Abstract
Integrating formative practice questions with text content is a highly effective learning method. Millions of AI-generated formative practice questions, embedded in thousands of publisher e-textbooks, are now available to students in higher education. This paper reviews findings from a multi-year research program to synthesize performance benchmarks for automatically generated questions and feedback derived from large-scale student interaction data. In addition, we report classroom-based applications that demonstrate how these questions can support learning when integrated into instruction. A central contribution of this review is to identify barriers to effectively scaling student engagement with formative practice, identifying both the successes of automatic question generation systems and the persistent challenges that must be addressed to maximize their potential for classroom impact.
