AI Fluency for Educators: teaching notes

Side-by-side notes for 12-praxis-ai-fluency-for-educators.html (7 slides). This applies the 4D framework (Delegation, Description, Discernment, Diligence) to teaching: course design, materials, assessment. It is the same spine as 05 (AI Fluency: Framework & Foundations) in one profession's world, so the 05 notes are the deeper reference for any 4D question that comes up. Written from the deck and 08-...-lessons.txt, against the Praxis teaching library. Method: teaching-notes-method.md.

How to use these notes

Guidance, not a script. Teach the person, not the slides.
This deck is short (four teaching slides) but each one compresses a full lesson. Don't race it. The 4Ds on slide 4 are the heart; everything else points at them.
Every teaching slide has a Pause, then ask block. Put the question, then stop. Don't fill the silence. If they freeze: "Take your time, type anything when you're ready, including 'I don't know'." Then wait again.
By background gives a line per archetype (tutor/scenarios.md). This is an educators course, so most rooms here are teachers, but the nine archetypes still map: read each as "how this 4D idea lands for that kind of professional", and reach for the one nearest whoever is in front of you. A palette, not a checklist.
The framework is abstract by nature. The job is to keep dragging it back to one real thing they teach. The two questions to keep returning to: "what does that mean to you?" and "how does that apply to your work?"
Hold the voice (train-the-trainer.md Part 4): no "great question", no emoji, one term: Claude Code. Underclaim. Cost is "about twenty dollars a month" and stop.

Title

Teaching with AI as a thinking partner, not an autopilot

The idea

Set the frame: AI in teaching is augmentation, not automation. You stay in the work; the framework outlasts the tools. Four lessons, one per screen, the watching is the lesson.

Say it like this

"The fear with AI in the classroom is that it does the teaching for you, and your judgment leaves the room. That's the opposite of this. AI here is a thinking partner for the parts around the teaching: designing the course, building the materials, setting the assessment. You stay at the centre. Four habits that do that, and they keep working whatever the tool does next."

Open with

The one thing they named in intake: a course they're redesigning, a deck they rebuild every term, a set of assessments they dread writing. "We'll run that through all four habits by the end."

Spine

The whole course on one screen

The idea

Two moves: get fluent in the 4Ds, then put them to work: first designing the course, then building the materials students actually use. Orient, don't teach.

Say it like this

"Here's the shape. First we set your teaching context and learn the four habits. Then we put them to work twice: designing a course, then building the materials and assessment. Foundation, then practice."

Cue

Twenty seconds. The 4D framework on slide 4 is the centre; everything points at it.

Lesson 1

Context first: write it once, reuse it everywhere

The idea

AI gets useful when it knows your situation. Spend the first session turning your teaching (your values, your constraints, who your students are, the methods you trust) into one reusable Teaching Context document, then start every later conversation from it.

Say it like this

"AI is generic until you tell it who you are. So before anything else, you spend ten minutes on your own (your values, your constraints, who your students actually are, the methods you trust), then you let it interview you and write it all into one document. From then on, every conversation starts already knowing your world. You write it once and reuse it everywhere." (Previews framings.md: the folder is the memory; here the Teaching Context is that memory in document form, and in Claude Code it becomes a file that persists.)

Pause, then ask

"If you had to put the way you teach into one paragraph (your values, your constraints, who your students are), what's in it?"
"What do you find yourself re-explaining to a new TA, a substitute, or a tool, every single time?"

(wait: their answer is, almost word for word, the start of the Teaching Context document.)

By background (what their "context document" holds)

Investigator: the house report rules and the standard a finding must meet, briefed once.
Analyst: the standard column set and the "no data" rule for every review.
Designer: brand tokens, the spec sheets, what "on-brand" means here.
Knowledge worker: the CSV schema and the supplier quirks so each batch starts briefed.
Reporter: the five-section format and who reads it.
Content: the house-voice guide and the back catalogue.
Form-filler: the template and the category definitions, reused each cycle.
Outreach: the knowledge base and the relationship history.
Exec: the KPI definitions and what counts as a decision, so each prep starts warm.

