Day 1 · Dimension 1 · ~4 hours
Foundations of Extended Cognition
Mapping the Extension System — Cognition Before Action
D1 measures what you understand about AI's nature before any tool use begins. The first two tasks are completed with no AI access — your unaided conceptual map is the primary evidence. The critical leap from Level 2 to Level 3 is whether your understanding is subject-specific (naming lesson activities, class characteristics, curriculum content) or generic.
Level 1–2
Entries are generic. A/B/C classifications reference tool names, not cognitive functions. Irreplaceability reasons say “teachers build relationships” without connecting to the subject.
Level 3
Each boundary ties a specific AI capability to a specific lesson activity in this unit. Irreplaceability reasons name the type of subject-specific judgment.
Level 4
Identifies second-order limits (AI’s inability to identify its own limits). Vendor annotation reveals structural assumptions about teacher role embedded in marketing language.
4 TASKS What you will produce on Day 1
01
Extension System Analysis
Pick five real teaching activities from your current unit — not hypothetical ones, real ones you’re planning to run this term. Classify each activity into one of three extension types:
- A-type (Amplification) — AI makes something you were already doing faster, wider, or more consistent. The activity’s pedagogical logic is unchanged; AI just expands its reach.
- B-type (Scaffolding) — AI reaches a student you couldn’t reach with the same effort, usually by providing personalised support you don’t have the bandwidth to give in real time.
- C-type (Co-creative) — AI does something neither you nor the student could have made alone. The rarest and most powerful category.
For each of your five activities, write:
- The specific learning objective it serves — in the vocabulary of your curriculum, not in generic language.
- Which extension type you’d classify it as, and why that classification and not one of the other two.
- What you would do if AI were removed tomorrow. If “what I’d do without AI” and “what I’m doing with AI” are the same thing, it’s A-type. If the answer is “I couldn’t run this activity at all without AI,” you may be looking at a C-type — or a dependency you should reconsider.
02
Subject-Specific Boundary Map
A three-column table. Columns: technical limits, contextual limits, value-judgment limits. At least five rows per column. Each row names:
- A specific AI capability (not “AI” in general — the specific thing AI does)
- A specific lesson activity in your unit where that capability is relevant
- Why the limit matters in this subject, at this level, for this class
Technical limit example: “Claude cannot play a physical musical instrument. In our Year 9 clarinet embouchure unit, every demonstration of tongue placement still has to come from me or another human player.”
Contextual limit example: “Claude will produce feedback on a Year 10 Macbeth essay, but its feedback rewards fluent phrasing over argumentative structure — which is exactly the skill my rubric is trying to develop.”
Value-judgment limit example: “The decision to flag a student’s work as possibly copied from AI is a decision with consequences for that student’s relationship with me, their record, and sometimes their family. I cannot delegate that decision to a detection tool.”
This is the hardest task of Day 1. Do it slowly.
Evaluation criteria: D1.2 Recognition of Extension Boundaries
03
Teacher Irreplaceability Declaration
Five decisions in this unit that cannot be delegated to AI. Each with a reason that sounds like your subject. A history teacher’s reasons should read like history. A PE teacher’s reasons should read like PE. Generic reasons (“teachers build relationships”) don’t tell us anything yet — everyone writes those, and they’re true but empty.
Aim for reasons that only you could write.
Evaluation criteria: D1.3 Anchoring of Teacher Agency
04
Mental Model Audit
Find three pieces of AI-in-education marketing — vendor websites, ed-tech conference talks, glossy explainers. Annotate each one. Mark accurate claims vs. overstatements about AI’s nature or teacher agency. Quote the exact phrases you’re marking.
You’re looking for two specific kinds of slippage:
- Capability inflation (“AI understands student thinking”) — claims that describe AI as if it had a mind, when what it has is pattern correlation over text.
- Role erosion (“AI personalises learning for every student”) — phrases that quietly redefine what “the teacher does” in ways that erase judgment from the picture.
Finding these takes practice. You’ll start noticing them everywhere once you’ve tagged a few.
Evaluation criteria: D1.1 Extension-System Understanding3 SUB-COMPETENCIES Evaluation criteria for Day 1
D1.1
Extension-System Understanding
A/B/C foundational
1 — Nascent
Names AI tool categories (chatbot, search, generator) but conflates AI capability with independent intelligence. Cannot distinguish A-, B-, and C-type extensions. Describes AI as a “smart assistant” without reference to training data or human knowledge origins.
