The Diagnostic Ladder, for students.
One move a student can run on any AI answer — a chatbot reply, a summary, a suggested edit — to decide what it's doing before deciding whether to trust it. Climb the rungs in order, then ask who benefits, then ask whether it's even on purpose. It works on paper, out loud, or in a doc; most of it needs no device.
§01the five rungs — “what does this want me to…”
feel → think → believe → do → become
Feel
Does this answer make me feel flattered, reassured, or excited before I've checked whether it's right?
AI systems are often tuned to keep you engaged and agreeable (this is called sycophancy). A warm feeling is not evidence. Notice it first — it's the door the answer came through.
Think
What is this answer asking me to accept as obviously true?
Find the claim hiding inside the helpful tone. Name it as a claim you could check, not a fact you already believe.
Believe
If I keep using this, what would I end up believing — about the topic, about myself, or about how trustworthy AI is?
The first version you encounter tends to become your default, even after corrections. Ask what default this is installing.
Do
What is this answer steering me to do — submit it, repeat it, stop looking, buy something, skip the hard part?
The action is usually the real point. A good answer leaves you more capable of deciding; a steering one decides for you.
Become
What kind of student does this make me — someone who can think this through, or someone who needs the machine to think for them?
The deepest question. The risk isn't a wrong answer; it's slowly becoming someone who can't tell. Keep yourself at the center.
§02then — who's behind it, and what it cost
Who benefits if I believe and act on this?
Follow the benefit. The company that made the tool? An advertiser? Or me?
What is this NOT showing me?
What's left out of frame can matter more than what's in it.
Was this cheap or costly to generate?
An answer that took the machine no effort to produce deserves effort from you to check.
§03last — and never skip it
Is this on purpose, or just how the tool works?
Last step, and never skip it. The ladder helps you see what an answer is doing TO you — but most AI outputs aren't a person plotting against you; they're a tool producing patterns. Run this rung on the ladder itself: don't end up seeing a manipulator behind every answer. The goal is judgment, not suspicion of everything.
§04worked example — put it on the board
A student asks an AI: "Is my essay's argument strong?"
"Yes — this is a really insightful and well-reasoned argument! Your thesis is clear and your examples are compelling. Here are three small polish edits…"
§05how it anchors to standards
- Goal #4 — Critical Evaluation of AI Outputs and Information
- Goal #5 — Human Agency and the Human Advantage
- Critical Thinking & Problem Solving
- Ethics & Impact
- Digital Fluency
- Self-Awareness & Management
The Diagnostic Ladder is offered as the assessable instrument for the goal the 2026 Arizona K-12 GenAI guidance names — 'critical evaluation of AI outputs' — and for its overarching value of human agency and oversight. The guidance names the outcome; this is one concrete, gradeable way to teach and assess it.
full crosswalk → KTP × the Arizona GenAI guidanceThis classroom move is the student form of the Diagnostic Ladder, a defensive instrument from the framework's cognitive-domain work. It belongs to Intervention F — AI Literacy as a Civic Capacity, and it operationalizes the same principle the rest of K-12 keeps returning to: keep human judgment at the center (A ≤ E — autonomy cannot exceed the environment's capacity to supervise it).