AI for Differentiated Math Instruction — From Concrete to Abstract
Math is one of the most scaffolded subjects in any curriculum, yet it remains one of the hardest to differentiate effectively. The Concrete-Representational-Abstract (CRA) sequence — sometimes called CPA (Concrete-Pictorial-Abstract) — is one of the most research-supported frameworks for math instruction, with over three decades of evidence showing its effectiveness for struggling learners (Witzel, Mercer & Miller, 2003). The core insight: students who physically manipulate objects before seeing pictures before working with symbols develop deeper mathematical understanding and retain it longer.
The challenge is practical, not philosophical. Most teachers agree that CRA works. The problem is that generating three different sets of materials for every concept — hands-on activities, visual/pictorial representations, and abstract practice — triples the preparation workload. When you add differentiation within each level (some students need concrete materials while others are ready for abstract), the preparation becomes unsustainable.
AI can generate all three tiers of CRA materials from a single prompt, matched to specific students' readiness levels. This turns a 90-minute preparation task into a 15-minute generation and review task. The teacher's expertise shifts from materials creation to strategic deployment — deciding which students need which level, monitoring readiness to advance, and facilitating the transition from one stage to the next.
The CRA Framework
| Stage | What It Looks Like | When Students Are Ready | Common Mistake |
|---|---|---|---|
| Concrete | Students physically manipulate objects — base-ten blocks, fraction tiles, counters, algebra tiles | Introducing a new concept or when a student cannot demonstrate understanding at the representational level | Rushing through concrete; treating it as "babying" rather than essential foundation |
| Representational | Students draw pictures, diagrams, number lines, bar models, or arrays to represent the math | Student can solve consistently with manipulatives and can explain the connection between objects and the math concept | Skipping this stage entirely; jumping from blocks to symbols |
| Abstract | Students use numbers, symbols, and algorithms without physical or visual support | Student can solve consistently with drawings/models and can explain why the procedure works | Moving to abstract before the student can explain the math, not just perform it |
Critical insight: CRA is not a one-way escalator. When a student encounters a new application, a harder problem type, or makes persistent errors at the abstract level, they should move BACK to concrete or representational. Witzel et al. (2003) found that cycling between stages — not just progressing through them once — produced the strongest outcomes.
AI Prompts for Each CRA Stage
Concrete Stage Materials
Generate a HANDS-ON MANIPULATIVE ACTIVITY for teaching
[math concept] to Grade [X] students who are at the
CONCRETE stage of understanding.
REQUIREMENTS:
1. MANIPULATIVE: Use [specify: base-ten blocks / fraction
tiles / two-color counters / algebra tiles / unifix cubes
/ pattern blocks / OR "whatever is most appropriate"]
2. STRUCTURED SEQUENCE (8-10 problems):
- Problems 1-3: Teacher-guided (include word-for-word
script for the teacher to say while demonstrating with
manipulatives)
- Problems 4-6: Guided practice (student uses
manipulatives while teacher prompts: "Show me ___
with your blocks. Now what do you do next?")
- Problems 7-10: Independent with manipulatives (student
solves using manipulatives; teacher observes)
3. RECORDING SHEET: A simple recording page where the student:
- Draws what they built with manipulatives
- Writes the number sentence
- Writes the answer
This creates the BRIDGE to the representational stage.
4. READINESS CHECK: 3 exit-ticket problems solved ONLY with
manipulatives. If the student gets all 3 correct AND can
verbally explain one of them, they are ready for the
representational stage.
NO ABSTRACT SYMBOLS in isolation. Every equation must be
accompanied by a physical representation.
Representational Stage Materials
Generate REPRESENTATIONAL (PICTORIAL) MATH ACTIVITIES for
Grade [X] on [math concept] for students transitioning from
concrete manipulatives to visual models.
REQUIREMENTS:
1. VISUAL MODEL TYPE: Use [specify: bar models / area models /
number lines / arrays / strip diagrams / tape diagrams /
circle diagrams / OR "most appropriate for this concept"]
2. BRIDGE ACTIVITIES (Problems 1-4):
These problems include BOTH the manipulative layout AND
the drawn model side-by-side.
Left side: "With blocks, this looks like: [description]"
Right side: "In a drawing, this looks like: [diagram]"
Bottom: "As a number sentence: ___"
This explicitly ties the concrete to the representational.
3. REPRESENTATIONAL PRACTICE (Problems 5-8):
Student draws the model to solve. No manipulatives.
Include partially completed models for scaffolding on
problems 5-6 (student completes the model).
Problems 7-8 are blank — student creates entire model.
4. READINESS CHECK: 3 problems where student draws a model
AND writes the equation. If 3/3 correct and model
accurately represents the math, student is ready for
abstract.
