This is the opposite of the “worksheet-industrial complex.”

5) Embodiment

Definition

Embodiment (Merleau-Ponty): cognition is not a detached “mind.” Understanding lives in the lived body—perception, action, gesture, spatial intuition, rhythm, tool-use. We learn by doing, not only by describing.

Five points

1) What’s wrong now: education treats thinking as disembodied symbol manipulation

Modern schooling often assumes:

• if students can read/listen, they can understand

• if they can repeat, they know

• “real learning” = silent sitting + abstract symbols

This produces a common failure mode:

• students can recite rules but cannot use them

• they can say words but cannot navigate the phenomenon

Embodiment predicts why: without sensorimotor grounding, concepts stay “floating.”

2) What education needs: “perception-action loops” as the core unit of instruction

A concept becomes real when students repeatedly loop:

• perceive → act → observe feedback → adjust

Examples across domains:

• math: manipulating representations (graphs, diagrams, transformations), not only algebra

• physics: feeling constraints (balance, friction, acceleration) through experiments

• writing: speaking arguments aloud, hearing ambiguity, revising structure

• programming: running code, observing behavior, debugging iteratively

This is not “play for fun.” It’s interactive contact with reality.

3) Dialogue is embodied too: speech, gesture, pacing, live reasoning

A lot of classroom talk is performative Q/A (“guess what’s in my head”). Embodied dialogue is different: it makes thinking visible and manipulable:

• talk while drawing the model

• gesture the causal structure (“this pushes that”)

• point to evidence in the artifact

• slow down and narrate the move (“I’m changing this variable because…”)

Embodied dialogue transforms “explanation” into shared perception.

4) AI in the embodiment frame: AI should create more doing, not more sitting

AI can either worsen disembodiment (more screen, more passive answers) or become a “coach of action.”

Good AI roles:

• micro-experiment generator (quick tests using household items, simple sensors, web data)

• interactive simulator (change variables, observe outcomes; student predicts first)

• skill coach (for speaking, writing, coding, design—iterative feedback loops)

• representation translator (turn verbal ideas into diagrams/checklists; then the student acts)

Design rule:

Every AI interaction should end with an action the student performs and verifies.

5) Future direction: education as studio + lab, not lecture + worksheet

Embodiment implies the future model:

• studio-based learning (make things)

• lab-based learning (test things)

• critique-based learning (discuss artifacts)

• iteration as the normal rhythm

Assessment shifts from:

• “can you answer” → “can you perform, diagnose, improve”

AI scales this by making iterative practice feasible for everyone, not only the top students.

6) Being-in-the-world (Dasein)

Definition

Heidegger: humans are not spectators observing a world; we are already involved—we care, we cope, we use tools, we pursue goals, we face risks. Meaning is practical before it is theoretical.

Five points

1) What’s wrong now: school pretends learners are not living real lives

Many school tasks are “as-if” tasks:

• write an essay no one will read

• solve a problem no one cares about

• memorize facts without consequence

• comply with procedures detached from agency

Students experience: “This is not my world.”
Heidegger would say: education breaks because it ignores the student’s mode of being: practical involvement, care, identity, reputation, fear, purpose.

2) What education needs: re-anchor learning in care, stakes, and responsibility

Not drama—real stakes in an age-appropriate way:

• students build something others rely on (a guide, a model, a tool, a briefing)

• students advise a decision (policy memo, design choice, budget tradeoff)

• students test claims that matter (local data, real controversies, measurable outcomes)

When learners are “in it,” attention becomes natural. You don’t need motivational tricks.

3) Dialogue changes when it’s about lived commitments

Dialogue becomes real when students are defending or improving something they own:

• “Here is our proposed solution—attack it.”

• “Which risk did we miss?”

• “What evidence would justify choosing Option A over B?”

• “What happens if our model is wrong?”

This is dialogue as coordination for action, not talk for grades.

4) AI in the being-in-the-world frame: AI as project partner + decision rehearsal

AI can powerfully support “being-in-the-world learning” by helping students operate like real practitioners:

• role-play stakeholders (customer, regulator, patient, voter)

• simulate consequences and second-order effects

• generate risk registers and mitigation options

• help students prepare interviews, surveys, experiments

• serve as “devil’s advocate” against their plan

But you must prevent AI from becoming the “doer.” The student must remain the agent.

