Khan Academy旗下的AI辅导工具Khanmigo在美国田纳西州进行了一项为期两年的教学效果评估。1这项集群随机对照试验涵盖18所中学，旨在考察该工具对数学教学的实际影响。1
试验结果表明，使用Khanmigo的学生每学期数学成绩提升1.3个百分点，折合每学年0.06至0.08个标准差。1若学生完整参与一年，推估效果可达0.14个标准差。1然而，学生的低参与度成为制约因素。1虽然96%的学生至少使用过该工具一次，但中位数学生仅在三分之一的练习日使用，且学生仅在17%的出错练习中与AI工具进行互动。1
进一步分析显示，学生与工具的互动方式存在问题。196%的学生消息为直接答案或点击建议提示，反映出学生倾向于寻求快速解答而非深度学习。1试验表明，学生参与度而非AI工具本身是主要的限制因素。1
A two-year randomized controlled trial conducted across 18 middle schools in Tennessee examined the effectiveness of Khanmigo, an artificial intelligence tutoring tool operated by Khan Academy, on mathematics instruction.1 The study found that students using Khanmigo improved their math performance by 1.3 percentage points per semester, equivalent to 0.06 to 0.08 standard deviations annually.1 For students who engaged with the tool consistently throughout a full year, the estimated improvement reached 0.14 standard deviations.1
Despite near-universal access to the technology, student engagement emerged as the primary constraint limiting the tool's impact.1 While 96 percent of students used Khanmigo at least once, the median student utilized it on only one-third of practice days.1 When students encountered errors, they interacted with the AI tool in just 17 percent of cases.1 Furthermore, 96 percent of student interactions with the system consisted of requests for direct answers or clicks on hint buttons, rather than deeper engagement with the tutoring features.1 These patterns of limited and surface-level usage suggest that student participation, rather than the capabilities of the AI tool itself, represents the main obstacle to achieving stronger academic outcomes.1
评论
还没有评论，欢迎留下第一条。