Abstract
The ubiquitous use of digital learning resources like etextbooks has shifted the learning experience in higher education. Digital learning has led to both new learning tools as well as research in how learning works via the large, rich data sets those digital resources generate. Advances in artificial intelligence have made it possible to develop and scale learning methods—such as formative practice integrated in etextbook content in a learning by doing approach. A primary benefit of generating formative practice is to bring this highly effective learning approach to millions of students using digital textbooks. This paper focuses on an automatic question generation system that has proven to generate effective formative practice for higher education textbooks, as measured through large-scale analyses of question performance metrics and in-classroom implementations. Open education resources (OER), such as OpenStax, offer students and faculty a learning resource without the high cost. In this paper, we evaluate performance metrics such as difficulty and persistence for the automatically generated questions added to an OER textbook for the first time. Used in several large, online chemistry courses at a major public university, this paper showcases the viability of automatically generated questions combined with OER content for increasing the access and affordability of formative practice as a feature in digital textbooks. Key question performance metrics, such as question difficulty and persistence, along with student interaction patterns and behavior, are analyzed, and future applications of OER content with automatically generated questions are discussed.