A good pattern:

• student proposes → AI critiques → student revises → student tests in reality → student reports evidence

5) Future direction: education for agency under complexity

The AI era punishes passive competence. The scarce resource becomes:

• making sense of messy situations

• choosing what to do next

• coordinating with others

• evaluating claims and tools

“Being-in-the-world” education trains students to navigate real complexity with judgment and responsibility—exactly what pure content schooling fails to produce.

7) Situatedness / Contextuality

Definition

Meaning is situated: understanding depends on context—goals, constraints, tools, culture, framing. Knowledge is not a universal “thing” you possess; it is a capability you can deploy in situations.

Five points

1) What’s wrong now: schooling trains brittle knowledge

A classic failure: students do well in the classroom but cannot transfer.

Why? Because they learned:

• procedures tied to one format (“this worksheet type”)

• definitions without usage conditions

• answers without sensing relevance

Situatedness predicts transfer failure: knowledge wasn’t learned as situational choice-making.

2) What education needs: variation, contrast, and conditions-of-use

To learn a concept, students must see:

• where it applies

• where it doesn’t

• how it changes under constraints

Concrete practices:

• “near-miss” examples (almost fits, but fails)

• changing constraints (time, resources, uncertainty)

• multiple representations (text, diagram, equation, simulation)

• scenario swaps (same concept in different domains)

This is how students build “when-to-use” intelligence, not just “how-to-do” memory.

3) Dialogue becomes “context negotiation”

Situated dialogue sounds like:

• “In which context is your solution valid?”

• “What constraint breaks your approach?”

• “What hidden variable matters here?”

• “What changes if we optimize for speed vs safety vs cost?”

This trains a major AI-era capability: conditional reasoning and tradeoff navigation.

4) AI in the situatedness frame: generator of contexts + adversary of overgeneralization

AI is extremely useful for:

• generating many contexts quickly

• producing edge cases

• offering counterexamples

• stress-testing a student’s claim

Powerful constraint:

Require students to state “validity conditions” for every explanation AI helps with.

AI prompt pattern:

• “Give 5 contexts where this applies, 5 where it fails, and 5 tricky edge cases.”
  Then the student must:

• explain why each is in that bucket

• propose a test for the edge case

5) Future direction: context-first competence (especially with AI)

In the AI era, anyone can get a plausible answer. The differentiator is:

• knowing whether it applies here

• what assumptions it relies on

• what failure modes exist

• how to adapt it to constraints

Situatedness becomes the backbone of:

• AI literacy

• decision-making

• real-world problem solving

8) Temporality (Lived time)

Definition

Understanding unfolds in time. Meaning is not captured instantly; it forms through cycles: exposure, confusion, practice, re-seeing, integration. There is also “kairos”—the right moment when something clicks.

Five points

1) What’s wrong now: education is paced like a factory, not like learning

School often moves as if:

• everyone should learn at the same speed

• understanding is immediate if explained clearly

• curriculum coverage matters more than integration

Results:

• shallow learning

• anxiety and shame for “slow” learners

• forgetting after exams

• no time for synthesis

Phenomenology says: you can’t force lived understanding into industrial time.

2) What education needs: spiral, revisit, and integration rituals

Temporality implies:

• you must return to ideas later, after new experiences

• you must re-encounter concepts at higher resolution

Practical design:

• short retrieval cycles (days)

• application cycles (weeks)

• synthesis cycles (months)

• “capstone re-seeing” where old ideas are reinterpreted

Also: build explicit “integration moments”:

• “What changed in your view since last month?”

• “What did you misunderstand earlier—and why?”

3) Dialogue should be staged across time, not only in the moment

A strong method:

• students commit to a model today

• revisit the same model after experiments

• compare early vs later thinking

This creates:

• intellectual honesty

• measurable growth

• revision skill (the core of real intelligence)

Education should normalize:

• “I was wrong, and here is how I updated.”

