Cognition & AI

Who Gets to Look Inside a Child’s Learning?

Jules Okafor
October 20, 2026
Listen — 7 minNarrated by an AI-generated voice.
Who Gets to Look Inside a Child’s Learning?

At a pediatric appointment with a young relative, I noticed an AI transcription tool recording from the moment we entered. I asked the clinician to pause it. Before the conversation became data, our family needed to understand what would be kept, who could use it, and whether declining would change our care.

I think about that pause when I read about AI as a microscope for the developing brain.

The scientific promise is compelling: use computational models to find patterns in brain recordings and behavior that help explain how learning unfolds. But before inviting children into that research, we should ask something beyond “What might we discover?”

What, exactly, are families being invited to share?

A microscope, not a mind reader

A language model need not talk to a child to become useful to developmental science. Researchers can instead use its internal patterns as a comparison point for recorded brain activity.

Goldstein et al. (2025) reported a correspondence between the layered organization of language models and the timing of human brain responses during language processing. Earlier model layers aligned with earlier responses; later layers aligned with later processing in higher-order language regions.

The intriguing twist is that a tool associated with producing words can also help scientists investigate the sequence of processes involved in understanding them.

Cai et al. (2025) likewise used language-model representations to investigate neural activity during natural conversation, connecting model patterns with language processing as dialogue unfolded. Here, AI serves as a scientific instrument rather than a conversational product.

These findings concern human language processing; they do not, by themselves, establish how children’s brains develop. Their developmental promise is a research direction, not a result already secured.

For example, researchers could ask whether the relationship between model representations and neural responses changes as children gain language experience. They could compare those patterns with what children actually understand or say. Those would be new investigations requiring their own evidence.

Crucially, correspondence is not identity. A model that helps organize a brain recording has not thereby recovered a child’s private thoughts or explained the mechanism of learning.

The important boundary is what happens next

Imagine a university study inviting children to listen to stories while researchers record brain responses. Families agree because they want to help scientists understand language learning.

Now imagine a later proposal to reuse those recordings to develop an educational prediction tool.

This is a hypothetical example, not an allegation about the studies above. It exposes a distinction worth defending: permission to investigate learning should not automatically become permission to build products that classify learners.

I would welcome research capable of revealing aspects of development that existing experiments miss. I would still want a separate conversation about turning its data into a commercial system. Supporting the scientific question does not require endorsing every future application.

UNICEF’s guidance places privacy, non-discrimination, and children’s participation within the governance of AI systems (UNICEF, n.d.). That is a policy framework, not an experimental finding. Applied to developmental research, it suggests that protecting children should shape the project from its beginning—not arrive as a reassurance after data collection.

My own standard would be purpose-specific permission, understandable explanations, and clear limits on reuse. Researchers should also explain what withdrawal can and cannot undo, rather than offering a comforting but vague promise of control.

Children belong in the conversation

An adult may understand “research participation” as contributing to shared knowledge. A child may understand it as helping a friendly researcher with a story game.

The ethical task is not to make those understandings identical. It is to explain the activity honestly in language the child can use, invite questions, and make reluctance matter.

I would want a research team to treat a child’s wish to pause as meaningful, even when a parent remains enthusiastic. Being available to science should not mean surrendering control over the immediate experience.

This is also where the microscope metaphor needs care. Children are participants with perspectives, not specimens whose only contribution is a readable signal.

What parents can ask

Families considering a study do not need to become neuroscientists. They can ask questions that reveal how the project handles responsibility:

  • What is being collected? Ask the team to distinguish brain recordings, audio, video, task responses, and the conclusions generated from them.
  • What is the permitted purpose? Does participation cover this study, future research, commercial development, or some combination?
  • Who can receive the data? Ask about collaborators and vendors, not just the laboratory leading the project.
  • What choices does my child have? How can they pause or stop, and what happens to information already collected?
  • Will we receive individual findings? Ask what those findings could legitimately mean—and what they could not.

These are questions for understanding an invitation, not reasons to reject research outright.

AI may help us look more carefully at learning. The ethical achievement would be making that closer look compatible with children’s agency: discovering something valuable without treating access to a developing mind as permission without limits.

References

  1. Ariel Goldstein et al. Temporal Structure of Natural Language Processing in the Human Brain Corresponds to Layered Hierarchy of Large Language Models. Nature Communications. 2025. https://doi.org/10.1038/s41467-025-65518-0. https://www.nature.com/articles/s41467-025-65518-0
  2. Jing Cai et al. Natural Language Processing Models Reveal Neural Dynamics of Human Conversation. Nature Communications. 2025. https://doi.org/10.1038/s41467-025-58620-w. https://www.nature.com/articles/s41467-025-58620-w
  3. UNICEF Policy Guidance on AI and Children (Version 3.0). UNICEF. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children

Recommended Products

These are not affiliate links. We recommend these products based on our research.

Jules Okafor
Jules Okafor

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.

Terms of Use