Assessing and guiding student science learning with pedagogically informed natural language processing

Abstract

Natural language processing (NLP) tools can score students' written explanations, opening new opportunities for science education. Optimally, these scores offer designers opportunities to align guidance with tested pedagogical frameworks and to investigate alternative ways to personalize instruction. We report on research, informed by the knowledge integration (KI) pedagogical framework, using online authorable and customizable environments (ACEs), to promote a deep understanding of complex scientific topics. We study how to personalize guidance to enable students to make productive revisions to written explanations during instruction, where they conduct investigations with models, simulations, hands-on activities, and other materials. We describe how we iteratively refined our assessments and guidance to support students in revising their scientific explanations. We report on recent investigations of hybrid models of personalized guidance that combine NLP scoring with opportunities for teachers to continue the conversation.