4) AI in the temporality frame: personal pacing + long-horizon coaching

AI can be a continuous tutor that:

• tracks misconceptions over weeks

• schedules spaced practice

• revisits earlier errors with new examples

• prompts reflection at the right time

• adapts pacing without stigma

But you must avoid the “instant answer = instant mastery” illusion.
So you structure AI use as:

• delayed reveal (student predicts first)

• forced retrieval (student explains without seeing notes)

• iterative refinement (AI critiques, student revises)

• spaced repetition (AI returns to the idea later)

5) Future direction: education as progression of capabilities, not synchronized content

Temporality implies the future isn’t:

• “everyone completes Unit 7 by Friday”
  but:

• “everyone reaches capability milestones, with different trajectories”

With AI, you can finally do this at scale:

• individualized learning paths

• continuous formative feedback

• portfolio evidence of growth

• mastery by repeated integration, not one-time exposure

9) Horizon

Definition

A horizon is the background of expectations, meanings, and possibilities that frames what a learner can even notice, ask, or imagine. Every experience comes with “more than is currently given”: implicit context + anticipated futures.

Five points

1) What’s wrong now: schooling narrows horizons instead of expanding them

Many students leave school with a shrinking sense of possibility:

• “There’s one right way.”

• “My job is to guess what the teacher wants.”

• “Big questions are dangerous; small answers are safe.”

• “I’m not that kind of person.”

Phenomenologically, this is catastrophic: if your horizon is narrow, you literally cannot see opportunities for inquiry, creativity, or agency.

2) What education needs: horizon-expansion as an explicit goal

Horizon expansion means enlarging:

• what counts as a good question

• what kinds of explanations are imaginable

• what methods are available (experiment, modeling, dialogue, critique)

• what futures a student can picture themselves inhabiting

Concrete moves:

• “Here are 5 different ways professionals would approach this.”

• “Here are 3 competing frames for the same situation.”

• “Here are the next questions this opens.”

3) Dialogue is the tool that stretches horizons safely

Good dialogue exposes students to “possible worlds” without forcing certainty:

• “What else could be going on?”

• “What would someone with a different goal see?”

• “What are we not allowed to assume?”

• “What becomes possible if this constraint disappears?”

A horizon expands when a student experiences:

• their interpretation isn’t the only one

• ambiguity is workable

• alternative futures can be reasoned about

4) AI in the horizon frame: generator of perspectives + futures, not a replacement for choice

AI can massively expand horizons by generating:

• alternative hypotheses and frames

• stakeholder viewpoints

• scenario trees and second-order effects

• “next question” maps

• analogies to distant domains

But the educational requirement is:

Students must choose and justify which horizon to operate in.

Good pattern:

• AI proposes 6 frames → student selects 1 → student runs an experiment or builds an argument within that frame → student compares results with another frame later.

5) Future direction: education for possibility-navigation

In an AI era, the limiting factor is not answers. It’s:

• selecting which questions matter

• selecting frames that generate leverage

• seeing option space

• anticipating consequences

So the future curriculum should train:

• framing skill

• scenario thinking

• “question-generation competence”

• the ability to deliberately expand and then narrow horizons through tests

10) Pre-reflective Experience (Tacit knowing)

Definition

Pre-reflective experience is what you “know” before you can say it: tacit pattern sense, bodily skill, intuitive recognition. We often grasp something implicitly long before we can articulate it.

Five points

1) What’s wrong now: school over-rewards verbalization and under-trains tacit skill

School privileges:

• definitions

• explanations

• written output

• explicit steps

But many real competencies grow as tacit perception first:

• sensing a flawed argument before naming the flaw

• feeling that a result is implausible

• recognizing a pattern in data

• hearing ambiguity in a sentence

When education ignores tacit knowing, students:

• become brittle “explainers” without judgment

• lose intuition instead of refining it

• can’t diagnose errors unless they match a known template

2) What education needs: “intuition → articulation → verification” loops

A powerful structure:

• Intuition: “What do you sense is happening?”

• Articulation: “Name it. What’s the pattern?”

• Verification: “How would you test it? What evidence would decide?”

This preserves intuition while preventing it from becoming superstition.

Practical methods:

• prediction before instruction

• estimation practices (“ballpark first”)

• error-spotting drills

• “which solution feels wrong—and why?”

3) Dialogue is how tacit understanding becomes shareable and improvable

Pre-reflective knowledge becomes educational when students can:

• externalize it into language, diagrams, demonstrations

• receive critique

• compare intuitions with others

• refine their “felt sense” into disciplined judgment

Dialogue prompts:

• “Point to where it breaks.”

• “What detail triggered your suspicion?”

• “Can you demonstrate it rather than explain it?”

• “What would change your mind?”

