The Cognitive Work Audit

A HARVEST KERNEL WORKING SESSION

The Cognitive Work Audit

Bring one assignment you are teaching this term. In about 20 minutes you will leave holding a finished AI use plan: the learning goal, every thinking step marked as student only, AI after a first attempt or AI is fine, a one sentence AI instruction for students, and a short check students do with the app closed. You do the deciding. Nothing here decides it for you, and nothing you type leaves this page.

Why this exists. On September 3, 2026, the American Psychological Association released an expert report on learning with educational technology. One of its six conclusions: AI may improve performance while reducing learning, so students can produce better work while developing less independent knowledge and skill. Another: the real test of learning happens after the app is closed. The report covers learners ages 5 to 18. In October, psychologists Beth M. Schwartz and Christopher Hakala argued in Inside Higher Ed and Faculty Focus that the same science of learning should guide AI decisions in college courses, starting from one question: which cognitive work must remain with the student. This session runs that question on one assignment you already teach.

SEED

Pick the assignment and write the goal

Choose one assignment where you are not sure what AI is doing to the learning. Then write the goal as something a student can do when it is finished, not something they understand. Explain, defend, solve, compare, apply.

SPROUT

List the thinking, then protect it

Write down every thinking step the assignment asks for: retrieve, organize, connect, explain, find a gap, correct a mistake, apply to a new case. Then pick the steps that are the point of the assignment. Those are student only. Be strict here. That is the learning.

GROW

Place the AI where it extends the thinking

AI helps most after the student has made a first attempt, as a challenger rather than an author. A student explains first, then asks AI for questions that push the reasoning. A student drafts first, then asks AI to find an unclear claim, missing evidence or a counterargument, and decides what to accept and why. Some steps cost no learning at all, and AI is fine there.

HARVEST

Design the after the app is closed check

The check is how you find out whether AI extended the thinking or did it. Keep it short and do it in class or live: explain the argument out loud, solve a fresh problem of the same kind on paper, or answer one question about the work that nobody saw coming. Decide now what a passing answer looks like, and what you will do if most students cannot give one.

Before you copy it out, one observation.

That is one assignment.

Every other lesson in the course carries the same question: which thinking stays with the student, where AI enters, and how you will know after the app is closed. Auditing this one by hand worked, and it took real attention for one assignment. The judgment about which thinking matters is the part that has to stay yours. Planning every lesson in the term that way is the part that does not scale by hand. It is a scale problem, not a quality problem.

When the whole term is the problem

The Lesson Plan Builder builds timed plans and handouts, grounded in your Course Record: your course, your outcomes, your policies and your voice, planted once and used by every tool in the toolkit. Bring the plan you built today as the pattern. Do the first one by hand anyway. Then you will know exactly what every lesson has to protect.

See the Lesson Plan Builder