Abstract
The rapid uptake of generative AI (GenAI) has exposed weaknesses in assessment design and policy. The AI Assessment Scale (AIAS) offers a practical way to align permitted AI use with intended learning outcomes. However, as the scale has grown, we have seen implementations that we feel could be improved: unenforceable ‘No AI’ labels, assigning labels to existing assessments without changing the assessments themselves, equity blind spots, and control-first policies that encourage ‘performance theatre’.
This commentary outlines our recommendations for AIAS implementations that replace detection with design: selecting levels by outcome and conditions, requiring light process evidence where suitable, adjusting criteria to assess judgement and voice, sequencing evidence across time, and planning for equitable access. We highlight sector-specific considerations for implementing and adapting the AIAS for K-12, higher education, Technical and Vocational Education and Training (TVET), English as a Foreign Language (EFL), and English for Academic Purposes (EAP).