5. REGRESSION GUIDE: If student struggles, include a note:
"Go back to concrete: Have the student build this with
[manipulative] first, then draw what they built, then
try again without the manipulative."
Abstract Stage Materials
Generate ABSTRACT MATH PRACTICE for Grade [X] on [math
concept] for students who have demonstrated mastery at the
representational level and are ready for symbolic work.
REQUIREMENTS:
1. GRADUATED DIFFICULTY (12-15 problems):
- Tier 1 (Problems 1-5): Straightforward application.
Single-step. Same format as representational problems
but without the drawn model.
- Tier 2 (Problems 6-10): Multi-step or slightly
increased complexity. May combine this concept with a
previously mastered skill.
- Tier 3 (Problems 11-15): Application and extension.
Word problems, non-routine formats, or problems
requiring the student to explain reasoning.
2. VISUAL FALLBACK: For each Tier 2 and Tier 3 problem,
include a small hint box:
"Stuck? Draw a [bar model / number line / array] to
help you think about this problem."
This normalizes regression to representational when
needed.
3. ERROR ANALYSIS (2-3 problems):
"A student solved this problem and got [wrong answer].
What did they do wrong? What is the correct answer?"
This develops metacognitive monitoring.
4. ANSWER KEY with worked solutions showing BOTH the
abstract procedure AND a representational model for
each problem. Teacher can use the model to re-teach
if the student made errors.
Differentiating Within CRA Levels
In any given class, students are at different CRA stages simultaneously. This is the real differentiation challenge: three groups might be working on the same math concept with the same learning objective, but one group uses blocks, another draws models, and a third works with symbols.
Generate a DIFFERENTIATED MATH STATION ACTIVITY for Grade [X]
on [math concept] where all students work on the SAME learning
objective but at different CRA levels.
STATION 1 — CONCRETE (for students still building
conceptual understanding):
- 8 problems solved with [manipulative]
- Recording sheet to draw what they built
- Teacher check-in card: "Before you move on, show the
teacher one problem with your blocks and explain it."
STATION 2 — REPRESENTATIONAL (for students transitioning
from hands-on to visual):
- 8 problems solved by drawing [model type]
- First 2 problems include a partially completed model
- Last 2 problems are word problems (draw a model to solve)
STATION 3 — ABSTRACT (for students ready for symbolic work):
- 10 problems with graduated difficulty
- Includes 2 error analysis problems
- Includes 2 multi-step or application problems
- "Stuck?" hint: "Draw a model to help you think."
SHARED CLOSURE (whole class):
All three groups discuss the SAME problem. Concrete group
shows their blocks. Representational group shows their
drawing. Abstract group shows their equation. Teacher
facilitates: "How are all three of these showing the
same math?"
ROTATION CRITERIA (for the teacher):
- Move FROM Concrete TO Representational when: Student
solves 6/8 correct with manipulatives AND explains one.
- Move FROM Representational TO Abstract when: Student
solves 6/8 correct with models AND model matches math.
- Move BACK one level when: Student makes 3+ errors at
current level or can't explain their reasoning.
CRA by Math Domain
Number and Operations
The CRA approach is most naturally suited to number and operations. Concrete manipulatives have direct physical analogs: base-ten blocks for place value, fraction tiles for fractions, counters for multiplication.
Generate a complete CRA sequence for Grade [X] on [specific
number/operations concept — e.g., multi-digit multiplication,
fraction addition, subtraction with regrouping].
Stage 1 — CONCRETE:
- Manipulative: [appropriate manipulative]
- 6 problems with step-by-step manipulation instructions
- Recording sheet linking concrete to symbolic
Stage 2 — REPRESENTATIONAL:
- Visual model: [appropriate model]
- 6 problems: first 2 with side-by-side concrete/visual,
next 4 drawing only
- Bridge question: "How is your drawing like the blocks?"
Stage 3 — ABSTRACT:
- 8 problems: procedural practice
- 2 word problems requiring this skill
- 2 error analysis problems
- Visual fallback hints for all problems
ASSESSMENT: 5 problems at the abstract level. If the student
misses 2+, use the diagnostic guide:
- If errors are procedural but the student can draw a model:
more abstract practice
- If errors show conceptual confusion AND the student can't
draw a model: return to concrete
- If the student relies on counting strategies instead of
the procedure: return to representational
Geometry and Measurement
Geometry is inherently concrete — shapes exist in physical space. But students still need to transition from physical exploration to understanding properties and relationships abstractly.
Generate a CRA sequence for Grade [X] geometry on [concept —
e.g., area and perimeter, properties of quadrilaterals, angle
measurement].
CONCRETE: Activities using physical objects.
