Scholarlink

Background <p>Bloom’s revised taxonomy is widely used to evaluate the cognitive level of examination questions. With the growing use of large language models (LLMs), interest has increased in their potential to support cognitive classification. However, evidence comparing LLM-based classification with expert consensus in dental specialty examinations remains limited. This study aimed to analyze the cognitive levels of dentomaxillofacial radiology (DMFR) questions in the Turkish Dentistry Specialization Examination (DUS) and to evaluate the agreement between expert consensus and ChatGPT v5.2.</p> Methods <p>A total of 130 text-based DMFR questions from the DUS were classified according to Bloom’s revised taxonomy using expert consensus and ChatGPT v5.2. Agreement between classifications was assessed using exact agreement and quadratic-weighted Cohen’s kappa.</p> Results <p>“Analyze” was the most frequent cognitive level in both expert consensus (41.5%) and ChatGPT classification (44.6%), followed by “Remember.” Exact agreement between ChatGPT and expert consensus was 80.0%, with almost perfect ordinal agreement (κw = 0.851; 95% bootstrap CI: 0.769–0.923).</p> Conclusions <p>ChatGPT v5.2 showed high agreement with expert consensus for Bloom-level classification in this dataset. These findings suggest that LLM-based classification may support Bloom-level mapping as an assistive tool; however, they should be interpreted as preliminary evidence of feasibility rather than practical validation. Expert oversight remains necessary, and further research is needed to assess performance across different models, settings, and educational contexts.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cognitive-level analysis of dentomaxillofacial radiology questions in the Turkish dentistry specialization examination: a Bloom’s revised taxonomy analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月21日

查看全文
  • Melisa Ocbe,
  • Hulya Cerci Akcay,
  • Birsay Gumru,
  • Sebnem Ercalik Yalcinkaya

摘要

Background

Bloom’s revised taxonomy is widely used to evaluate the cognitive level of examination questions. With the growing use of large language models (LLMs), interest has increased in their potential to support cognitive classification. However, evidence comparing LLM-based classification with expert consensus in dental specialty examinations remains limited. This study aimed to analyze the cognitive levels of dentomaxillofacial radiology (DMFR) questions in the Turkish Dentistry Specialization Examination (DUS) and to evaluate the agreement between expert consensus and ChatGPT v5.2.

Methods

A total of 130 text-based DMFR questions from the DUS were classified according to Bloom’s revised taxonomy using expert consensus and ChatGPT v5.2. Agreement between classifications was assessed using exact agreement and quadratic-weighted Cohen’s kappa.

Results

“Analyze” was the most frequent cognitive level in both expert consensus (41.5%) and ChatGPT classification (44.6%), followed by “Remember.” Exact agreement between ChatGPT and expert consensus was 80.0%, with almost perfect ordinal agreement (κw = 0.851; 95% bootstrap CI: 0.769–0.923).

Conclusions

ChatGPT v5.2 showed high agreement with expert consensus for Bloom-level classification in this dataset. These findings suggest that LLM-based classification may support Bloom-level mapping as an assistive tool; however, they should be interpreted as preliminary evidence of feasibility rather than practical validation. Expert oversight remains necessary, and further research is needed to assess performance across different models, settings, and educational contexts.