Automatic Question Generation

Automatic question generation allows us to scale the Doer Effect.

Using artificial intelligence methods, we can use textbook content to generate questions, and these questions can be created for textbooks nearly instantaneously to provide students with the learning by doing practice they need. However, it’s crucial to know that these questions are good. Our award winning research analyzed engagement, difficulty, and persistence of automatically generated questions and human authored questions answered by the same students in the same courseware. This research has shown:

Students do not treat automatically generated questions differently than human authored questions.
The differences in engagement, difficulty, and persistence were actually related to the cognitive process dimension of the question (recognition versus recall), not how the questions were created.
Easy questions did not deter students from answering, but very difficult questions did decrease student persistence in answering.

Recent Publications About Automatic Question Generation

Search our catalog of recent publications authored by our team.

2026

LLM feedback isn't automatically better: Static scaffolds outperform dynamic feedback in textbook-embedded practice

Seventh International Workshop on Intelligent Textbooks

Benny G. Johnson, Jeffrey S. Dittel, Oscar J. Ortiz, Rodrigo Bistolfi, Michelle W. Clark, Bill Jerome, Richard Benton, and Rachel Van Campenhout (2026)

2026

Extending an automatic question generation pipeline with LLM-based free-response tasks: An analysis of performance metrics using student data

CSEDU 2026 – CSEDU 18th International Conference on Computer Supported Education

Rachel Van Campenhout, Jeffrey S. Dittel, Bill Jerome, Benny G. Johnson (2026)

Automatic Question Generation Generative AI Teaching & Learning

See all publications