2 — Developing
Explains basic operating principles of one or two AI types; partial grasp of A- vs B-type. Conflates B- and C-type extensions. Acknowledges AI uses training data but cannot draw pedagogical implications.
3 — Proficient
Accurately explains operating mechanisms of generative AI, recommendation algorithms, and intelligent agents. Reliably distinguishes A/B/C extension modes in concrete teaching scenarios with principled reasoning. Articulates that AI outputs are products of human knowledge encoding, not independent creation, and applies this in practice.
Anchor: Classifications consistent across ≥5 activities; each cites a pedagogical function, not a tool name.
4 — Advanced
Independently maps novel AI tools onto A/B/C before deployment with explicit justification. Develops shared mental models for colleagues; identifies when popular descriptions misrepresent AI’s nature. Uses extension system cognition to evaluate vendor claims and inform institutional adoption.
500–700 words. Classify ≥5 teaching activities from your current unit into A/B/C extension types. Each classification must cite the specific learning objective, class level, and what you would do if AI were removed.
- Activities are real (from your current unit), not hypothetical
- Each A/B/C label is justified by the cognitive function AI serves, not the tool name
- The “if AI removed” counterfactual reveals whether you understand what AI actually contributes
- If two activities use the same tool but get different types, the reasoning explains why
Annotate 3 AI education marketing texts: mark accurate claims vs. overstatements about AI’s nature or teacher agency. Cite specific phrases.
- Annotations target specific phrases, not vague “this is misleading”
- Distinguishes factually wrong claims from framing overclaims
- At Level 3+, reveals understanding of how marketing language shapes teacher identity
D1.2
Recognition of Extension Boundaries
A/B/C foundational
1 — Nascent
Aware that AI makes errors but describes limits only in technical terms (hallucination, outdated data). Cannot identify contextual or value-judgment limits. Treats AI limits as the AI’s own deficiencies rather than features of how the extension system is constituted.
2 — Developing
Identifies technical and some contextual limits when prompted; inconsistently applies to specific scenarios. Occasionally identifies value-judgment limits but cannot reliably distinguish AI’s functional range from areas requiring human judgment.
3 — Proficient
Systematically identifies technical, contextual, and value-judgment limits before deploying AI. Explains that limits derive from how the extension system is built. Applies boundary analysis routinely for each A/B/C type.
Anchor: All 3 limit types present; each row ties a specific AI capability to a specific lesson activity.
4 — Advanced
Proactively maps extension boundaries for unfamiliar AI systems using a transferable framework. Identifies second-order limits (limits on AI’s ability to identify its own limits). Contributes to institutional deployment policies based on boundary analysis.
p1t2Subject-Specific Boundary Map
Three-column table — technical limits, contextual limits, value-judgment limits. Each row names a specific AI capability, a specific lesson activity, and why the limit matters in this subject at this level.
- Three distinct limit types are present and correctly distinguished
- Each row is grounded in a real lesson activity, not a hypothetical
- Limits are specific enough to be falsifiable (not “AI doesn’t understand nuance”)
- At Level 3+, contextual limits reference this class’s characteristics
D1.3
Anchoring of Teacher Agency
Critical in C-type
1 — Nascent
Acknowledges teacher importance in general terms but cannot specify which roles are non-substitutable. Defers to AI outputs without rationale; cognitive offloading is unrecognised. Professional identity noticeably destabilised by AI capability.
2 — Developing
Names non-substitutable roles (relationship, emotional support) but grounds them in convention rather than principled reasoning. Sometimes overrides AI but rationale is implicit. Can identify cognitive offloading in others but not consistently in own practice.
3 — Proficient
Articulates non-substitutable roles grounded in value judgment, relational responsibility, and developmental purpose. Consistently exercises professional judgment when AI is involved. Maintains stable professional identity anchored in teacher-as-purposive-core.
Anchor: ≥5 roles named; each has a subject-specific reason that sounds like this teacher’s discipline.
4 — Advanced
Develops explicit frameworks for preserving teacher agency in AI-rich environments. Analyses systemic risks (algorithmic management, automated grading dominance) and advocates for structural protections. Contributes to policy discussions on teacher identity.
p1t3Teacher Irreplaceability Declaration
≥5 decisions in this unit that cannot be delegated to AI, each grounded in a subject-specific or relational reason. A history teacher’s reasons must sound like history.
- Each decision names a specific judgment call from this unit
- Reasons could not be copied to a different subject without rewriting
- Articulates what would be lost if this decision were delegated