Situating AI (and Big Data) in the Learning Sciences: Moving toward Large-Scale Learning Sciences

Abstract

Interdisciplinary research in the learning sciences incorporates diverse constructs, measures, and processes with the goal of advancing theories of learning and to inform the design and implementation of effective instructional methods and learning technologies. The complexity of these goals is further influenced by a large array of factors stemming from the learning context, learning task, and the characteristics of the individual learners. Yet, traditional approaches to research may not explain these factors such that we can truly understand learning in context. We argue that artificial intelligence (AI) and big data can be leveraged to implement large-scale testing of interventions with diverse learners in multiple contexts (i.e., Large-Scale Learning Sciences). AI and machine learning have altered the landscape of instruction as it has shifted from solely classroom-based to a variety of approaches that range from some technological enhancement (e.g., hybrid) to completely online (e.g., synchronous, self-paced asynchronous). We discuss how AI (and big data) can be used to advance the learning sciences in five areas: Deep student models, causal learning outcome models, natural language processing (NLP), sensor-free student factor measures, and instructional policy learning. All of these areas have the potential to examine what works best for whom and in what conditions. In turn, by combining AI and the learning sciences, scientists may begin to address the inequalities and inequities present in educational research with access to a diverse array of students, contexts, and interventions. Merging these domains and the knowledge and techniques therein is crucial to understanding learning and improving learning outcomes for all learners.