When AI can make the project, grade the thinking behind it
Ali Shakiba describes an AI-permissive university engineering course. The design pairs a finished project with its source, checks, student explanations and individual review. By week 11 it involved 87 students in 17 teams. That is an early course report, not proof of a K-12 assessment method.
What the source reports
Ali Shakiba describes a university engineering course where students may use AI to build projects. The course does not treat a polished final product as enough proof of skill. Its design also asks for the work’s origin, the student’s checks and explanations, reflection and individual review. By week 11, the first deployment involved 87 students and 17 teams. A teacher still has to judge that evidence; a trail does not score itself. This is an early course design and deployment, not a tested K-12 assessment. It does not show that collecting more records alone proves what each student knows.
Original source
- Title
- When AI Can Build It, What Should We Assess? Rethinking Evidence of Engineering Competence
- Author
- Ali Shakiba
- Publication
- arXiv
- Date
- Monday, October 26, 2026