▲ Claude Academy
Sign in

Vault note

Course: AI Fluency for Students

Mirrors: Anthropic Academy — AI Fluency for Students · https://anthropic.skilljar.com/ Audience: Students at any level (high school, undergrad, grad). No technical background needed. · Time: ~60 min + project Prereqs: claude-101 and a claude.ai account (setup-checklist). · Backing notes: ai-fluency-framework-foundations, ai-capabilities-and-limitations Project: p03-ai-fluency-delegation-audit

This course re-teaches the 4D framework — Delegation, Description, Discernment, Diligence — for the one job that defines student life: learning. The framework was developed as open educational material by Rick Dakan and Joseph Feller. The hardest idea here is also the most important: AI can either accelerate your learning or quietly replace it. Fluency is knowing the difference and choosing acceleration.

Learning objectives

After this course you can:

  • Use AI to learn faster without outsourcing the thinking that learning requires.
  • Apply the four Ds — Delegation, Description, Discernment, Diligence — to coursework.
  • Build study workflows (explain-back, practice generation, feedback loops) that deepen understanding.
  • Discern when an AI answer is wrong, shallow, or fabricated before you rely on it.
  • Disclose AI use honestly and stay inside your institution's academic-integrity policy.

Module 1 — The 4D framework, for learners

🎞 Frame 1 · Learning vs. cheating is a Delegation question · ⏱ ~3 min

🎬 Scene — Split screen. Left: a student pastes an essay prompt and submits the AI's answer. Right: a student asks the AI to quiz them on the same topic until they can explain it.

🧠 Concept — Delegation is deciding what to hand to AI and what to keep for yourself. The line between learning and cheating isn't "did you use AI" — it's "did you delegate away the very thinking the assignment was meant to build." Delegate the chores; keep the cognition.

🖼 On screen

Safe to delegateKeep for yourself
Reformatting notes, making flashcardsForming your own argument / thesis
Generating extra practice problemsDoing the practice problems
Explaining a stuck concept a new wayWriting the words you'll be graded on
Checking your reasoning after you tryRecalling facts on an exam

✅ Checkpoint — Name one task in a current assignment that is safe to delegate and one you must keep.

🎞 Frame 2 · Description: ask like a student, not a search box · ⏱ ~3 min

🎬 Scene — A vague prompt "explain the French Revolution" returns a Wikipedia-flavored wall of text. A described prompt returns a targeted, level-appropriate explanation.

🧠 Concept — Description is telling the AI your role, goal, what you already know, and the format you need. For learning, the magic words are your current level and what's confusing you — that turns a generic answer into a tutor's answer. (Deeper: prompt-engineering-basics.)

🖼 On screen

❌ "Explain integration by parts."
✅ "I'm in first-year calculus and I get the formula but never
    know which part to call u. Explain how to choose u with a
    simple rule, then walk one example slowly. Assume I know
    the product rule."

✅ Checkpoint — Rewrite a topic you're stuck on into a prompt that states your level and your specific confusion.

🎞 Frame 3 · Discernment and Diligence: the student's safety net · ⏱ ~3 min

🎬 Scene — The AI confidently cites a study that doesn't exist. A diligent student tries to find the source — and can't.

🧠 Concept — Discernment is judging whether an output is correct, complete, and yours-to-use. Diligence is being responsible: verifying claims, disclosing AI use, and respecting policy. For students these two Ds protect both your grade and your integrity. (Deeper: ai-capabilities-and-limitations.)

🖼 On screen

Discernment → Is this right? Deep enough? Could I defend it in class?
Diligence   → Did I verify the facts? Disclose the help? Follow policy?

⚠️ Gotcha — AI invents citations, dates, and quotes with total confidence. Never paste an AI-generated source into a bibliography without finding the real thing.

✅ Checkpoint — In your own words, what's the difference between Discernment and Diligence?

