Bloom-Aligned Question Generation and Evaluation: A Systematic Literature Review

Bloom-aligned automatic question generation is an emerging research area in the educational domain. It enables the assessment of higher-order and more complex cognitive skills beyond simple recall and understanding. With the advancements in NLP generation models, studies have adopted different approaches and techniques for generating and evaluating educational questions. Therefore, this systematic review aims to explore recent studies in this field in terms of applied generation techniques and evaluation methods. A thematic analysis was also conducted to develop an evaluation taxonomy. A systematic search was conducted for studies published from 2021 to 2026; however, only studies from 2023 to 2026 met the predefined inclusion criteria. The results show that recent studies used prompt-based methods more commonly than the fine-tuning approach and applied zero-shot prompting more frequently than few-shot prompting. The findings also indicate that experts conducted human evaluations more frequently than automatic evaluations. The proposed evaluation taxonomy, informed by feedback from two experts in NLP in education, consists of three main evaluation dimensions: educational quality, linguistic quality, and Bloom evaluation. The reviewed studies still struggle with generating higher-order questions, the lack of Bloom-based datasets, and the absence of a unified evaluation rubric. The findings of this review provide insights into current generation and evaluation techniques and highlight key gaps and challenges that should be addressed in future research.