4) AI in the tacit frame: a mirror that forces articulation and tests intuition

AI can help students convert tacit sense into explicit, testable claims by:

• asking for reasons behind a hunch

• offering candidate labels (“Is it contradiction, equivocation, missing variable, base-rate neglect?”)

• generating minimal tests

• producing counterexamples to stress intuition

Design rule:

AI should never accept “I just feel it” as final; it should help turn feelings into hypotheses.

5) Future direction: disciplined intuition as a core AI-age advantage

When AI outputs fluent text, humans need:

• the ability to sense when something is off

• the ability to probe assumptions quickly

• the ability to test rather than trust

So education should explicitly train:

• calibrated intuition

• anomaly detection

• uncertainty awareness

• fast experimental thinking (“what quick check would validate this?”)

11) Interpretation / Hermeneutics

Definition

Hermeneutics is the theory of interpretation: we never receive “pure facts” without a frame. Meaning is always interpreted through prior assumptions, language, culture, and purpose.

Five points

1) What’s wrong now: school pretends meaning is automatic and texts are transparent

Students are often trained to treat:

• reading as decoding

• listening as absorption

• “correct interpretation” as a single static thing

This breaks in real life, where:

• arguments manipulate

• data is framed

• narratives compete

• incentives distort meaning

Without interpretive skill, students become easy targets for misinformation—especially amplified by AI.

2) What education needs: interpretation as a method with explicit steps

Teach interpretation as disciplined practice:

• identify the speaker’s goal

• map the argument structure

• separate claim vs evidence

• find ambiguities and missing premises

• compare alternative readings

• test the reading against the whole context (part–whole loop)

This is not “subjective opinion.” It’s a craft.

3) Dialogue is the engine of interpretive rigor

Interpretation improves when interpretations collide:

• “Show me where the text implies that.”

• “What would the author disagree with in your reading?”

• “What alternative reading explains the same lines better?”

• “Which reading predicts what comes next?”

Classroom dialogue should shift from:

• “What did the author mean?” (guessing)
  to:

• “What readings are possible, and which is best supported?” (reasoning)

4) AI in the hermeneutics frame: multi-reading generator + argument mapper + bias detector

AI can support interpretation by:

• producing multiple plausible readings

• mapping arguments into premises/conclusions

• flagging loaded terms and rhetorical devices

• generating “what would count as evidence” prompts

• proposing questions to ask the author (simulated interview)

But again: the student must decide.
Use patterns like:

• AI gives 3 interpretations → student defends 1 with textual evidence → AI attacks it → student revises.

5) Future direction: interpretive literacy as civilization infrastructure

In an AI media environment, everyone will be surrounded by:

• persuasive synthetic narratives

• plausible but distorted summaries

• “evidence-looking” claims

Education must therefore train:

• interpretive discipline

• rhetorical and framing awareness

• evidence standards

• cross-checking habits

Hermeneutics becomes a survival skill.

12) Intersubjectivity

Definition

Intersubjectivity is the shared world of meaning between persons. Understanding is not purely private; it is formed, stabilized, and corrected through social exchange, trust, recognition, and shared standards.

Five points

1) What’s wrong now: education treats learning as individual performance, not shared sense-making

Typical schooling:

• isolates students

• penalizes collaboration

• grades individual output

• creates competition for status

This undercuts the real mechanics of learning:

• we learn by explaining, arguing, imitating, correcting

• we calibrate meaning socially

• we build standards through community

When intersubjectivity is suppressed, students lose the most powerful correction mechanism: other minds.

2) What education needs: classrooms as communities of inquiry

A community of inquiry has:

• shared norms: “claims need reasons,” “revision is respected”

• distributed cognition: students build knowledge together

• real roles: skeptic, explainer, tester, summarizer, connector

• collective artifacts: shared models, living documents, experiment logs

Education improves when the “unit” is not the isolated student but the thinking group.

3) Dialogue is not optional—it’s the core medium of shared truth

Intersubjective dialogue should be structured, not chaotic:

• rules for critique without humiliation

• protocols for turn-taking and steelmanning

• explicit evidence standards

• “disagreement maps” that track where people differ

This trains:

• cooperative truth-seeking

• epistemic humility

• conflict navigation

• leadership through clarity

4) AI in the intersubjective frame: amplify group dialogue, don’t replace it

AI can help groups by:

• summarizing discussion and extracting claims

• tracking disagreements and unresolved questions

• generating tests to resolve disputes

• ensuring quieter voices are surfaced (“Who hasn’t spoken?” prompts)

• providing neutral “judge” functions (argument structure, missing premises)

But if AI becomes the authority, intersubjectivity collapses.
Design rule:

AI is a facilitator and mirror, never the final arbiter.

