Automatically Assess Elementary Students’ Hand-Drawn Scientific Models Using Deep Learning of Artificial Intelligence
Tingting Li, Feng Liu, Joseph S. Krajcik · Proceedings. · 2023
Because of the complexity of scoring open-end tasks, machine learning (ML) has been utilized for automatically assessing students' constructed responses.However, most existing research focuses on grading text-based responses.No studies have investigated the automatic scoring of hand-drawn models created by elementary students.In this study, we applied ML to automatically score hand-drawn scientific models developed by elementary students for evaluating knowledge-in-use.We first developed algorithms using human-scored responses and then validated these algorithms with new data.We also implemented a data augmentation technique to enhance accuracy.Our findings demonstrate the potential of the developed algorithm to achieve high performance in automatically scoring hand-drawn models.