Abstract
Textbooks have been the traditional method of providing learning content to students for decades, and therefore have become the standard in high quality content. Yet the static textbook format is unable to take advantage of the cognitive and learning science research on effective interactive learning methods. This gap between quality content and highly efficient methods of learning can be closed with advances in artificial intelligence. This paper will contextualize the need for improving textbooks as a learning resource using research-based cognitive and learning science methods, and describe a process by which artificial intelligence transforms textbooks into more effective online learning environments. The goal of this paper is to evaluate textbook-based automatic question generation using student data from a variety of natural learning environments. We believe this analysis, based on 786,242 total observations of student-question interactions, is the largest evaluation of automatically generated questions using performance metrics and student data from natural learning contexts known to date, and will provide valuable insights into how automatic question generation can continue to enhance content. The implications for this integration of textbook content and learning science for effective learning at scale will be discussed.
