AI-based diagnosis of student reasoning patterns in NGSS assessments

Abstract

Scientific reasoning engages students in developing and testing hypotheses about how to make sense of natural phenomena. In this chapter, we describe a project exploring how a machine-learning approach can be utilized to identify student reasoning patterns related to the three dimensions of the Next Generation Science Standards (NGSS). We discuss the importance of identifying student reasoning patterns in the context of NGSS assessments and describe a framework for identifying those patterns in the context of one ecosystem item. We used student responses to this item to develop and validate an AI tool to automate the classification of reasoning patterns and provide feedback to students. We report the results of a cognitive interview study with middle and high school students to evaluate the AI tool. We conclude the chapter by discussing validity issues related to the AI-based classification of student reasoning patterns, potential uses and limitations, and future research.