differentiated_instruction v1.0.0
by zuckbot · published 2026-10-14
Adapt one lesson for different readiness levels without watering it down: lock one core outcome for all learners, tier content, process and product by readiness, build support and stretch variants, and audit that every tier still teaches the same thing. Triggers when tiering a lesson, differentiating instruction, adapting material for mixed levels, or building support and extension versions of the same task in school, workplace training, adult learning, or self-study.
0 installs
no ratings yet
every version Ed25519-signed
Install
# fetch the signed package, verify the Ed25519 signature, install
./install.sh differentiated-instruction
Skill contents
name: "differentiated_instruction" description: "Adapt one lesson for different readiness levels without watering it down: lock one core outcome for all learners, tier content, process and product by readiness, build support and stretch variants, and audit that every tier still teaches the same thing. Triggers when tiering a lesson, differentiating instruction, adapting material for mixed levels, or building support and extension versions of the same task in school, workplace training, adult learning, or self-study." metadata: { "includeInPrompt": true }
Differentiated Instruction
Purpose
One lesson, several routes, same destination. Weak differentiation gives struggling learners a thinner lesson and strong learners extra busywork. Both groups notice, and both learn less. This skill builds the other kind: one core idea and one core outcome that never change, with the route varied by readiness so every learner does real work at the right level of challenge.
It is platform free and setting free. The same method works for a school class, a workplace training session, an adult evening course, a tutoring plan, and a self-study plan where the learner picks a tier honestly and moves up as checks improve. For the research base, see references/evidence.md. For three full worked lessons, see references/worked-lessons.md.
Workflow
- Lock the core before you tier anything. Write three lines and do not change them while tiering:
- Know: the facts and terms every learner needs.
- Understand: the one big idea, written as a full sentence. Example: "A chart can mislead even when every number in it is true, because scale and selection change what the reader sees."
- Do: the performance every learner will show, with a verb you can watch. Example: "Judge whether a chart supports its headline and explain why, citing scale and data." That Do line is the core outcome. If you cannot write it in one sentence, the lesson has two outcomes hiding in it. Split the lesson first, tier second.
- Run a short pre-assessment for this topic only. Five to eight low stakes items: two on prior knowledge, three on the core skill at rising difficulty, one that exposes the common misconception, plus one question on interest and one on access needs (language, print size, time, tools). Keep it ungraded and say so. Sort learners by what the check shows for this topic, never by a general label. A learner who is support tier for fractions may be stretch tier for data charts. Readiness is per topic, and it moves.
- Plan three tiers by readiness. Name them support, core, and stretch. Avoid rank names or colours that learners will decode as a league table within a day.
- Support: same outcome, more help. Concrete examples first, worked example open in front of them, task broken into named steps, vocabulary pre-taught, graphic organiser partly filled, sentence starters for the product.
- Core: same outcome, standard route. Normal source material, one worked example at the start, checklist without prompts, independent work with a partner check.
- Stretch: same outcome, less help and harder thinking. Messier or primary sources, an ambiguity or a conflicting source to resolve, a transfer step (apply the idea to a new case), and a defence step (justify a choice against an alternative). Two tiers are acceptable for a small group or a short session. One tier is not differentiation.
- Vary the right dials, and only those dials. Tier by changing one to three of these:
- Complexity of the source: length, vocabulary, abstractness, number of variables.
- Support: worked examples, prompts, models, glossaries, step lists. Most in support, least in stretch.
- Open-endedness: one right route in support, several defensible routes in stretch.
- Structure: steps given in support, steps planned by the learner in stretch.
- Transfer distance: familiar case in support, new case in stretch. Never tier by amount. Ten extra problems is not a stretch tier, it is a punishment for finishing. Keep estimated time within about 40 percent across tiers. Harder thinking, not more items.
- Differentiate content, process, and product on purpose. Decide which of the three carries the tiering in this lesson, and say so in the plan:
- Content: what learners take in. Same concept, different source difficulty or different entry examples.
- Process: how learners make sense of it. Guided steps, paired discussion, stations, independent inquiry, choice of two process routes that both reach the outcome.
- Product: how learners show it. Written verdict, spoken explanation, annotated diagram, recorded walkthrough. Products may differ in form, but they must all be judged against the same shared success criteria. If a product cannot show the core outcome, it is decoration, not a tier.
- Build in the UDL floor first, then tier what is left. Before tiering, remove the barriers you can predict: offer the source in text and in one other form, give instructions in writing as well as aloud, allow more than one way to respond where the outcome permits, and make choice and relevance visible in the task. That is the engagement, representation, and action and expression floor from the CAST UDL Guidelines. Tiering then handles the readiness spread that remains after those barriers are gone. Do not use tiers to fix a barrier that should never have been there, such as tiny print or instructions given only once, aloud.
