Assessment Redesign for Generative AI: Program-Level Assurance of Learning

About this course

Assessment redesign for generative AI has mostly happened one assignment at a time, and the surveys show it: 87 percent of United States faculty set their own AI rules while only 35 percent report a department rule and 48 percent an institutional one (Elon and AAC&U, January 2026), and only 20 percent of provosts describe their institution's approach as "primarily generative" (Inside Higher Ed, September 2026). This course teaches the program-level alternative. You will learn why Corbin, Dawson and Liu call course-by-course traffic-light schemes an "enforcement illusion," why detectors cannot carry the load when seven of them flagged non-native writers' TOEFL essays as AI-written 61.3 percent of the time on average, and how Sydney's two-lane model, TEQSA's three assurance pathways and the revised AI Assessment Scale fit together into one program map. You will build that map for a sample program, place secure points such as interactive orals where they assure the learning outcomes, audit a full set of assessment briefs with an AI agent, and leave with a rollout plan and its metrics.

What you will be able to do

1. Distinguish discursive from structural assessment change, and classify their institution's current AI assessment rules as one or the other with evidence.
2. State the published false-positive and accuracy figures for AI-text detectors and explain, in a paragraph suitable for a faculty senate, why detection cannot serve as a program's control.
3. Map every assessment in one program onto Sydney's two lanes and the five levels of the revised AI Assessment Scale, and identify where the program has no secure assurance of a learning outcome.
4. Choose and justify one of TEQSA's three assurance pathways (whole-program, unit-level, hybrid) for a given program and describe what changes at each level.
5. Design an interactive oral assessment for one unit, including its rubric structure and staffing, drawing on the Griffith and Dublin City University case studies.
6. Run an AI agent over a full set of assessment briefs to produce a first-pass lane and level classification, verify it by hand, and use the result to write a rollout plan with measurable program-review metrics.

Who this course is for

Roles: provosts and vice provosts for academic affairs, directors of assessment and institutional effectiveness, quality-assurance and accreditation liaison officers, deans and department chairs, program directors, and center-for-teaching-and-learning leads who support program review.

Institutions: universities and colleges of any size that run program review or assurance-of-learning processes, including those preparing accreditation self-studies. The two frameworks the course uses most, Sydney's two-lane model and TEQSA's pathways, are Australian; the course presents them as international models and asks learners to map them onto their own accreditor's expectations.

Before you start: familiarity with your institution's program-review or assessment cycle. No statistics or coding is required. The lab uses an AI agent through a browser on a sample program's assessment briefs, or on the learner's own.

Format

• Estimated time: 11 to 13 hours.
• Level: intermediate. Assumes program-review experience; assumes nothing about AI.
• Hands-on: one lab that runs an AI agent over a supplied sample program (a 24-unit bachelor's program with 71 assessment briefs) or the learner's own briefs, four templates the learner keeps (program assessment map, secure-point register, interactive oral design, rollout plan), three discussions, and a capstone on one real program.
• Structure: 4 sections, each ending in an AI conversation assessment, then a capstone assessment.

Self-paced. Every lesson is taught by an AI tutor that opens with a short story and works through one idea and one question at a time. Each section ends with an AI conversation assessment, and the course closes with a capstone assessment. Section assessments count for 60% and the capstone for 40%; 70% passes. There are no multiple-choice quizzes.

Currency of this course

Last verified 2026-09-28. The frameworks taught here are expert guidance, not binding regulation, and the documents are still being revised. Dates are stated on every unit and unit 4.1.3 shows how to re-check them.