Module 2 — Study workflows that actually teach you

🎞 Frame 4 · The explain-back loop · ⏱ ~3 min

🎬 Scene — A student explains a concept to the AI; the AI plays a confused peer and asks "but why?" until the gaps surface.

🧠 Concept — The fastest way to learn is to teach. Use AI as a patient student you explain things to (the Feynman technique). When you can't explain it, you've found exactly what to study.

🖼 On screen

You: "I'll explain osmosis; act like a curious 12-year-old and
      ask 'why' when I'm vague or wrong."
You: <your explanation>
AI:  "Wait — why does water move toward the saltier side?"
You: <realize you don't fully know → go look it up>

✅ Checkpoint — Pick a concept from this week and explain it to the AI as a curious novice. Where did you stall?

🎞 Frame 5 · Practice generation, not answer generation · ⏱ ~3 min

🎬 Scene — Instead of "solve these problems," the student asks for new problems and hides the answers until they've tried.

🧠 Concept — Delegate the making of practice, never the doing. Ask for fresh problems at your level, do them yourself, then have the AI grade and explain only what you missed. Retrieval practice is what moves knowledge into memory.

🖼 On screen

"Generate 5 practice questions on this chapter at exam difficulty.
 Don't show answers yet. After I answer, mark each and explain
 only the ones I got wrong."

⚠️ Gotcha — If you let the AI solve the problem set, you'll feel like you learned and discover on the exam that you didn't.

✅ Checkpoint — Why is asking AI to create practice safer for learning than asking it to complete it?

🎞 Frame 6 · The feedback loop on your own work · ⏱ ~3 min

🎬 Scene — A student pastes their own essay draft and asks for feedback on argument structure — not for a rewrite.

🧠 Concept — AI is a tireless feedback partner. Ask it to critique your reasoning, find weak spots, or play devil's advocate. The words stay yours; the feedback sharpens them. This keeps you on the right side of the Delegation line.

🖼 On screen

✅ "Here's my thesis and three reasons. Which reason is weakest
    and why? Don't rewrite it — just point at the problem."
❌ "Rewrite my essay so it gets an A."

✅ Checkpoint — Phrase a request that gets feedback on your draft without the AI writing it for you.

Module 3 — Integrity, disclosure, and over-reliance

🎞 Frame 7 · Read your institution's policy first · ⏱ ~3 min

🎬 Scene — Two syllabi side by side: one bans AI outright, one requires an "AI use" appendix. Same campus, different rules.

🧠 Concept — Diligence starts with knowing the rules. AI policy varies by course, instructor, and assignment. The safe default: assume you must ask and disclose unless told otherwise. "I didn't know" is not a defense for an integrity violation.

🖼 On screen

Before using AI on graded work, check:
  • Course/assignment policy (syllabus, rubric, LMS)
  • Whether disclosure is required, and in what form
  • Whether *this specific task* is permitted
When unsure → ask the instructor, in writing.

✅ Checkpoint — Where would you find the AI policy for your next graded assignment?

🎞 Frame 8 · How to disclose AI use honestly · ⏱ ~3 min

🎬 Scene — A student appends a short note: "I used Claude to generate practice questions and to critique my draft's structure; all writing and analysis are my own."

🧠 Concept — Good disclosure is specific: what you used AI for and what you kept. Transparency is the cheapest way to stay above suspicion — and it forces you to be honest with yourself about how much was really you.

🖼 On screen

AI Use Statement (example)
  Tool: Claude (claude.ai)
  Used for: brainstorming counter-arguments; checking my math;
            generating practice problems.
  NOT used for: the thesis, the prose, the final analysis.

✅ Checkpoint — Draft a one-line AI-use statement for a paper where you used AI only to find weaknesses in your argument.

🎞 Frame 9 · The over-reliance trap · ⏱ ~3 min

🎬 Scene — A student who "understood everything" with AI's help freezes on a closed-book exam.

🧠 Concept — The danger isn't a wrong answer — it's fluency illusion: AI makes hard things feel easy, so you stop building the skill. Periodically work without AI to test what's actually in your head. If you can't, you've over-delegated.