5) Future direction: hybrid intelligence—humans + AI as a thinking ecology

The future classroom can become a “hybrid intelligence lab”:

• students collaborate with each other

• AI facilitates, stress-tests, and personalizes practice

• truth emerges from dialogue + experiment + evidence

This is exactly what modern education rarely achieves: a scalable culture of rigorous inquiry.

13) Empathy

Definition

In phenomenology, empathy isn’t “being nice.” It’s the capacity to access another person’s experience as experience—to grasp how the world appears from their standpoint (their fears, aims, constraints, meanings). It’s how intersubjectivity becomes precise instead of vague.

Five points

1) What’s wrong now: education trains viewpoint collapse

School often treats perspectives as:

• irrelevant (“just learn the facts”)

• performative (“write what the teacher wants”)

• moralized (“agree with the ‘right’ view”)

Students don’t learn how to reconstruct a worldview. They learn compliance or tribal argument. That destroys dialogue quality and makes disagreement unproductive.

2) What education needs: empathy as a method (reconstruction, not agreement)

Teach empathy as a disciplined procedure:

• Reconstruction: “What is the other person trying to protect or achieve?”

• Constraint mapping: “What constraints make their choice rational?”

• Value inference: “What do they prioritize?”

• Evidence standards: “What would they accept as proof?”

• Prediction test: “If I truly understand them, I can predict their next move/argument.”

Empathy becomes a cognitive tool for truth-seeking and coordination.

3) Dialogue improves when empathy is enforced structurally

Add dialogue rules like:

• steelman before critique

• summarize their position to their satisfaction

• separate values disagreements from facts disagreements

• ask “What would change your mind?” genuinely

This transforms the classroom from debate theatre into collaborative inquiry.

4) AI in the empathy frame: perspective simulator + misunderstanding detector

AI can help by:

• generating plausible stakeholder perspectives

• role-playing an opponent who has coherent values

• highlighting where a student caricatured the other side

• suggesting clarifying questions that reduce conflict

But the student must still do real reconstruction.
Design rule:

AI can generate candidates, but students must validate them against real humans, texts, or evidence.

5) Future direction: empathy as core AI-age competence

In an AI world:

• social fragmentation rises

• persuasion becomes cheap

• misunderstandings scale fast

Empathy becomes infrastructure for:

• collaboration

• governance

• negotiation

• leadership

• conflict de-escalation

Education should treat it as “applied cognition,” not “soft skills.”

14) Intentional Arc / Skill Incorporation

Definition

Merleau-Ponty’s intentional arc: as skills develop, the whole field of perception and action reorganizes. Tools become extensions of the body. A novice sees noise; an expert sees structure and can act fluidly.

Five points

1) What’s wrong now: education over-indexes explanation and under-builds incorporation

Students are often asked to talk about competence instead of becoming competent:

• lots of “definitions”

• few real reps

• little feedback

• weak iteration loops

So the intentional arc never forms. Students stay in brittle “step-following” mode.

2) What education needs: deliberate practice + tight feedback loops

Skill incorporation requires:

• high-quality repetitions

• immediate feedback

• progressive difficulty

• attention to error patterns

• reflection that extracts principles

This applies to:

• reasoning

• writing

• math

• coding

• experimentation

• collaboration

Key shift:

Curriculum should be organized around “capabilities built by practice,” not “topics covered.”

3) Dialogue should track skill growth, not just correctness

Dialogue that builds incorporation sounds like:

• “Show your move.”

• “Where did it start to go wrong?”

• “What cue did you miss?”

• “What would you do first next time?”

This makes learning about improving perception-action coupling, not winning.

4) AI in the intentional arc frame: infinite coach, not infinite answer

AI is excellent at:

• generating practice sets tuned to weaknesses

• giving immediate formative feedback

• offering alternative strategies

• tracking a student’s error signature over time

• replaying “similar but different” tasks for transfer

But: if AI supplies final products, incorporation dies.
Rule:

Use AI to create reps + critique, never to remove the learner’s performance.