Watch for

Teachers may treat the ten minutes of self-reflection as the soft bit to skip. It's the opposite: it's the whole foundation. Everything later is only as good as this. Don't let them rush to the tool before they've named who they teach and what they value.

Lesson 2

The 4Ds: how you collaborate, not just what you ask

The centrepiece. If they take one thing away, it's the four Ds.

The idea

Four habits, each with three parts. Delegation (problem awareness, platform awareness, the task): what to hand off and why. Description (product, process, performance): say what you want clearly. Discernment (the same three, judged): weigh what comes back.

Diligence

(responsible, transparent, accountable): own the outcome. Description and Discernment are a loop in the middle, and that loop is what turns a tool into a thinking partner. Augmentation, not automation: you stay in the work.

Say it like this

"Four habits. Delegation: decide what's even worth doing together, and what stays yours. Description: say what you want, and how. Discernment: judge what comes back. Diligence: own the result (accuracy, bias, transparency). The middle two are a loop you run over and over: describe, look, describe better, go again. That loop is the difference between a tool and a thinking partner. Delegation sets it up, Diligence keeps it honest." (For the full breakdown of any one D, the deck points to Anthropic's Framework & Foundations course; the Praxis 05 notes carry it.)

Pause, then ask

"Of those four, which do you think you're already weakest at? Most people know."
"Take that course or deck you named: what would delegation mean for it, before you typed a word? What's genuinely yours, and what could be the AI's?"

(wait)

By background (one teaching task through all four Ds)

Investigator: delegate the extraction; describe "cite every source, never infer"; discern the citations hold; stand behind the filing.
Analyst: delegate the table; describe the columns and basis; discern the blanks are honest; own the comparison.
Designer: delegate the variants; describe the brand rules; discern they're on-brand; sign off the campaign.
Knowledge worker: delegate the extraction; describe the schema; discern the flagged rows; trust the output.
Reporter: delegate the consolidation; describe the sections; discern nothing's misattributed; send it.
Content: delegate the draft; describe the house voice; discern it's on-voice; publish under your name.
Form-filler: delegate the allocation; describe the categories; discern the totals reconcile; submit it.
Outreach: delegate the per-recipient drafts; describe the format; discern no invented numbers; send it.
Exec: delegate the synthesis; describe "facts, not editorial"; discern risks are stated straight; present it.

Watch for

"Augmentation not automation" matters more in teaching than almost anywhere: the fear is that AI hollows out the thinking students (and teachers) are supposed to do. Name it: we keep your judgment at the centre on purpose, not as a compromise. This also maps onto the skill-file idea they'll meet in Claude 101: "describe what good looks like, once" is writing the rules down. Plant the seed, don't dwell. (framings.md: skill file as institutional memory.)

Lesson 3

Designing the course: you hold the vision

The idea

Apply all four Ds to course design in four stages: plan (what's the work, what needs you), set up (share your context, describe it richly), the Describe → Discern loop across three design tasks (essential content, the learning journey, measurable objectives), then document the decisions and your reasoning. Naming what good looks like up front is what makes the judging possible later.

Say it like this

"Course design becomes four stages. Plan: what's the work, and what genuinely needs you. Set up: hand it your teaching context and describe this course richly. Then the loop, run across three things: the essential content, the learning journey, the measurable objectives. For every suggestion it makes, you ask one question: does this serve my students? Then you say why, build on the good ideas, redirect the weak ones. Last, you document what you decided and why. You're not outsourcing the design. You're holding the vision and using it to think."

Pause, then ask

"For that course, what does 'good' actually look like? Specific enough that you'd know it when you saw it."
"Where in the design is your expertise the thing that catches what a generic suggestion would miss?"

