我用真实对话取代传统编程测试，考察学生如何借助 AI 解决问题
I replaced traditional coding tests with real conversations about how my students solve problems using AI.
一位数据挖掘课程教师用 Gemini in Google Colab 充当学生的个人导师，取代传统编程测试，改以口头面试让学生为建模决策辩护、排查 AI 建议的错误。教师借助 Gemini Canvas 和 Gemini Notebook 规模化评估这些作业，认为 AI 已能产出可用模型，关键在学生能否对 AI 行使判断力。
Embedding AI as a collaborative thinking partner completely transformed how I assess my data mining students.
Instead of spending hours grinding through coding syntax, they use Gemini in Google Colab as a personal tutor. While the AI suggests code and alternatives, students must decide what to use, question the suggestions, and justify final modelling decisions.
This creates a meaningful form of human–AI teaming. Because AI can increasingly produce a working model, what matters is whether students can exercise judgment over the AI supporting them. I conduct verbal interviews where students defend their choices, troubleshoot and identify where AI's recommendations may be incorrect. Behind the scenes, Gemini Canvas and Gemini Notebook help me evaluate this work at scale.
Balancing my roles as a professor and Chief Digital Officer, my core belief remains: we aren't replacing learning with AI, but intentionally redesigning it to build lasting human capability rather than tech dependency.
来源：Google Blog：AI · blog.google