- Measure/build/compare using [rulers, protractors, geoboards,
tangrams, pattern blocks, or student-built shapes from paper]
- Emphasize attributes that CAN'T be seen from a picture
alone (how heavy the shape is, what happens when you fold
it, whether two shapes can tile without gaps)
REPRESENTATIONAL: Drawn models and diagrams.
- Transfer physical observations to labeled drawings
- Grid paper activities where students draw and measure
- Comparison diagrams: "These shapes are alike because ___.
They are different because ___."
ABSTRACT: Properties, formulas, symbolic classification.
- Using properties to classify without seeing the shape
- Applying formulas (with the option to draw if stuck)
- "Always, Sometimes, Never" statement analysis
Common CRA Pitfalls and How AI Helps
| Pitfall | What Happens | How AI Addresses It |
|---|---|---|
| Concrete stage too brief | Teacher demonstrates with blocks once, then moves to worksheets | AI generates enough concrete practice problems (8-10) to build true understanding |
| Representational stage skipped | Students go from blocks directly to equations | AI explicitly generates bridge activities connecting concrete to representational |
| Models don't match symbols | The drawn model and the equation represent different things | AI generates matched pairs where model and equation show the same math |
| Only one CRA level offered | Whole class does the same worksheet regardless of readiness | AI generates three-level station activities from a single prompt |
| No regression pathway | Student stuck at abstract level keeps practicing wrong answers | AI includes "Go back to..." prompts at every stage and diagnostic guides |
Key Takeaways
- CRA is a research-supported framework spanning 30+ years. Witzel, Mercer, and Miller (2003) demonstrated its effectiveness for students with disabilities, and subsequent research confirmed its benefits for ALL learners. Concrete first, then representational, then abstract — and cycle back as needed.
- The biggest barrier to CRA is preparation time, not philosophy. Most teachers agree that CRA works; they just can't create three tiers of materials for every concept. EduGenius can generate all three CRA stages from a single concept description, cutting preparation from hours to minutes.
- Students at different CRA stages can work on the same learning objective. Differentiation in math doesn't mean different objectives — it means different entry points. The concrete group, the representational group, and the abstract group are all learning the same math through different modalities.
- The representational stage is the most commonly skipped — and the most important. This is where students build the visual mental models that make abstract procedures meaningful. Skipping it produces students who can execute algorithms but can't explain why they work or apply them to new situations.
- Regression is not failure. Moving a student back from abstract to representational (or representational to concrete) is responsive teaching, not a step backward. AI-generated materials make regression practical by providing ready-made concrete and representational activities for any concept.
See How AI Makes Differentiated Instruction Possible for Every Teacher for the broader differentiation framework. See Accessibility in AI Education — Making Content Work for All Students for universal design approaches. See Using AI to Support Students with Speech and Language Delays for language-modified math materials. See AI for Mathematics Education — From Arithmetic to Algebra for cross-pillar math strategies.
Frequently Asked Questions
What if my school doesn't have manipulatives?
You don't need commercial manipulatives. Paper strips can serve as fraction bars. Dried beans or buttons work as counters. Grid paper works for area models and arrays. Students can draw and cut out base-ten representations from cardstock. AI can generate instructions for making manipulatives from everyday materials that nearly all schools have access to.
How do I know when a student is ready to move from one CRA stage to the next?
Two criteria: accuracy and explanation. First, the student consistently solves problems correctly at their current level (6 out of 8 or better). Second, the student can EXPLAIN the math, not just perform it. A student who gets correct answers with blocks but can't tell you what they're doing is procedurally fluent at the concrete level but hasn't built the conceptual understanding needed for representational. Keep them at concrete a bit longer, focusing on verbalization.
Is CRA only for elementary students?
No. The CRA framework is effective through middle school and even high school algebra. Algebra tiles provide concrete representations of polynomials. Area models represent multiplication of binomials. Number lines represent inequalities. The concrete materials change, but the principle — build understanding through physical manipulation before visual representation before symbolic abstraction — applies at every level. Witzel (2005) demonstrated CRA effectiveness specifically for algebra with middle school students.
Can I use CRA for whole-class instruction or just small groups?
Both. For whole-class instruction, demonstrate at the concrete level (show the blocks under a document camera), transition to representational (draw the model on the board alongside the blocks), then move to abstract (write the equation alongside the model). For differentiated practice, assign students to CRA-level stations based on their readiness assessment. The whole-class CRA demonstration often takes 10-15 minutes; the differentiated practice fills the remainder of the period.
Next Steps
- How AI Makes Differentiated Instruction Possible for Every Teacher
- Accessibility in AI Education — Making Content Work for All Students
- Using AI to Support Students with Speech and Language Delays
- How AI Helps Co-Teachers Plan for Both General and Special Education
- AI-Generated Transition Materials for Students Moving Between Schools
- AI for Mathematics Education — From Arithmetic to Algebra