Problem-Solving Skills in Kids: What the Research Actually Shows

Walk into almost any elementary or middle school and ask to see the school’s problem-solving curriculum. What you’ll usually find is a math problem-solving framework (likely something similar to George Polya’s four-step method) and a social-emotional learning component about “problem-solving” in conflict contexts. What you won’t find is explicit instruction in how to solve novel problems — problems with no predetermined method, unclear information, and multiple possible right answers.

This gap matters because the problems children will actually face as adults — professional, civic, and personal — are almost exclusively this second type. They are not math textbook problems with known solution methods. They are situations where the right approach isn’t obvious, where information is incomplete, and where multiple paths forward exist. The cognitive science on how this kind of problem-solving develops, and what actually builds the skill, offers parents and educators a more useful framework than “critical thinking” and “21st century skills” tend to provide.

What Cognitive Science Says About Problem-Solving Development

The foundational distinction in problem-solving research separates three types of problems: well-defined problems with known solution algorithms, ill-defined problems that require heuristic search, and insight problems that require restructuring the problem entirely before a solution becomes possible.

George Polya’s 1945 work How to Solve It remains one of the most influential frameworks in mathematical problem-solving education. Polya’s four steps — understand the problem, devise a plan, carry out the plan, look back — are explicitly taught in many schools. This framework is valuable for well-defined problems where the path forward is learnable through decomposition. It is less helpful for ill-defined problems where “devising a plan” requires knowing which plan is even plausible, or for insight problems where the solution depends on reconceptualizing the problem itself.

Researcher Jonathan Schooler, in studies published in the journal Psychological Science in 1993 and subsequent replications through the 2010s, demonstrated that insight problem-solving — the “aha” experience — is neurologically distinct from analytical problem-solving. Insight solutions tend to arrive after a period of incubation (being away from the problem), often involve right-hemisphere processing, and are disrupted by verbal articulation of the problem-solving process. This finding has practical implications: telling a child to “talk through” their thinking while they’re working on a creative or insight problem can actually impair their solution-finding. The instruction useful for algorithmic problems is counterproductive for insight problems.

The research on heuristic problem-solving — the messy middle ground between algorithms and insight — focuses on what researchers call “problem-solving schemas.” Schemas are patterns of problem recognition: a child who has solved many problems involving rate-time-distance relationships builds a schema for that problem type, and can apply it to novel problems that have the same underlying structure. Schema development is the mechanism through which experience with varied problems produces transferable skill.

Problem-Solving Type	When Kids Develop This	Activities That Build It	What Undermines It
Algorithmic (following a known procedure to a known answer)	Develops throughout K-12 with instruction; reliable by age 8-9 for simple algorithms	Math practice with immediate feedback, coding with defined outputs, science experiments with known results	Skipping steps, over-reliance on calculators before conceptual understanding is built
Heuristic (strategic search through uncertain problem space)	Emerges meaningfully around age 9-11; strengthens significantly through adolescence	Open-ended engineering challenges, strategy games (chess, puzzles), design projects with multiple constraints, debugging code	Premature hints that bypass the search process; problems that are always solvable by the same method
Creative / Insight (“aha” problems requiring problem restructuring)	Present in some form from early childhood; refined through exposure to insight-type problems	Lateral thinking puzzles, creative constraints (build a bridge using only these 5 materials), incubation time, brainstorming followed by reflection	Pressure for immediate answers; verbal articulation requirements during solving; high-stakes environments

Productive Struggle: The Most Important Finding Parents Need to Know

The concept of productive struggle — coined by researcher Warrington J. Heaton and developed extensively by educational psychologist Jo Boaler at Stanford — is the most practically useful finding from the problem-solving research literature. The basic finding: a period of genuine struggle with a problem, before any help is given, significantly improves both immediate learning and long-term skill transfer. Conversely, providing a hint or worked example before a child has had substantial struggle time reduces what they learn from the problem.

This finding runs directly counter to most parenting instincts. When a child is stuck and frustrated, every parental impulse says to help. The research says to wait longer than feels comfortable.

A landmark 2010 study by Kapur and Bielaczyc published in Cognition and Instruction compared two groups of students learning a mathematics concept. One group received explicit instruction first, then problems. The other group received challenging problems first, struggled with them without instruction, then received explicit instruction. The second group — the “productive failure” group — performed significantly better on transfer problems (novel problems requiring application of the same principle) despite, or because of, the difficulty they experienced. The failure itself was instructional.

