# Rethinking Pediatric Human-AI Interaction for Building Safer Digital Health Ecosystems

## TL;DR

This paper discusses how AI systems designed for adults can mislead younger users in healthcare settings and suggests ways to improve safety and understanding.

## Contribution

The paper introduces developmental design features to reduce misinterpretation of AI outputs by adolescents.

## Key findings

- Adolescents may misinterpret probabilistic AI outputs as authoritative clinical advice.
- Technical opacity and model drift can lead minors to perceive unstable system behavior as stable clinical logic.
- Developmental design features like uncertainty cues and layered explanations can improve AI safety for minors.

## Abstract

Artificial intelligence (AI) systems used in everyday digital spaces often rely on design assumptions shaped by adult patterns of reasoning, which creates specific interpretive gaps for younger users. This editorial examines how narrative-style outputs produced through epistemic automation can make probabilistic estimates appear more authoritative than intended for some adolescents. It also considers how technical opacity and model drift introduce shifts in system behavior that minors may misread as stable clinical logic, since there are few cues that distinguish computational changes from expert reasoning. When adolescents independently consult conversational agents or symptom-oriented tools, these interactions can influence clinical encounters without being systematically discussed. Therefore, this editorial outlines practical ways for clinicians to ask about AI-mediated information seeking and describes developmental design features, such as explicit uncertainty cues, layered explanations, and age-responsive prompting, that can reduce misinterpretation. Treating the pediatric digital ecosystem as a distinct design and regulatory setting allows for more precise alignment between algorithmic behavior, developmental cognition, and clinical practice.