Generating Differentiated Math Reasoning Problems With AI

AI generates differentiated math reasoning problems effectively when prompts specify the reasoning demand — not just the difficulty level. Mathematical reasoning tasks must be distinguished by the type of thinking required: recognising a pattern, applying a known strategy in a new context, evaluating whether an approach is efficient, or generating one's own mathematical argument. These distinctions produce genuinely differentiated problems; adjusting only the number size produces harder calculations wearing a reasoning costume.

Quick Answer: Build a three-tier reasoning framework for AI prompt differentiation. Tier 1: apply a given strategy or rule to a provided problem (procedural reasoning). Tier 2: identify the appropriate strategy without being told (strategic reasoning). Tier 3: generate, justify, or evaluate — construct an argument, determine if a claim is always true, or evaluate two competing approaches (generative reasoning). Each tier requires a separate prompt; number difficulty is secondary to reasoning type.

---

Why Differentiated Reasoning Is Harder Than Differentiated Calculation

The moment a teacher hears "differentiated math," the most common instinct is to assign easier or harder numbers. Give the lower group 2-digit problems and the extension group 4-digit problems. This is differentiated calculation — it modifies the size of the computation, not the type of thinking required.

Mathematical reasoning is different. NCTM (2025) defines mathematical reasoning as the ability to analyse situations, draw logical inferences, justify conclusions, and recognise generalisable patterns — skills that are not inherently connected to the complexity of the numbers involved. A student who can recognise that 5 × 12 = 6 × 10 without needing to calculate both is demonstrating multiplicative reasoning that cannot be captured by a "harder multiplication worksheet."

This distinction matters practically for several reasons. What Works Clearinghouse (2024) notes that students who receive only computation differentiation often plateau in mathematics achievement at Grades 5-7 — they become fast and accurate at calculation without developing the flexible thinking that underlies algebraic reasoning and applied problem-solving. The reasoning gap does not show up in calculation tests, which is partly why it persists.

AI is particularly useful for generating reasoning differentiation because the hardest part — designing tasks that differ in cognitive demand rather than computational complexity — is precisely where AI prompt engineering adds the most value. Once a teacher knows how to specify reasoning type, AI generates high-quality differentiated reasoning sets in minutes.

---

A Three-Tier Reasoning Framework

The three-tier framework distinguishes reasoning levels by cognitive demand, not number size:

The jump from Tier 1 to Tier 2 is strategy selection. The jump from Tier 2 to Tier 3 is meta-mathematical — generating an argument, not just executing one.

---

Building Three-Tier Reasoning Problems for Common Math Topics

Number and Operations (Grades 3-6)

Tier 1 (Procedural):

"Write 6 Tier 1 number reasoning problems for Grade 4 students. Each problem states a multiplication fact and asks students to use it to derive a related fact without calculating. Example: 'If 6 × 7 = 42, use this fact to find 6 × 8 without a new calculation.' Use multiplication facts within the 12-times table. Provide the reasoning chain in the answer key."

Tier 2 (Strategic):

"Write 6 Tier 2 number reasoning problems for Grade 4 students. Each problem presents a multiplication or division task and asks students to choose between two strategies (e.g., partitioning vs. rounding and adjusting) and show which they chose and why. Do not tell students which strategy to use. Present both strategy names in the problem so they know what their options are. Provide a model answer showing both strategies and explaining which is more efficient for this number combination."

Tier 3 (Generative):

"Write 4 Tier 3 number reasoning problems for Grade 4-5 students. Each problem presents a mathematical claim about multiplication and asks students to determine: always true, sometimes true, or never true — and to justify their answer with at least two examples. Claims: (1) 'Multiplying a number by an even number always gives an even result'; (2) 'Adding two odd numbers gives an odd result'; (3) 'The product of two single-digit numbers is always a single-digit number'; (4) 'Doubling a number and then halving it gives the original number.' Provide the answer and the justification for each."

Why "always/sometimes/never" tasks are powerful: These tasks require students to test claims rather than apply them — a fundamentally different cognitive process that reveals whether students understand the mathematical structure behind a rule or just the rule itself. EdWeek Research Center (2025) identifies conditional reasoning tasks as one of the highest-impact, lowest-cost strategies for developing algebraic thinking at the upper primary level.

