Scholarlink

With the recent popularity of Large Language Model (LLM) applications such as GPT-4o, o3-mini and DeepSeek-R1, subject instructors, class teachers, and educational practitioners generally perceive LLMs as virtual assistants in education, creating multiple types of assessments such as assignments, quizzes, examination papers, marking schemes and suggested solutions in formal and informal settings. The present study is to explore the use of contemporary LLMs such as GPT-4o, o3-mini and DeepSeek-R1 in preparing assessment questions, suggested solutions and marking schemes of Probability and Engineering Statistics in response to research questions on the completeness, correctness, consistency and cognitive levels of assessment questions and solutions. It is found that the LLM-created assessment questions are complete, and the total score of the assessment questions is consistent with that of its marking scheme. The suggested solutions to the questions are complete and mostly correct. Despite arithmetic errors, there are about two-thirds of application-oriented questions, and other hierarchies of cognitive learning questions such as understanding-oriented, analysis-oriented and evaluation-oriented questions in the assessment, aligning with learning outcomes of the subject satisfactorily.

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

Evaluating Cognitive Learning Hierarchies of LLM-Created Assessment Questions in Engineering Education

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Christopher Chung Lim Kwan

摘要

With the recent popularity of Large Language Model (LLM) applications such as GPT-4o, o3-mini and DeepSeek-R1, subject instructors, class teachers, and educational practitioners generally perceive LLMs as virtual assistants in education, creating multiple types of assessments such as assignments, quizzes, examination papers, marking schemes and suggested solutions in formal and informal settings. The present study is to explore the use of contemporary LLMs such as GPT-4o, o3-mini and DeepSeek-R1 in preparing assessment questions, suggested solutions and marking schemes of Probability and Engineering Statistics in response to research questions on the completeness, correctness, consistency and cognitive levels of assessment questions and solutions. It is found that the LLM-created assessment questions are complete, and the total score of the assessment questions is consistent with that of its marking scheme. The suggested solutions to the questions are complete and mostly correct. Despite arithmetic errors, there are about two-thirds of application-oriented questions, and other hierarchies of cognitive learning questions such as understanding-oriented, analysis-oriented and evaluation-oriented questions in the assessment, aligning with learning outcomes of the subject satisfactorily.