Assessment in an AI World

Examine assessment validity, formative and summative use, process evidence, feedback, grading and task redesign when generative AI can produce increasingly polished work.

• Start with assessment validity — Ask what inference the assessment is meant to support before redesigning the task.
• Use AI formatively without letting it become the teacher — Use AI to generate information for learning while keeping diagnosis and response connected to teacher judgement.
• Treat process as evidence, not surveillance — Collect a few meaningful traces of thinking, revision and decision-making.
• Add performance where it matters — Use explanation, application and transfer to reconnect knowledge with the learner.
• Separate feedback support from grading authority — Use AI to support feedback workflows without handing consequential judgement to an opaque system.
• Redesign the task, not just the rules — Change the learning experience when AI has weakened the original evidence model.
• Make assessment expectations legible — Align policy, task instructions and classroom practice so learners know what counts as valid evidence.