Fractions (Grades 4-7)

Tier 1:

"Write 5 Tier 1 fraction reasoning problems for Grade 5 students. Each problem provides a fraction benchmark (½ is given as the benchmark) and asks students to identify whether a given fraction is greater than, equal to, or less than ½ — using the benchmark only, without converting to decimals or finding common denominators. Fractions to compare: ⅓, ⅝, 4/7, 3/8, 6/11. Provide the reasoning approach in the answer key ('⅝ > ½ because 5 is more than half of 8')."

Tier 2:

"Write 4 Tier 2 fraction reasoning problems for Grade 6 students. Each problem presents two fractions and asks students to compare them using the most efficient strategy — benchmarking, common numerator, common denominator, or missing-piece reasoning. Students must name the strategy and explain why they chose it. Fraction pairs: ⅗ vs. ⅔; 7/8 vs. 5/6; ¼ vs. 3/13; 11/12 vs. 8/9. Provide the most efficient strategy for each pair in the answer key, with the explanation."

Tier 3:

"Write 3 Tier 3 fraction reasoning problems for Grade 6 students. Each asks students to evaluate a claim about fractions and justify their answer: (1) 'A fraction gets larger when you add the same number to the numerator and denominator' — is this always, sometimes, or never true? (2) 'The fraction with the smaller denominator is always the larger fraction' — is this always, sometimes, or never true? (3) 'Multiplying two proper fractions always gives a smaller result than either original fraction' — is this always, sometimes, or never true? Provide the justification, including an example and a counterexample where appropriate."

Algebra and Pattern Reasoning (Grades 5-8)

Tier 1:

"Write 6 Tier 1 pattern reasoning problems for Grade 6 students. Each problem presents a numerical pattern and asks students to apply the rule to find the next two terms. All rules are stated explicitly in each problem. Use linear (constant difference) and simple quadratic (constant second difference) patterns. Provide the rule and the next two terms in the answer key."

Tier 2:

"Write 5 Tier 2 pattern reasoning problems for Grade 7 students. Each problem presents a number sequence without stating the rule. Students must: (a) identify the rule (describe it in words), (b) extend the sequence by three more terms, and (c) use the rule to find the 10th term. Include sequences with: constant difference, constant ratio, alternating signs, and one sequence where the rule involves adding increasing odd numbers. Provide the rule and the 10th term in the answer key."

Tier 3:

"Write 4 Tier 3 reasoning problems for Grade 8 students on algebraic generalisations. Each problem presents two student claims about sequences or expressions and asks students to evaluate both — which is correct, which is incorrect, and why. Include problems where both students have partially correct reasoning, not just one right and one wrong. Provide the evaluation and the correct generalisation in the answer key."

---

A Classroom Scenario: Weaving Reasoning Into a Grade 5 Fractions Unit

Say you teach Grade 5, and your class of 30 students is mid-year in a fractions unit. Suppose end-of-unit tests from prior years show a consistent pattern: about a third of the class can apply fraction procedures (finding common denominators, converting, calculating) but cannot explain why a procedure works or identify when a different approach would be more efficient.

You could decide to incorporate one Tier 3 reasoning task per week — specifically, "always/sometimes/never" fraction claims — as a 15-minute Friday activity for the whole class.

Week 1 Friday — First Tier 3 Task:

You generate this prompt:

"Write a single 'always/sometimes/never' task for Grade 5 students about fractions. Claim: 'If you make the denominator bigger, the fraction gets smaller.' Ask students to: (a) test the claim with at least three examples; (b) decide if it is always, sometimes, or never true; (c) write one sentence explaining their decision. Provide the teacher answer notes: when is it true and when is it not?"

The discussion that could follow — 15 minutes, student-led with you facilitating — has the potential to be among the most mathematically rich conversations your class has all term. Several students might immediately say "always true" (it is sometimes true — only when the numerator stays the same), a few could correctly identify the condition, and the resulting conversation about what has to stay the same for the claim to hold builds a conceptual understanding of fraction size that cannot be generated from fraction calculation practice alone.

Over several weeks, a routine like this aims to help students independently evaluate mathematical claims with consistent justification. The goal is improvement not just in reasoning tasks but in fraction calculation accuracy — because students who understand why a procedure works tend to make fewer systematic errors.

ASCD (2025) describes this "meta-mathematical" instruction — teaching students to evaluate claims rather than just apply rules — as one of the most underused high-impact strategies in primary and lower secondary mathematics. AI makes generating the raw material (the claims and the teacher discussion notes) fast enough to be a weekly routine rather than an occasional project.

---