Scholarlink

In the age of AI in education, the utilization of educational data by AI to provide individually optimized feedback to learners is crucial. Furthermore, understanding how this can be implemented in actual educational settings is essential for envisioning a future where humans co-evolve with AI. In this study, we implemented our recently developed self-explanation based eXplainable AI learning system in real-world settings with first-year middle school students to investigate how AI learning systems can be put into practice. Our system automatically detects stuck points and recommends quizzes based on the learner’s self-explanation text data and pen stroke data. Over a period of approximately one month, we gradually introduced this system to 80 students in a mathematics class. The results showed that the system was capable of automatically detecting stuck points from self-explanations and recommending related problems as personalized feed-back.

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

Pilot Study of Self-explanation Based Automate Stuck Point Detection and Personalized Feed-Back Recommendation by Educational eXplainable AI Tool in Middle School Math Classes

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kyosuke Takami,
  • Ryosuke Nakamoto,
  • Yiling Dai,
  • Brendan Flanagan,
  • Hiroaki Ogata

摘要

In the age of AI in education, the utilization of educational data by AI to provide individually optimized feedback to learners is crucial. Furthermore, understanding how this can be implemented in actual educational settings is essential for envisioning a future where humans co-evolve with AI. In this study, we implemented our recently developed self-explanation based eXplainable AI learning system in real-world settings with first-year middle school students to investigate how AI learning systems can be put into practice. Our system automatically detects stuck points and recommends quizzes based on the learner’s self-explanation text data and pen stroke data. Over a period of approximately one month, we gradually introduced this system to 80 students in a mathematics class. The results showed that the system was capable of automatically detecting stuck points from self-explanations and recommending related problems as personalized feed-back.