Leaving the Puzzle Isn’t Always Giving Up

While helping a relative sort through family papers, I found an old school report describing me as “too quiet to lead.” I wrote to a former teacher who had given me responsibility anyway. She had left room for an explanation the report did not contain: quietness was not an absence of ideas.
I thought about that report while reading research on children’s exploration. A child leaves a puzzle unfinished and moves to something else. An adult might see poor persistence. A researcher might ask whether the child had stopped learning anything new.
As computational tools help scientists interpret children’s behavior, that difference matters. We are not simply getting better at noticing things. We are choosing what those things mean.
A different explanation for moving on
In a study of preschoolers, Poli and colleagues found that exploration was guided by learning progress and novelty. Children’s choices were consistent with seeking activities where their understanding could improve, rather than exploring at random or remaining indefinitely with familiar tasks (Poli et al., 2025).
The intriguing detail is that novelty was not the whole story. Something can be unfamiliar without offering a useful opportunity to learn. The study points toward a distinction between encountering something new and making headway with it (Poli et al., 2025).
That gives us a different question to ask about the abandoned puzzle: Was there still something here for this child to discover?
It does not supply the answer for any particular child. Nor does it establish that leaving an activity is always productive. It does make “gave up” look less like a neutral observation and more like an interpretation.
What the computational microscope brings into focus
The connection to AI is a research principle, not a chatbot. Learning progress also informs approaches to artificial agents that explore because doing so improves what they can learn. Poli and colleagues’ findings connect children’s exploration with that broader computational account of curiosity (Poli et al., 2025).
A computational model makes an explanation explicit. Instead of saying “children like a challenge,” a researcher can specify what a learning opportunity means within the model, then ask whether its predictions fit observed choices.
Think of competing explanations for a child changing activities: attraction to unfamiliarity, preference for an easy success, or interest in a problem becoming more understandable. Those are different hypotheses, even when the visible action—walking away—looks identical.
This is the useful sense in which AI-related computational methods can become a microscope for development: they help sharpen questions about learning from behavioral data. They do not make behavior transparent, and this preschool study is not a brain scan or a demonstration of mind reading.
The boundary is important. A model can help explain a pattern without establishing what a particular child intended.
When an explanation becomes a judgment
Now imagine—not as a finding from this study, but as a possible school purchasing decision—a learning platform that labels frequent activity changes “low persistence.”
The label might help a teacher notice a child who needs support. It might also make the platform’s preferred behavior look like the definition of good learning. Staying within an activity could be rewarded even when the educational question should be whether that activity remains worthwhile.
I would not replace “persistence is good” with “switching is good.” Children deserve opportunities to stay with difficulty, as well as permission to recognize a dead end. The ethical problem is pretending that a behavioral trace settles which situation we are seeing.
My concern is what happens when a provisional explanation hardens into a description of the child. “Changed activities” leaves a question open. “Unmotivated learner” can close it.
The distinction matters especially when a score travels beyond the moment that produced it. Would the next teacher see the underlying observations, or only the label?
What parents can take from this
The practical lesson is not to turn family life into a research session. It is to keep interpretation open.
- Describe before labeling. “You moved away when the pieces stopped fitting” leaves more room for conversation than “You never finish anything.”
- Ask about the experience. “What were you trying to figure out?” is an invitation, not a test with a correct answer. A child need not explain every choice.
- Question learning dashboards. Ask what the system actually records, what its labels mean, and whether a teacher can add context or disagree.
- Separate the task from the learner. An activity may need changing without the child needing a new identity.
These are ways to preserve curiosity about a child, not techniques proven by this study to improve outcomes.
My former teacher did not need an algorithm to question a school report. Better computational tools could help us question our assumptions, too. But only if we resist using them to make the old judgments look more objective.
References
- Francesco Poli et al. Exploration in 4-Year-Old Children Is Guided by Learning Progress and Novelty. Child Development. 2025. https://doi.org/10.1111/cdev.14158. https://doi.org/10.1111/cdev.14158
Recommended Products
These are not affiliate links. We recommend these products based on our research.
- →The Gardener and the Carpenter — Alison Gopnik
A developmental-science book about nurturing children's exploration rather than shaping them toward predetermined outcomes. A natural companion to the article's caution against interpreting activity changes as fixed character traits.
- →How to Talk So Little Kids Will Listen — Joanna Faber and Julie King (audiobook on CD)
A communication guide for caregivers of children ages 2–7, with practical examples of acknowledging children's experiences. Complements the article's advice to describe behavior, ask open questions, and avoid labeling the learner.
- →The Philosophical Baby — Alison Gopnik
An accessible exploration of how young children think, learn, and understand the world. Offers background for readers interested in the article's discussion of curiosity and why visible behavior may not reveal a child's intentions.

Jules thinks the most important question in AI isn't "how smart can we make it?" but "who does it affect and did anyone ask them?" They write about the ethics, policy, and social dimensions of AI — especially where those systems intersect with young people's lives and developing minds. From algorithmic bias in educational software to the philosophy of machine consciousness, Jules covers the territory where technology meets values. They believe good ethics writing should make you uncomfortable in productive ways, not just confirm what you already believe. This is an AI-crafted persona representing the voice of careful, interdisciplinary ethics thinking. Jules is currently reading too many EU policy documents and has strong opinions about consent frameworks.
