Back to news
research2026-11-17

Caltech professor stops grading math homework on correctness as AI disrupts university math classes

A Caltech professor said that with generative AI the final answer no longer discriminates between students, and switched grading to focus on derivation process and written explanation, reflecting reassessment of AI-assisted learning evaluation.

A professor at the California Institute of Technology said that with generative AI widely available, the final answer on coursework has lost its ability to discriminate between students, and no longer grades homework on correctness. Attention shifts to the derivation process and the completeness of the written explanation. Student solutions must justify each step, and giving only a conclusion is no longer accepted as valid work.

The change reflects a direct contest between assessment design and tool capability. When models can cheaply produce structurally complete and correct solutions, outcome-based assignments neither distinguish student ability nor reveal the learning process, so the standard moves up to the verifiability of reasoning chains and the rigor of expression.

This discussion appears alongside other developments in AI in education. Reporting in the same period described AI tool deployment in US higher education moving from pilot to routine, and the programme giving academics free access to frontier models is expanding. Institutions face the same trade-off: lowering the barrier to tool use in support of learning while preserving assessment as a filter for independent thinking.

教育AI 辅助学习评估方式