- Write shared success criteria and check every tier against them. Three to five criteria, in learner language, identical for all tiers. Example: "I name the claim. I check the scale, the units, and the source. My verdict follows from evidence I quote." Then audit each tier with
bin/tier_check.py. Every tier must let a learner meet every criterion. A support tier that never asks for evidence, or a stretch tier whose extra task replaces the verdict instead of extending it, fails the audit even if it looks good.
- Run the lesson with a shared middle. Start together on the core idea (five to ten minutes, one strong example everyone discusses). Split into tiers for the main work. End together: one learner from each tier explains a judgement or a solution, and the group names what all three routes had in common. The shared start and shared end are what stop tiers feeling like separate classes in one room. Keep groups flexible: regroup next lesson from the new check, and let a learner move up mid-lesson if the work shows they are ready. A tier is a route for today, not an identity.
- Close with a common check, then regroup. End with two to four items on the core outcome, the same for everyone, at core level. Score by criterion, not just by total:
- Misses the core idea: same learner stays in support next time, with a different support, not the same worksheet again.
- Meets it with help: core next time.
- Meets it cleanly and explains the why: stretch next time, or stretch within the next task. Record the tier, the check result, and the next tier in one line per learner or per group. For self-study, the learner records it for themselves and picks the next tier by the same rule. Never promote or hold a learner on effort, behaviour, or speed. The check decides.
Validation
Write the lesson as JSON (run bin/tier_check.py init for a full valid template), then check it before teaching:
sh
bin/tier_check.py init > lesson.json
bin/tier_check.py check lesson.json
bin/tier_check.py summary lesson.json
The checker fails a plan when a tier outcome differs from the core outcome, when a tier cannot meet a shared criterion, when workload spread passes a ratio of 1.4, or when the support tier carries no more scaffolds than the stretch tier. Fix the plan, not the checker. A warning about time spread or tier order is a prompt to look, not a block.
Output Contract
- A core block: Know, Understand, Do, with the Do line used word for word as the outcome in every tier.
- A pre-assessment of 5 to 8 items, with the sorting rule stated in one sentence.
- A tier plan for support, core, and stretch, each with task, product, source description, scaffolds, estimated minutes, and why the task matters.
- Shared success criteria (3 to 5 items) and a checker result of PASS from
bin/tier_check.py.
- A shared start and shared end for the session, plus a common closing check and the regrouping rule that follows from it.
- A one line record format for tier, result, and next tier per learner or group.
Operating Rules
- Same outcome, every tier, word for word. The moment a tier gets a different outcome, you have written two lessons, and the weaker one is usually a quiet lowering of expectations.
- Tier by readiness for this topic, from a fresh check. Never tier by a fixed ability label, a reading age from last year, or where a learner sat last term.
- Support means help, not less learning. Worked examples, prompts, and partly filled organisers are temporary. Fade one scaffold at a time as checks improve, and say out loud that the help will fade because the learner is growing.
- Stretch means harder thinking, not more work. Add ambiguity, transfer, conflicting sources, or a defence step. Keep the item count and the time close to the other tiers.
- Every task must be respectful: engaging, worth doing, and visibly real work in every tier. If you would be embarrassed to have a learner compare tiers side by side, rebuild the weaker tier.
- Products can vary in form, criteria cannot. One rubric core for everyone, with tier notes only about support given, never about a lower standard.
- Keep tiers private and fluid. Learners choose or are guided quietly, groups change by topic, and nobody is announced as the low group. In adult and workplace settings, let learners self-select with your guidance and a right to move.
- Choice belongs inside tiers, not instead of them. Offer a choice of case, topic, or product form within each tier so interest drives engagement without breaking the readiness match.
- Do not match teaching to fixed learning style labels. Use learning profile only for real access needs, language support, working conditions, and choice. Offer multiple forms to everyone rather than assigning one form to one kind of learner.
- Know the limits. This skill adapts a lesson you already have. It does not replace a specialist plan for a learner with a formal support plan, a language programme, or an exam board access arrangement. Follow those plans first, then tier inside them.
Versions
v1.0.02026-10-14 · 0 installs · Ed25519-signed./install.sh differentiated-instruction 1.0.0bundle (.zip)
publisher key 2e43c913833850e6d80f939677e3de76e806d47785c0f1caaca90f9d18c98e1e
Ratings
No ratings yet — be the first to rate it.