(wait: the loop is the human-premium moment; let them feel that the "does this serve my students?" judgment is theirs alone.)

By background (the loop, in their world: what "does this serve them?" becomes)

Investigator: does this finding hold against the evidence, or is it inferred?
Analyst: are these two figures even comparable on the same basis?
Designer: does this variant actually meet the brief and the spec?
Knowledge worker: is this field real, or did it guess a blank?
Reporter: is this attributed to the right person, and does it belong?
Content: is this line in our voice, or borrowed?
Form-filler: is this receipt in the right section, and do the totals reconcile?
Outreach: is every claim in here backed by the file, no invented numbers?
Exec: is this stated as fact, or quietly editorialised?

Watch for

They'll be tempted to accept good-looking output and move on. The whole value is in the loop: describe, discern, describe better. If they accept the first draft, they've used it as a vending machine, not a partner. Also useful here: ask the AI to take the students' view at a transition point ("what might confuse them, what scaffolding is missing?"), which surfaces gaps their own expertise can then judge.

Lesson 4

One loop builds anything students use

The idea

Every material follows the same three steps: set the task on top of your established context, refine it in the Describe → Discern loop, then run the Diligence check (accuracy, bias, integrity) before it reaches a student. The same loop builds lecture slides, a study guide, an in-class exercise, a knowledge quiz. Shared context is what makes them cohere.

Say it like this

"Once your context exists, building any material is the same three steps. Set the task on top of that context. Refine it in the loop. Then the Diligence check before it ever reaches a student: is it accurate, is it biased, is it honest about AI's role. Slides, a study guide, an exercise, a quiz: same loop every time. And because they all share one source, they line up: the quiz tests what the slides taught." (framings.md: the system improves through use; here the shared context is what makes the whole set coherent.)

Pause, then ask

"Which material do you spend the most time on, and the most time formatting rather than thinking?"
"What's the check you'd have to run before you'd put your name on something a student sees?"

(wait)

By background (the Diligence check that bites for them)

Investigator: is the privacy of the file respected, is disclosure made where required, do you vouch for the filing?
Analyst: where might this data go, is the basis transparent, do you sign the comparison?
Designer: rights and brand safety, AI credited where the client expects, you ship it.
Knowledge worker: data handling, a note the output was AI-assisted, you trust the numbers.
Reporter: who sees the raw inputs, and the briefing is yours to stand behind.
Content: your publication's disclosure norms, your name on it.
Form-filler: the privacy of the source documents, you submit and you're accountable.
Outreach: recipient-data handling, you vouch for every claim.
Exec: confidentiality of the inputs, you own what you present, AI-assisted or not.

Watch for

Academic integrity is the live wire in this room and it cuts both ways. The defensive read ("students will cheat with AI") is real, but the deck's point is the constructive one: when students have AI too, your own fluency lets you design assessments that ask for authentic learning rather than the thing AI can do in a second. Name both, land on the second. This is also the natural home for the data-safety reassurances for school IT: the floor (Claude on the web, no install) and the ceiling (a local model, nothing leaves the network). (objections.md: IT/compliance; no web portal.)

Augment, don't automate. Keep your judgment at the centre.

The idea

The whole course in one line: build your teaching context, run the 4Ds, and let the describe–discern loop make AI a thinking partner for design, materials and assessment. The framework holds as the tools change.

Say it like this

"Build your context once. Run the four habits. Let the describe–discern loop do the thinking with you, not for you. That's the difference between AI as an autopilot and AI as a thinking partner. If the tools change next year, and they will, this still holds."

Close with

Their task from the start, and the next concrete step. For Praxis, point forward: the Teaching Context document and the 4Ds are the why; the skill file and Claude Code are the how that makes "describe what good looks like, once" permanent on their own machine. End on the habit: "at the end of a session, tell it to look back at what you did badly and save how you fixed it."