🖼 On screen

Reliance self-check (weekly):
  Could I do this WITHOUT the AI right now?
   ├─ Yes  → healthy delegation
   └─ No   → you delegated the learning; reclaim it

⚠️ Gotcha — Smooth ≠ learned. The exam is closed-book; train for it.

✅ Checkpoint — Name one skill you've been letting AI carry that you should be able to do alone.

Module 4 — Discerning AI output in coursework

🎞 Frame 10 · Spotting confident nonsense · ⏱ ~3 min

🎬 Scene — An AI explanation of a historical event mixes a real cause with an invented one, seamlessly.

🧠 Concept — Discernment for coursework means cross-checking against your source material. Ground the AI in your textbook or lecture notes (upload them) and treat any claim it can't tie to a source as unverified. (Deeper: ai-capabilities-and-limitations.)

🖼 On screen

Trust ladder:
  Grounded in MY uploaded notes/textbook   → likely reliable
  General knowledge, I can verify it       → check, then use
  A specific stat/quote/citation           → verify the source exists
  "It sounds right"                         → not good enough

✅ Checkpoint — Give two kinds of AI claims you must always verify before putting them in an assignment.

🎞 Frame 11 · Use AI to deepen, not to shortcut · ⏱ ~2 min

🎬 Scene — A student who has a draft answer asks AI for three harder follow-up questions to test their understanding.

🧠 Concept — The most fluent move is to use AI to make a topic harder for yourself on purpose: "What would a professor ask next? What's the strongest objection?" That's deliberate practice, not a shortcut.

✅ Checkpoint — Ask AI for the toughest follow-up question on something you think you've mastered. Can you answer it?

🎞 Frame 12 · You've got the student's 4Ds · ⏱ ~1 min

🎬 Scene — Recap slide: delegate chores not cognition, describe your level, discern the output, do the diligence.

🧠 Concept — You can now use AI as a study accelerator that makes you smarter rather than a crutch that makes you look smart. Next: pressure-test these habits in the project.

✅ Checkpoint — Without looking, list the four Ds and one student example of each.

🛠 Project

Complete p03-ai-fluency-delegation-audit — Delegation & Discernment Audit (Student edition). You'll take one real assignment from this term, draw the line between what you delegate and what you keep, run an explain-back and a practice-generation loop, verify three AI claims against your sources, and write an honest AI-use disclosure statement.

🧪 Self-check quiz

  1. What are the four Ds, in order?
  2. What single question separates "learning with AI" from "cheating with AI"?
  3. Why is asking AI to generate practice safer than asking it to solve it?
  4. What does Discernment ask vs. what does Diligence ask?
  5. Give one element every honest AI-use disclosure should contain.
  6. What is the "fluency illusion" and how do you test for it?
  7. Which kind of AI output most often needs source-verification before you cite it?
<details><summary>Answers</summary>
  1. Delegation, Description, Discernment, Diligence. 2. "Did I delegate away the thinking the assignment was meant to build?" 3. You still do the retrieval practice that moves knowledge into memory; solving it for you skips the learning. 4. Discernment: "Is this right/deep/mine to use?" Diligence: "Did I verify, disclose, and follow policy?" 5. What you used AI for (and ideally what you did not). 6. The feeling of understanding AI gives without the skill; test by working the task with no AI. 7. Specific facts, stats, quotes, and citations — AI fabricates these confidently.
</details>

🎓 Certificate criteria

You've "passed" AI Fluency for Students when you can:

  • State the four Ds and give a student example of each.
  • Run an explain-back loop and a practice-generation loop on real coursework.
  • Verify three AI claims against your own source material.
  • Write an honest, specific AI-use disclosure for an assignment.
  • Complete p03-ai-fluency-delegation-audit and journal one over-reliance habit you fixed in learning-journal-template.

Tick this course off in progress and record the date you earned Anthropic's official certificate.

🔗 Sources & deeper notes