ChatGPT for Next Generation Science Learning

Abstract

The K-12 Framework for Science Education set forth a vision for Next Generation Science Learning, which is to engage students in scientific practices to learn and use disciplinary core ideas and crosscutting concepts. This vision posits significant challenges in terms of how to track students' learning, provide feedback and learning guidance, recommend learning materials, and meet the special needs of students with diverse backgrounds. This study pilots ChatGPT in meeting these challenges. I used one performance expectation of the Next Generation Science Standards to develop a prompt, using which ChatGPT automatically generates a performance-based assessment task. I supplied a response and asked ChatGPT to grade and provide feedback. I then asked ChatGPT to provide learning guidance and learning materials based on the response. At last, I told ChatGPT that the learner was with dyslexia, and eventually ChatGPT recommended specific learning materials for the learner. Results suggest that ChatGPT has the potential to tackle the most challenging problems of science learning through automatic assessment development, automatic grading, automatic learning guidance, and automatic recommendation of learning materials. Even with exciting findings, this study suggests that ChatGPT cannot substitute teachers. Teachers need professional knowledge to use ChatGPT for instructional purposes. Further, the BlackBox of ChatGPT, in terms of how it generates automatic results, needs to be explainable so that students, teachers, and parents can fully appreciate its merit.

Keywords: ChatGPT, Education, Science learning, Assessment, Artificial Intelligence, Natural Language Processing