Manu Kapur, who has published extensively on productive failure since 2010, describes the mechanism: when students struggle with a problem before receiving instruction, they activate prior knowledge, generate multiple possible approaches (many incorrect), and become cognitively “ready” for the instruction that follows. The struggle creates the conditions for deep encoding. Students who receive instruction first are passive recipients; students who struggle first are active participants in a process of hypothesis testing.

The practical translation for parents is uncomfortable but clear: help later than you want to. When a child is working on a challenging problem and asks for help, it is almost always more valuable to ask a question than to provide a hint. “What have you tried so far?” and “What do you know about this problem?” extend the productive struggle rather than ending it. “Here’s how to approach this” ends the productive struggle and, with it, a significant portion of the learning opportunity.

Why Giving Hints Too Early Undermines Skill Development

The hint-giving research is specific and worth understanding. A 2014 meta-analysis by Wittwer and Renkl, reviewing 42 studies on instructional explanations and hints during problem-solving, found a consistent pattern: hints that explain what to do next (procedural hints) reduce transfer more than hints that ask questions about the problem structure (conceptual prompts). The least helpful hint is the one that immediately resolves the child’s current impasse. The most helpful response to a stuck child is a question that helps them see the problem differently without giving them the next step.

The research identifies three levels of hint specificity, from least to most helpful for long-term skill development:

1. Restating the goal (“Remember, you’re trying to find the fastest route, not just any route”) — minimal guidance, maximal engagement
2. Directing attention (“What do you know about the distances between each point?”) — guides attention without prescribing approach
3. Providing the next step (“Try adding the distances from A to B and from B to C”) — resolves the impasse but removes the cognitive work

The third level — what most parents default to because it stops the child’s frustration most quickly — is the most cognitively costly for the child’s skill development.

This connects directly to the research on executive function and self-regulation in children. Problem-solving is not purely a cognitive skill — it requires the regulatory capacity to tolerate uncertainty, maintain effort in the absence of immediate reward, and redirect attention when an approach fails. These are executive function capacities that develop through practice. Problem-solving challenges that are resolved quickly by adults who provide answers are not practice for those capacities — they are practice for asking adults for answers.

What Types of Activities Build Transferable Problem-Solving

The research on which activities build problem-solving that transfers to new contexts converges on several features that distinguish high-transfer from low-transfer experiences.

Open-ended, multi-solution problems are more effective than single-solution problems for building transferable skill. A challenge that has one right answer and a clear method for reaching it is an algorithm practice problem. A challenge that has multiple workable solutions, each with different tradeoffs, is a problem-solving development problem. Examples: designing a structure to hold a specific weight (multiple solutions), writing a program that could use any of several approaches, planning a garden within a set of constraints.

Real-world grounding improves transfer. A 2016 study by Kaminski and Sloutsky found that abstract problems transferred less reliably than problems with real-world context. When a problem involves something children genuinely care about or can touch and test, the engagement is higher and the encoding is deeper. This is one mechanism behind the effectiveness of maker education and project-based learning — the problems are real, the stakes are real, and the testing is real.

Explicit reflection after problem-solving dramatically improves what is retained and transferred. The “looking back” step in Polya’s framework is the most neglected in practice. Research by Schoenfeld (2014) found that students who spent as little as 5 minutes explicitly reflecting on what they had just done, what worked, and what they’d do differently, showed significantly better performance on structurally similar novel problems compared to students who simply moved on after solving.

The engineering mindset framework provides useful language for this reflection practice: framing failure as information, multiple attempts as normal, and “what did I learn from this attempt?” as the default question after any problem-solving effort.

The Difference Between Mathematical Problem-Solving and Real-World Problem-Solving

One limitation of school-based problem-solving instruction is that it is almost entirely mathematical. Math problems have definite answers, clear feedback, and closed solution spaces. Real-world problems have none of these properties. A child who is an excellent math problem-solver may have difficulty with problems in which:

• The problem itself is not clearly stated and must be defined
• Information is missing, ambiguous, or contradictory
• Multiple reasonable solutions exist with different tradeoffs
• The quality of a solution cannot be determined until after it’s implemented
• Failure doesn’t produce an error message but a situation that must be interpreted

These features describe most significant adult problems. Engineering design problems have all of these features. So do organizational problems, ethical dilemmas, and interpersonal conflicts. Teaching children to work with ambiguous, open-ended problems is not supplementary to academic problem-solving skills — it is the preparation for problems that academic contexts, by design, don’t present.

Computational thinking bridges part of this gap — decomposition, pattern recognition, abstraction, and algorithmic thinking are generalizable skills that appear in both academic and real-world problem contexts. But computational thinking alone doesn’t address the tolerance for ambiguity and multi-path searching that characterizes complex real-world problem-solving.