5) Future direction: “AI-assisted mastery trajectories”

You can redesign schooling into mastery pathways:

• students progress when capabilities stabilize

• AI provides adaptive drills and feedback

• teachers focus on motivation, meaning, group inquiry, and project design

• assessment becomes performance evidence across time

This is the practical way to escape one-size-fits-all pacing.

15) Authenticity / Ownership

Definition

Authenticity (Heidegger and later existential phenomenology) is not “be yourself” as a slogan. It’s owning your possibilities—relating to your learning and life as something you choose and take responsibility for, rather than something imposed.

Five points

1) What’s wrong now: schooling trains inauthenticity as a survival strategy

Students learn:

• perform for grades

• hide confusion

• mimic expected language

• optimize for evaluation

• outsource meaning to authority

This creates “learned non-ownership.”
Students may succeed academically and still feel:

• alienated

• passive

• incapable of initiating real projects

2) What education needs: agency structures, not motivational speeches

Authenticity emerges from structural conditions:

• choice within constraints (real options)

• responsibility for outcomes

• visible impact (work matters to someone)

• permission to revise identity (“I’m becoming capable”)

• environments where honesty about confusion is safe

3) Dialogue must shift from “answer recitation” to “position-taking”

Ownership grows when students must:

• make claims

• justify them

• revise them publicly

• choose methods

• explain tradeoffs

Dialogue prompts:

• “What do you believe and why?”

• “What would you do next?”

• “What did you choose not to do—and why?”

• “What standard are you using to judge success?”

That’s agency training.

4) AI in the authenticity frame: personalized pathways + reflective mirror

AI can support ownership by:

• helping students set goals and plans

• reflecting their progress back as a narrative

• suggesting projects aligned with interests

• offering multiple ways to approach the same capability

• prompting metacognition (“What are you optimizing for?”)

But AI can also destroy authenticity by becoming the student’s “voice.”
So require:

• voice constraints (student must speak in their own words)

• provenance (what is yours vs assisted)

• oral defense and live performance

• portfolio evidence of iteration

5) Future direction: identity formation through real work

Future education should create people who:

• initiate

• build

• test

• collaborate

• revise

• take responsibility

AI should free time from clerical work so students can do real work:

• experiments

• projects

• investigations

• designs

• community contributions

Authenticity becomes a measurable outcome: “Can you author a path?”

16) Alienation / Reification

Definition

Reification is when living meaning turns into dead objects. In education: learning becomes grades, procedures, tokens, compliance—while the real phenomenon (curiosity, understanding, capability) disappears. This is the phenomenology of “school feels pointless.”

Five points

1) What’s wrong now: education is optimized for metrics, not meaning

Common reifications:

• learning = test score

• intelligence = speed of recall

• writing = formula

• science = facts

• school = credential factory

Students adapt rationally:

• maximize grades

• minimize risk

• avoid deep confusion

• outsource thinking when possible

This isn’t laziness; it’s system incentives.

2) What education needs: de-reification through inquiry and consequence

To restore meaning:

• tasks must connect to real questions

• work must produce artifacts with audiences

• evaluation must reward thinking quality and revision

• students must experience “knowledge as power to act”

Core mechanism:

Replace token incentives with epistemic incentives: curiosity, prediction, testing, improvement.

3) Dialogue is the antidote to reification

Reification thrives in monologue and bureaucracy.
Dialogue restores:

• living questions

• active disagreement

• shared standards

• real-time correction

• human recognition (“I see your mind working”)

But the dialogue must be about evidence and models, not status.

4) AI risk: maximal reification (instant output, zero meaning)

AI can intensify reification brutally:

• students submit perfect-looking work with no ownership

• teachers grade artifacts disconnected from student capability

• credentials lose signal

• learning collapses into “content generation”

AI opportunity: de-reification via experiment and critique:

• AI generates hypotheses, counterexamples, tests

• students run experiments and defend conclusions live

• assessment focuses on process evidence and performance

5) Future direction: assessment redesign (the real bottleneck)

If you don’t change assessment, AI will force reification.
The future needs:

• oral defenses

• live problem solving

• project portfolios with iteration logs

• peer critique records

• experiment notebooks

• “validity conditions” statements for claims

• evaluation of questioning and testing skill

In short:

Grade what AI cannot fake easily: judgment, experimentation, dialogue, revision, and real agency.