post-interviews indicated a shift in students' perceptions towards a more critical awareness of ML's limitations, underscoring the importance of understanding underlying principles. While students responded positively to the data-driven research experience, this response was limited to the novelty effect of the new scientific approach. Additionally, the findings suggest that employing a more diverse range of phenomena in SRPC, where universal laws are not yet established, may better support genuine scientific discovery using SR.