Calibrate and fade project scaffolds

1

Research

Why this matters, and what good looks like

A scaffold is temporary support for work students cannot yet do alone. Fading means reducing that support after a teacher reviews evidence of independence. The retained EEF landing-page passages concern planning, monitoring and evaluation in subject learning. They do not directly establish fading, tiers or review schedules. PBLWorks supplies professional guidance; UNESCO supplies general AI guidance. None establishes AI benefit here. The tiered menu and review triggers below are authorial design choices. The catalogue strength remains an editorial classification.

EEF describes explicit planning, monitoring and evaluation strategies embedded in subject learning.
PBLWorks connects varied instructional supports with project goals and student independence.
UNESCO retains human agency and cautions that fluent AI output cannot be assumed accurate.

Where the evidence comes from

Strong for scaffolding and self-regulation; AI benefit unvalidated.

2

Workflow

Brief it, steer it, check it

Identify one observed barrier

Bring one task, criteria and nonidentifying observations. Optional D07 access plans and C01/C03 evidence can supply these, or use your own notes. Keep private support records outside AI. Bound one support trial and review; stop beyond 60 active minutes.

Prompt 1 · Map the barrier

CONTEXT
Task, unchanged learning demand and verified subject material: [task and sources]
Nonidentifying observed task evidence and teacher diagnosis: [observations and diagnosis]
Teacher-approved access constraints and current scaffold or none: [access and support]
REQUEST
Map the teacher-identified barrier to observed evidence and the subject demand that must remain. Identify missing material before drafting supports.
QUALITY BAR
Do not diagnose learners, infer needs from profiles/stereotypes or request personal data. Separate observation from teacher interpretation. Do not lower demand or solve the project. If direct observations, teacher diagnosis, demand or subject material are missing, return only HOLD and the missing inputs. Unknown outcomes are UNOBSERVED.
FORMAT
Return a table: barrier ID | exact observation | teacher interpretation | unchanged demand | access constraint | current support | unknowns. Then list teacher decisions required.

Draft support levels

Approve or correct the evidence map yourself. Ask for lighter and fuller temporary supports for the same barrier, with an independent route. Check every subject statement against your sources before choosing; levels describe support, never learner ability.

Prompt 2 · Draft the menu

CONTEXT
Exact step 1 output: [step 1 output]
Teacher approval and edits to the map: [map decision]
Task, verified subject sources and access constraints: [task sources and access]
REQUEST
Draft a three-option menu: independent route, light checklist, fuller frame or partial model. Link each option to the same observed barrier and retain the same learning demand. Include access notes, a demand-preserving extension, a proposed independence check and a proposed review/fade trigger.
QUALITY BAR
Do not assign options to students or ability groups. No stereotypes, personal profiles, answer scripts or complete project solutions. Cite supplied subject locators for every subject claim/example. Missing approval or source support means HOLD. A trigger prompts human review, never automatic removal. Preserve required access arrangements; fading instructional prompts must not withdraw accommodations. Mark all outcomes UNOBSERVED.
FORMAT
Return a table: option | barrier ID | trigger for teacher consideration | model/checklist/frame | subject locator | access notes | extension | proposed independence check | proposed review/fade trigger | outcome. Follow with teacher selection and accuracy checks needed.

Plan and record a short trial

Select and correct a support after checking accuracy and demand. Let students try it and attempt a comparable task with fewer instructional prompts, preserving access. Record what happened and hear students' explanations before deciding whether to retain, adjust or reduce.

Prompt 3 · Record the trial

CONTEXT
Exact step 2 menu: [step 2 output]
Teacher-selected support, accuracy checks, edits and approved trial/review plan: [trial decision]
Nonidentifying trial and independent-attempt observations, student explanation and teacher interpretation, or UNOBSERVED: [trial evidence]
REQUEST
Record the selected support and observed effect on access, learning demand and independence. Identify gaps for a human review. Preserve attribution to observation, student explanation or teacher interpretation.
QUALITY BAR
Do not infer success, diagnose need or decide fading. Missing selection or accuracy check means HOLD. Outcomes without observed evidence stay UNOBSERVED. Keep required access arrangements. Do not turn the proposed trigger into an automatic decision or invent a review date. If the support supplies the answer, flag replacement for teacher decision.
FORMAT
Return a table: selected option | supplied accuracy check | support used | task evidence | student explanation | teacher interpretation | independence outcome | review point | human decision needed. Then list holds and UNOBSERVED outcomes.

Audit the teacher's revision

Choose retain, adjust or reduce from the trial evidence. Record a review point, even when evidence is missing. Audit your selected menu and trial record, then make the release decision yourself. Future reviews require fresh observations.

Prompt 4 · Audit the support decision

CONTEXT
Exact step 3 output and original task/criteria: [step 3 and criteria]
Teacher-selected revised menu, edits, retain/adjust/reduce decision and review point: [selected menu and decision]
Versioned source checks, access requirements and trial evidence: [audit evidence]
REQUEST
Compare the teacher-selected result with observed evidence, source checks and unchanged demand. Flag answer scripts, lost access, unsupported independence claims and automatic fading. Do not rewrite the result.
QUALITY BAR
You audit supplied text only. Teacher diagnosis, selection, fading and release remain human decisions. If selection or evidence is missing, return HOLD. Unknown outcomes stay UNOBSERVED. No profiles, inferred needs, lowered expectations or automatic support removal.
FORMAT
Return a table: item | exact selected text | supplied evidence/version or MISSING | concern | human action. Finish with Unresolved holds and Human release decision required. If no concern is found, say only that no concern was found in the supplied comparison.

Check before you use it

Is each support tied to an observed task barrier and a teacher diagnosis, without individual profiles?
Have you checked subject claims and examples against your original materials?
Does every option preserve learning demand, inquiry and required access arrangements?
Have students tried the support and explained a comparable attempt with fewer instructional prompts?
Are unknown outcomes UNOBSERVED and any answer scripts flagged for replacement?
Have you personally chosen retention, adjustment or reduction and recorded the next review point?

Stop rule

Replace scaffolds that become scripts or shortcuts; never infer needs from stereotypes or personal data.

3

Reuse it

So the next one takes minutes

Suitable

Build a skill

Save this as a reusable instruction you can run on any project.

The bounded sequence can be reused with new teacher materials and fresh human decisions.

Not suitable

Build an agent

Chain the steps, with a checkpoint where you approve.

Original evidence checks and consequential support or use decisions require direct human control in this bounded workflow.