Check the Machine (CtM)

A 4-step protocol for verifying AI outputs. Use the one-pager as a reference. Use the worksheet with students.

The CtM Protocol — One-Pager

Every time you or your students use AI to generate an answer, run this 4-step cycle before trusting the output.

TASK

What did you ask the AI to do? Write the prompt or describe the request. Be specific.

BEFORE

What do you expect the answer to be? Write your prediction, estimate, or prior belief BEFORE looking at the AI's output. This is the most important step — it activates your own knowledge.

AFTER

What did the AI actually produce? Describe the output. Compare it to your prediction. Flag anything that surprises you, contradicts what you know, or that you can't independently verify.

TAKEAWAY

What did the comparison reveal? This is about TWO things: (1) What did you learn about the AI tool — where is it reliable? Where does it fail? (2) What did you learn about your OWN understanding?

Principle 1: AI is never the first step. You must have your own expectation before checking the AI's output.

Principle 2: Errors are features, not failures. Finding an AI error means you're learning.

Principle 3: Verification is a skill that transfers. The same mindset applies to checking any data source, calculation, or claim.

Worked Example

Example: 7th Grade Math

Task: I asked the AI: "Write a word problem about proportional relationships using a recipe context for 7th graders."

Before: I expect a problem that gives a recipe for a certain number of servings and asks students to scale it up or down. The numbers should be reasonable (not fractions of eggs) and the proportion should be straightforward.

After: The AI produced a problem about tripling a cookie recipe. It was well-written BUT it asked students to use "3/4 × 3 = 9/4 cups of flour" — mathematically correct, but the AI didn't convert to mixed numbers (2¼ cups) which is how recipes actually work. A student could get the right math answer but an impractical cooking answer.

Takeaway: The AI is good at creating the structure of word problems but doesn't always check whether the answer makes practical sense in context. I need to verify that AI-generated problems have contextually reasonable answers, not just mathematically correct ones.

Student Worksheet

Print or share this section with students. Editable fields below — customize the header for your class.