What to Watch For Over the Next 3 Months

AI tools are reshaping the productive struggle question in classrooms and at home. When a child can ask an AI assistant for the next step at any moment, the conditions for productive struggle are fundamentally altered. This is worth discussing explicitly with children: the value of the struggle is in the struggle, not in being stuck. Using AI as a reference (what are some approaches to this type of problem?) is different from using AI to skip the struggle entirely (solve this problem for me).

Project-based learning certification for teachers is expanding in several states. If your child’s school has recently adopted PBL or design-thinking frameworks, understand what that means in practice — the quality of implementation varies enormously, and the research benefits depend on genuine open-endedness, not just relabeled worksheets.

Summer challenge programs — robotics competitions, design challenges, hackathons for middle and high schoolers — offer concentrated problem-solving development in a real-world context. Registration for fall-season programs often opens in spring.

Frequently Asked Questions

My child gives up as soon as a problem gets hard. How do I build persistence? Start with problems at the right difficulty level — hard enough to require effort, easy enough that success is possible with sustained effort. The zone is roughly where they fail on the first attempt but succeed within 15-20 minutes of trying. Repeated failure without success discourages; repeated easy success doesn’t build persistence. Calibrating the challenge level is the most important variable.

Does playing strategy games like chess or puzzles actually build problem-solving? The research on transfer from strategy games is mixed. Chess training has been studied extensively; findings show improvement in chess-specific problem-solving and some transfer to spatial reasoning, but limited evidence for broad transfer to academic problem-solving. Puzzles and strategy games are valuable for building comfort with uncertainty and multi-step planning, but they are not a substitute for varied, open-ended, real-world problem-solving experiences.

At what age can children tackle genuinely open-ended problems? Children can engage with open-ended problems much younger than most adults expect. A 5-year-old given a set of blocks and the challenge “build something that can hold a cup of water” is doing open-ended problem-solving. The sophistication scales — the complexity of constraints, the number of variables to manage, the abstraction required — but the fundamental capacity is present very early. Waiting until middle school to introduce open-ended challenges is waiting too long.

What’s the role of failure in problem-solving development? Failure is not an obstacle to problem-solving development — it is the mechanism. Kapur’s productive failure research shows that failed attempts, if they involve genuine cognitive engagement rather than random guessing, improve subsequent learning. The goal is not protecting children from failure but ensuring that failure is followed by reflection, not discouragement. “What did I learn from that attempt?” is the question that converts failure into skill.

How do I know if a problem is too hard or at the right level of challenge? Vygotsky’s Zone of Proximal Development — the range of tasks a child cannot yet do independently but can do with scaffolding — is the right concept. A problem is in the productive zone when the child can make progress with sustained effort, even if they can’t reach a solution independently. A problem is too hard when even sustained effort produces no meaningful progress. The sign of “too hard” is random guessing rather than strategic exploration.

Should I teach my child problem-solving steps like Polya’s framework explicitly? Polya’s framework is worth knowing but should be introduced as a description of what effective solvers do, not as a prescription for what to do. A child who mechanically applies “Understand, Plan, Execute, Review” to every problem will apply it ineffectively to insight problems and ill-defined problems. The more valuable lesson is: effective problem-solvers spend time defining the problem before jumping to solutions, generate multiple approaches before committing to one, and always look back at whether their solution actually addresses the original problem.

My child is good at following instructions but struggles with open-ended tasks. Is this a problem? It depends on the age and how extreme it is. Children who are strong algorithmic thinkers often find open-endedness uncomfortable because it removes the feedback signal they’re used to. This is not a deficit — it is a profile that benefits from specific kinds of practice: problems with multiple right answers, design challenges, open-ended making. The discomfort with open-endedness is exactly why explicit practice with it matters.

---

About the Author

About the author Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

• Polya, G. (1945). How to Solve It. Princeton University Press.
• Kapur, M. (2010). Productive failure in mathematical problem solving. Instructional Science, 38(6), 523–550.
• Schooler, J. W., Ohlsson, S., & Brooks, K. (1993). Thoughts beyond words: When language overshadows insight. Journal of Experimental Psychology: General, 122(2), 166–183.
• Wittwer, J., & Renkl, A. (2014). Why instructional explanations often do not work. Educational Psychologist, 43(1), 49–64.
• Schoenfeld, A. H. (2014). Mathematical Problem Solving. Elsevier.
• Kaminski, J. A., & Sloutsky, V. M. (2016). Abstraction, context, and learning. Current Opinion in Behavioral Sciences, 10, 100–106.

---