Assessment Redesign and the AI Assessment Scale

Most assignments quietly do three jobs at once: production (practice that builds skill), measurement (evidence that certifies it), and logistics (course plumbing). This lecture shows why AI makes that ambiguity expensive — the highest-risk pattern is letting students practice with AI and then grading as if they practiced without it — and walks through annotating each sentence of a prompt by its job. Small relabeling moves in the spirit of TILT let you attach different AI expectations to different components and give differently prepared students a visible sequence of practice, support, and proof.

An introduction to the AI Assessment Scale (AIAS), five assessment contracts running from no AI, through AI-assisted planning and collaboration, to full AI use and AI-as-object-of-critique. Each level changes three things together — what the student does, what AI does, and what evidence is required — which is why the scale works best applied per component rather than pasted across a whole course. The lecture pairs the scale with an access plan (no-cost path, privacy rule, fallback) and distinguishes the assignment contract from the integrity policy that backs it.

The Stress Test treats your assignment as the object of a failure analysis before students arrive: what can a current AI tool produce, what learning sits invisibly below the polished product, what small piece of process evidence would restore confidence, and what equity cost would that fix create? Results feed a decision tree — keep, revise, split, replace, or integrate AI — with detect-and-punish deliberately absent. A risk heat map focuses limited redesign time where bypass is likely and the learning consequence is high, and the output is a short Decision Memo a colleague could challenge.

Transparent and Authentic Assessment combines TILT, authentic assessment, and the AI Assessment Scale into a clear student-facing contract: why the work matters, what students must produce, what strong performance looks like, and where AI is permitted, prohibited, or documented. Authenticity comes not from attaching a prestigious job title to a generic task, but from aligning audience, genre, evidence, and constraints around credible disciplinary judgment. A before-and-after prompt demonstrates how transparency preserves intellectual rigor while reducing hidden expectations, improving equity and feedback. The resulting Decision Memo becomes a design brief for rewriting the assignment in Module Four.