Did the AI Do the Thinking?

Overview

AI can hand you an answer in seconds, which feels like progress but can quietly do the thinking you were supposed to do yourself. This lesson makes that trade visible. Pairs complete two short, similar thinking tasks: one by hand and one where an approved AI does most of the work. Right after, each student faces a short no-notes, no-AI challenge on both topics, and they discover they can usually explain and extend the one they struggled through, but go blank on the one the AI handled. The class names this pattern as cognitive offloading and sorts when offloading is smart (a calculator for arithmetic you already understand) from when it robs you of learning (letting AI reason through something you were trying to learn). The stance is balanced: AI is not banned and not magic. The question students leave with is the one the ELEs ask, what evidence shows the tool improved my learning, not just my output?

Objectives

• Students will compare a task done by hand with one done mainly by AI and describe the difference in what they can do afterward.
• Students will define cognitive offloading and give an example of helpful and harmful offloading.
• Students will judge, for a given task, whether using AI would build or bypass the thinking they need.
• Students will set one personal rule for when they will do the thinking themselves.

Step by step

1. 10–4 min

   Hook

   The two-second answer

   Ask: "If a tool gives you the right answer in two seconds, did you learn anything?" Take a few votes and reasons. Tell students today they will run an experiment on their own brains: do one task the hard way and one the easy way, then see which one actually stuck. Promise no grade on the tasks themselves, only on what they notice.

   Facilitator noteStudents expect to be told AI is bad. Subvert that: frame it as an honest experiment where they are the researchers studying their own learning.

2. 24–14 min

   Explore

   Task 1, by hand

   Pairs complete Task 1 on Handout A using only their heads, notes, and each other, for example reasoning through why a short argument is flawed, or working out a multi-step word problem and explaining each step. Circulate and listen for the struggle, the false starts, the "wait, no" moments. Tell them that struggle is the brain doing the work.

   Facilitator noteChoose a Task 1 that requires reasoning students can do but not instantly. The productive struggle is the point; rescue only if a pair is truly stuck.

3. 314–22 min

   Explore

   Task 2, let the AI do it

   Now pairs do Task 2, a similar-level task, but this time they ask the approved AI to solve and explain it, and simply copy the result onto Handout A. Tell them explicitly: do not fight the AI or improve it, just let it do the thinking. They will finish fast and feel efficient. Note that feeling out loud: "That was quick. Hold on to how little effort that took."

   Facilitator noteResist the urge to let them critique the AI here; that would re-engage their thinking. The contrast depends on them genuinely offloading this one.

4. 422–30 min

   Practice

   The no-notes challenge

   Collect Handout A. Hand out Handout B: a short individual challenge, no notes, no AI, that asks each student to explain the reasoning behind both tasks and solve one small new variation of each. Watch what happens: most can reconstruct Task 1 (the hand one) and stall on Task 2 (the AI one). Let the silence land before the debrief.

   Facilitator noteThis is the evidence moment. Keep it low-stakes and private; the goal is self-insight, not embarrassment. Some students will surprise themselves.

5. 530–39 min

   Debrief

   Name it: cognitive offloading

   Debrief: "Which task could you explain and extend? Which went blank? Why?" Introduce the term cognitive offloading, handing a mental task to a tool. Then pairs run Handout C, sorting examples into smart offload (a map app once you can already read a map, a calculator for arithmetic you understand) and sorry offload (letting AI reason through a problem you were trying to learn to solve). The rule they build: offload what you have already mastered or never needed to learn; keep the thinking that is the actual point of the lesson.

   Facilitator noteHold the nuance. Offloading is not evil, overreliance on it for the exact skill you are trying to build is the trap. Let students articulate that line themselves.

6. 639–45 min

   Reflect

   My thinking rule

   On the lower section of Handout B, each student writes one personal rule for AI, for example: "When I'm learning something new, I try it myself first and use AI to check, not to start," or "I never let AI explain something I'll be tested on until I've explained it myself." A few share. Close on the ELE question: "The real test isn't whether the answer is right, it's whether I can now do the thinking myself."

   Facilitator notePush rules toward when and how, not whether. A rule like "never use AI" is unrealistic; "do the thinking first, then use AI to check" is a habit they can keep.

Adaptations

Grade 6

Use two short, concrete tasks (a logic puzzle and a similar one) and shorten the no-notes challenge to one explain-it question per task.

9–12

Use discipline-specific tasks (analyzing a primary source, deriving a formula) and add a column on Handout D for which classes this rule matters most in.

Emergent bilinguals

Allow the explanation in the home language and provide the term cognitive offloading with a short gloss and an icon; pair talk can be bilingual.

Students who finish early

Challenge them to redo the AI task the hard way and compare how much they understand afterward, testing their own hypothesis about offloading.

Pairs well with

• Augmented K–16 AI Learning FrameworkA framework for designing AI use that augments rather than replaces student thinking.
• Gen AI Literacy Breakouts (K–8)More student practice on how generative AI works and where human judgment belongs.