MindOS Learning Manual: Problem-Decomposition State | A Hard Problem May Be Several Smaller Problems Hiding Together

Wait, What?

A difficult problem can become easier without becoming simpler.

Sometimes the learner does not need an easier question. They need to see that the current question contains several different jobs.

A complex Mathematics problem may require interpreting the situation, defining variables, forming equations, solving them and checking whether the answer makes sense. A Science investigation question may require identifying variables, predicting a relationship, interpreting data and then explaining the mechanism. A comprehension question may require locating evidence, inferring meaning and constructing a justified response.

Those jobs are connected. Treating them as one undifferentiated block can overwhelm a learner who actually possesses many of the component skills.

Quick Answer

Problem-Decomposition State is the learner operation of breaking a genuinely complex problem into meaningful subproblems, preserving the dependencies among them, solving or resolving the parts, and recombining them into a valid whole.

Decomposition is not “do half the worksheet now and half later.” It is not merely shortening the task. It changes the learner’s representation of the problem.

Owned Learning Operation

PROBLEM-DECOMPOSITION STATE = identify whole goal → locate functional subproblems → order dependencies → solve parts → recombine → verify the whole.

This is deliberately separated from Chunking State. Chunking is mostly a bottom-up operation: several understood elements become one larger usable unit. Decomposition is the complementary top-down operation: one difficult whole becomes several functional parts. A mature learner often needs both.

The Wrong Way to Break a Problem Apart

Not every smaller piece is a useful subproblem.

Suppose a learner is solving a multi-step algebra word problem. Splitting it into “read the first sentence”, “read the second sentence”, “write something”, “calculate something” reduces size but does not reveal structure.

A stronger decomposition might be:

What quantities are unknown?
What relationships are given?
Which relationship should be represented first?
What equation or system represents those relationships?
How can the equations be solved?
Does the solution satisfy the original conditions?

The subproblems are functional. Each exists because solving it changes what can be done next.

Four Decomposition Failure States

Fragmentation: the task is chopped into pieces that no longer preserve the logic of the whole.
Missing dependency: a later subproblem depends on information that was never established earlier.
Subgoal completion without recombination: every part is solved, but the learner never checks whether the parts answer the original question.
Permanent external decomposition: the tutor always supplies the subgoals, so the learner never learns to derive them.

That last failure is especially important. A scaffold that makes today’s problem solvable can still block tomorrow’s independence if it never fades.

The MindOS Problem-Decomposition Protocol

Step 1 — State the Whole Goal

Before breaking the task apart, make sure the learner knows what the complete problem is asking. Decomposition without a whole-goal representation can produce locally correct work that never recombines.

Step 2 — Ask What Must Be True Before the Final Answer Is Possible

This reveals candidate subgoals. If the final explanation requires a mechanism, the learner may first need evidence and a causal relationship. If the final mathematical answer requires an unknown quantity, a representation must be built first.

Step 3 — Order the Dependencies

Some subproblems can be solved in parallel. Others must come first. Ask: “Which result does another step depend on?”

Step 4 — Solve One Functional Unit at a Time

Keep the whole goal visible, but temporarily narrow attention to one meaningful subproblem. This can reduce coordination demand without lowering the conceptual standard.

Step 5 — Recombine Explicitly

After resolving the parts, ask how they jointly answer the original problem. Do not assume recombination happens automatically.

Step 6 — Verify the Whole, Not Only the Parts

A correct calculation can sit inside an incorrect model. A correct quotation can sit inside an invalid inference. A correct observation can sit inside a poor scientific explanation. Verification returns to the original goal.

Step 7 — Fade the Subgoal Prompts

Next time, provide fewer labels. Eventually ask the learner to generate the decomposition independently.

Worked Examples Across Subjects

English: for a demanding inference question, decompose into evidence identification → meaning of each clue → combined inference → answer wording → evidence check. The learner should eventually derive this structure without being given the labels.

Mathematics: for an optimisation problem, decompose into define variables → establish constraint → express target quantity → reduce to one variable → optimise → check domain and reasonableness. The subgoals make the architecture visible without solving the problem for the learner.

Science: for experimental evaluation, decompose into claim → evidence → reliability/validity issue → alternative explanation → improved design. Each subproblem serves the final judgement.

Competing Explanations When a Learner Cannot Decompose

The whole goal may be misunderstood.
Prerequisite concepts may be missing.
The learner may not recognise the problem type.
Working-memory load may prevent the learner from holding the whole while generating parts.
The learner may know subgoals but not their order.
The learner may rely on surface templates from familiar examples.
The problem may not actually benefit from decomposition because it is already simple or because its parts are too tightly coupled.

Do not interpret every failure to decompose as weak “problem-solving skills”. Identify which representational operation is missing.

How Do We Know?

Research on subgoal learning provides one important evidence base for decomposition. Studies in procedural domains have found that grouping steps under functional subgoal labels can help novices recognise problem structure and improve near-term problem solving. A 2016 Learning and Instruction study found that subgoal-labelled expository text and worked examples helped learners solve novel programming problems better than unlabeled materials.

A semester-long 2020 study in introductory programming found improved quiz performance and lower withdrawal/failure patterns for students receiving subgoal-oriented instruction, while average exam performance was not significantly better. That difference is exactly the sort of boundary MindOS should preserve: an intervention can improve early problem organisation without guaranteeing a broad long-term advantage on every outcome.

More recent intelligent-tutoring research has also examined training learners to derive subgoals themselves, reporting advantages when learners practised a backward, subgoal-directed strategy rather than merely receiving examples.

Evidence Boundary

Much of the direct subgoal evidence comes from procedural STEM and programming tasks. We should not assume identical effects in every school subject. Nor should we equate teacher-provided subgoal labels with learner-generated decomposition. Provided labels can scaffold the structure; the stronger independence target is for the learner to derive useful subproblems themselves and know when decomposition is unnecessary.

Scaffold Fade

Stage 1: tutor provides all subgoal labels.
Stage 2: tutor provides the first and final subgoal; learner fills the middle.
Stage 3: tutor asks only, “What has to happen before the final answer is possible?”
Stage 4: learner generates and orders subgoals independently.
Stage 5: learner decides whether decomposition is needed at all.

If every difficult problem still requires an adult to draw the roadmap, the learner is using decomposition but does not yet own it.

Immediate, Delayed and Transfer Checks

Immediate: can the learner explain why each subproblem exists?
Delayed: can the learner reconstruct the decomposition after a gap?
Transfer: can the learner derive different subgoals for a new problem with the same deep structure?
Recomposition: can the learner show how all solved parts answer the whole?

Teaching Guide for Parents, Tutors and Teachers

When a learner freezes at a complex task, do not immediately tell them the first calculation. Ask structural questions:

“What is the final thing we need?”
“What would we need to know before that becomes possible?”
“Which part can we solve now?”
“What does this part give us for the next part?”
“How do these pieces answer the original question together?”

These prompts preserve learner cognition better than quietly solving the decomposition for them forever.

MindOS Direction

If the learner knows the pieces but cannot coordinate them, inspect Working Memory Load. If several substeps should eventually become one functional unit, use Chunking State. If the learner cannot recognise the deeper type of problem, use Rule-Induction, Comparison or Analogical-Mapping State. If the learner can decompose only a familiar format, test Transfer State.

MindOS rule: decomposition succeeds when the learner can turn one complex problem into meaningful functional subproblems, preserve how they depend on one another, and recombine them into a valid whole without permanent external prompting.