High-Order Question Generation in a Multilingual Educational Context
Merged summary
TL;DR — This study evaluates how well LLMs generate "high-order" (critical-thinking) questions across Basque, Spanish, and English using two alternative prompting frameworks, showing prompt design and evaluation matter beyond the usual English/Bloom's Taxonomy setup.
- Introduces prompts grounded in two non-Bloom frameworks—Claim-Evidence-Reasoning and Divergent Questioning—applied in a multilingual (Basque, Spanish, English) educational context.
- Both an open-source and a proprietary LLM generated answerable questions fairly effectively across all three languages.
- However, teachers judged only about half of the answerable questions as genuinely high-order, indicating a quality/evaluation gap.
- The alternative frameworks produced structurally and conceptually varied questions, suggesting they can complement each other and serve as viable alternatives to Bloom's Taxonomy.
Sources (1)
High-Order Question Generation in a Multilingual Educational Context
TL;DR — This study evaluates how well LLMs generate "high-order" (critical-thinking) questions across Basque, Spanish, and English using two alternative prompting frameworks, showing prompt design and evaluation matter beyond the usual English/Bloom's Taxonomy setup.
- Introduces prompts grounded in two non-Bloom frameworks—Claim-Evidence-Reasoning and Divergent Questioning—applied in a multilingual (Basque, Spanish, English) educational context.
- Both an open-source and a proprietary LLM generated answerable questions fairly effectively across all three languages.
- However, teachers judged only about half of the answerable questions as genuinely high-order, indicating a quality/evaluation gap.
- The alternative frameworks produced structurally and conceptually varied questions, suggesting they can complement each other and serve as viable alternatives to Bloom's Taxonomy.