研究人员开发了一种方法，通过分析学生向AI助教提出的问题来识别在线课程中的知识差距。该方法使用一个少样本文本分类器，该分类器由GPT-4提取的先决知识图提供信息，将问题映射到特定的课程主题。该系统在跨43个标签分类问题时达到了80%的准确率，并显示出问题数量与学生报告的主题难度之间存在显著相关性，表明其有潜力突出需要教师关注的领域。 AI
影响 为教师提供了一种利用AI互动数据识别和解决学生知识差距的新颖方法。
排序理由 该集群包含一篇详细介绍分析AI互动新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.10736v1 Announce Type: cross Abstract: Large online courses generate thousands of student questions directed at conversational AI teaching assistants, yet these interaction logs remain largely untapped as diagnostic signals. We present a pipeline that maps student ques…
Large online courses generate thousands of student questions directed at conversational AI teaching assistants, yet these interaction logs remain largely untapped as diagnostic signals. We present a pipeline that maps student questions from a conversational AI teaching assistant …