Assessment shifts from catching AI use to judging how students evaluate AI.

Teachers often struggle to evaluate student work when students use generative artificial intelligence. Instead of trying to catch whether AI was used, researchers created a low-infrastructure system called the Human-AI Evidence-Based Assessment Protocol, or HAEA-P. This protocol focuses on Human Evaluative Agency, which is a learner's ability to critically judge AI output. Students show evidence of their thinking by using a Decision Triad Matrix. Through this tool, they record whether they accept, modify, or reject AI-generated content, prove their choices with evidence, and reflect on their thinking. The system also gives teachers a fast grading rubric that works well in low-resource schools. When seven experts reviewed the design, their feedback confirmed the core setup worked and helped refine it into version 1.1.

Why it matters

For educators and institutions, this changes what gets measured: not whether a student cheated with AI, but whether they can critically appraise and decide what to accept, modify, or reject. That makes assessment fairer and more focused on skills students actually need, without demanding expensive monitoring tools.

Questions answered