AI Fluency: Framework & Foundations

• QuestionWhat is "Classify failure into a 4D"?Tap to revealAnswerWhen an LLM interaction fails, the 4Ds tell you which kind of failure it is — Delegation (wrong task), Description (weak prompt), Discernment (skipped verification), or Diligence (deployment miss).
• QuestionIn "Classify failure into a 4D", explain:Tap to revealAnswerDelegation = wrong task for the model. Fix: change task shape (tools, decomposition, different system).
• QuestionIn "Classify failure into a 4D", explain:Tap to revealAnswerDescription = right task, weak prompt. Fix: re-describe (role, constraints, format, examples).
• QuestionIn "Classify failure into a 4D", explain:Tap to revealAnswerDiscernment = output trusted without verification. Fix: check facts, format, edge cases.
• QuestionIn "Classify failure into a 4D", explain:Tap to revealAnswerDiligence = deployment-side miss (oversight, bias, misuse). Fix: process and policy.
• QuestionAnalogy for "Classify failure into a 4D":Tap to revealAnswerThink of the 4Ds as four different ways a delivery can go wrong: wrong address (Delegation), unclear instructions (Description), didn't check the package (Discernment), or no insurance (Diligence). The fix depends on which one happened.
• QuestionWhat is "Identify mis-delegation"?Tap to revealAnswerMis-delegation is asking the LLM to do something it's structurally bad at. The fix is architectural — a tool, decomposition, or a different system — not a better prompt.
• QuestionIn "Identify mis-delegation", explain:Tap to revealAnswerMis-delegation = task structurally bad for an LLM. Architecture problem, not prompt problem.
• QuestionIn "Identify mis-delegation", explain:Tap to revealAnswerFive canonical classes: exact arithmetic, fresh/real-time data, authoritative facts, large deterministic transforms, irreversible high-stakes calls.
• QuestionIn "Identify mis-delegation", explain:Tap to revealAnswerDiagnostic signature: prompt iteration plateau. If quality won't move with better prompts, the task is wrong.
• QuestionIn "Identify mis-delegation", explain:Tap to revealAnswerFix vocabulary: add a tool, decompose, change the surface, or don't use an LLM.
• QuestionAnalogy for "Identify mis-delegation":Tap to revealAnswerAsking an LLM to do exact arithmetic is like asking a writer to be your accountant. They can talk about numbers, but they're not the right hire for the books. Get them a calculator (a tool) or hire an accountant (a different system).
• QuestionWhat is "Write a structured description"?Tap to revealAnswerA good prompt names four things: who you want the model to be, what to do, what rules to follow, and exactly what shape the output should take. Add a worked example.
• QuestionIn "Write a structured description", explain:Tap to revealAnswerStructured description = role + task + constraints + format. All four, named explicitly.
• QuestionIn "Write a structured description", explain:Tap to revealAnswerOne worked example beats three sentences of constraints.
• QuestionIn "Write a structured description", explain:Tap to revealAnswerDiagnostic: a junior with the prompt and no context should know exactly what to produce.
• QuestionIn "Write a structured description", explain:Tap to revealAnswerUnder-specification is silently filled by the model's priors — not a feature.
• QuestionAnalogy for "Write a structured description":Tap to revealAnswerBriefing the model is like briefing a contractor. 'Build me a house' is bad. 'Build me a 3-bed bungalow with a flat roof, here's a sketch' gets you what you wanted on the first try.
• QuestionWhat is "Run description ↔ discernment loop"?Tap to revealAnswerWhen the output disappoints, ask 'what did the prompt leave the model to guess?' and add it to the prompt. That's the loop.
• QuestionIn "Run description ↔ discernment loop", explain:Tap to revealAnswerDescription and Discernment are a loop, not a sequence.
• QuestionIn "Run description ↔ discernment loop", explain:Tap to revealAnswerDisappointing output → ask 'what did I leave to guess?' → re-describe.
• QuestionIn "Run description ↔ discernment loop", explain:Tap to revealAnswerRetrying without re-describing is sampling, not iteration.
• QuestionIn "Run description ↔ discernment loop", explain:Tap to revealAnswerWhen re-description stops helping, you've hit Delegation or Diligence.
• QuestionAnalogy for "Run description ↔ discernment loop":Tap to revealAnswerIt's like editing a recipe. The cake came out dry — instead of baking the same recipe again hoping for better, change the recipe. Each disappointment teaches you what to add to the recipe; the next bake is better.
• QuestionWhat is "Name the diligence concern"?Tap to revealAnswerDiligence is the responsible-deployment layer: oversight, bias, misuse, ownership, transparency. Match the concern to the shape of the deployment.
• QuestionIn "Name the diligence concern", explain:Tap to revealAnswerDiligence = deployment-side concerns: oversight, bias, misuse, ownership, transparency.
• QuestionIn "Name the diligence concern", explain:Tap to revealAnswerConcern is matched to deployment shape, not blanketed.
• QuestionIn "Name the diligence concern", explain:Tap to revealAnswerDiligence interventions live in design, not in the prompt.
• QuestionIn "Name the diligence concern", explain:Tap to revealAnswerFailure mode is news-story shaped, not bug-shaped — design for it upfront.
• QuestionAnalogy for "Name the diligence concern":Tap to revealAnswerDiligence is to LLM apps what safety engineering is to physical products. You don't bolt it on after the recall — you design for it.

1. 1Diagram the core idea

   Analyse~10 min

   In your own words, diagram the central idea of "Classify failure into a 4D". Aim for one paragraph, no jargon you haven't earned.

   Done when: A non-expert could read your paragraph and explain it back to you.

   Re-read the source concept

2. 2Translate it to a real situation

   Apply~15 min

   Find a real situation from your own work / life where "Name the diligence concern" applies. Translate the idea — describe the situation, the move you'd make, and the alternative you considered first.

   Done when: You can name a specific situation (not a hypothetical), state your move, and explain the trade-off in one paragraph.

   Re-read the source concept

3. 3Teach it back

   Evaluate~5 min

   Pick someone (a colleague, a peer, a rubber duck) and explain the section to them in under 5 minutes. Use your own example, not the one from the lesson.

   Done when: Your listener can repeat the core idea + the trade-off back to you without prompting.

Why this matters. Reading + quizzing prove recall; applying it in a real situation is the only evidence of transfer. Treat at least one of these as non-negotiable before you move on.