Artificial intelligence models built for adult medicine cannot simply be handed to children, and the peer-reviewed literature now offers named frameworks to prove it. Two complementary guides—the PEARL-AI framework, introduced by Chng and colleagues in npj Digital Medicine in 2025, and the earlier ACCEPT AI framework proposed by Muralidharan et al. in 2023—translate children's unique vulnerabilities into actionable recommendations for developers, clinicians, and regulators. A third contribution, a proactive, risk-sensitive interim review model published in Frontiers in Digital Health by Chih-Shung Wong and Tsui-Wen Hsu, extends this thinking to trial oversight. Together they map the technical and ethical terrain of pediatric AI.
The Unique Landscape of Pediatric AI: Challenges and Ethical Imperatives
The starting point for every one of these frameworks is a simple clinical truth: children are not miniature adults. As a review in Clinical and Experimental Pediatrics puts it, pediatric patients "necessitate specialized, tailored approaches that account for their unique developmental and physiological characteristics." A model trained predominantly on adult data inherits adult physiology, adult disease distributions, and adult assumptions about consent—none of which transfer cleanly to a population that changes developmentally year by year.
The gap is not merely technical but structural. The authors of the npj Digital Medicine review note that before their work, "there does not exist any previous comprehensive review on AI ethics in child health or any guidelines for management, unlike in adult medicine." Adult medicine has an established ethical apparatus for AI; child health has had to build one from scratch, which explains why several frameworks appeared in quick succession between 2023 and 2026.
Age-related algorithmic bias is the practical consequence. When age-specific best practices are absent, models can systematically misserve younger patients—not because the algorithms are broken, but because the data and the ethical guardrails were designed for someone else. That is the problem ACCEPT AI was explicitly proposed to address.
Guiding Principles: Ethical Frameworks for Pediatric AI
ACCEPT AI, proposed by Muralidharan et al. in 2023, is founded on six core principles: age, communication, consent, equity, data protection, and technological transparency. Each section of the framework carries key recommendations that, in the words of the Clinical and Experimental Pediatrics review, "enable researchers, regulators, and clinicians to translate these ethical principles into practical measures to mitigate age-related algorithmic bias." Its focus is the ethical prerequisite stage—what must be true before AI is introduced into pediatric medicine at all.
PEARL-AI, the Pediatrics EthicAl Recommendations List for AI proposed by Chng et al. in 2025, shifts the emphasis toward clinical practice. According to the review comparing the two, PEARL-AI "addresses the unique needs and vulnerabilities of children and focuses on an actionable ethical strategy for clinical practice." Its authors describe it as "a practical, actionable resource for clinicians, academics, healthcare administrators, and AI developers," and it includes new recommendations the authors consider important for ethical AI in pediatric healthcare, elaborated in their preceding section on child-centric medical AI.
The two frameworks are therefore complementary rather than competing: ACCEPT AI establishes the ethical prerequisites for introducing AI into pediatric settings, while PEARL-AI supplies an operational strategy for using it responsibly once introduced.
Comparative Analysis of Pediatric AI Ethical Frameworks
The table below contrasts the stated goals, core components, and intended users of the three models. The comparison shows a deliberate division of labor: ACCEPT AI targets age-related bias at the introduction stage, PEARL-AI guides day-to-day practice across multiple stakeholder groups, and the interim review model equips oversight bodies for governance of evolving trials.
| Framework/Model | Primary Goal | Key Principles/Components | Target Audience |
|---|---|---|---|
| ACCEPT AI (Muralidharan et al., 2023) | Translate ethical principles into practical measures that mitigate age-related algorithmic bias | Six core principles: age, communication, consent, equity, data protection, and technological transparency, each with key recommendations | Researchers, regulators, and clinicians |
| PEARL-AI (Chng et al., 2025) | Provide ethical recommendations and an actionable strategy for child-centered medical AI in clinical practice | Ethical principles plus new recommendations for child-centric medical AI, addressing children's unique needs and vulnerabilities | Clinicians, academics, healthcare administrators, and AI developers |
| Proactive, risk-sensitive interim review model (Wong and Hsu, 2026) | Translate ethical principles into concrete oversight mechanisms for AI-driven pediatric trials | Structured interim oversight grounded in protection, proportionality, respect for autonomy and assent, and procedural fairness | Institutional review boards, regulatory authorities, and healthcare institutions |
For a developer choosing where to start, the distinction matters: ACCEPT AI's principles shape design and validation choices, PEARL-AI shapes deployment and clinical workflow, and the review model shapes institutional governance. A project can, and arguably should, draw on all three at different lifecycle stages.
Proactive Oversight: The Interim Review Model for Pediatric AI Trials
Ethical review of pediatric AI research cannot end at initial approval, because the technology itself evolves during a study. The proactive, risk-sensitive interim review model developed by Wong and Hsu addresses this by building structured oversight into the life of AI-driven pediatric trials. Their recommendations are designed to be adapted to local legal, cultural, and organizational contexts while preserving four core commitments: protection, proportionality, respect for autonomy, and procedural fairness.
In their conclusions, the authors frame the model as "structured interim oversight grounded in core ethical principles—protection, proportionality, respect for autonomy and assent, and procedural fairness—provides a foundation for enhanced governance of evolving research technologies." The addition of assent alongside autonomy is notable: in pediatrics, the child's own agreement matters ethically even where legal consent rests with a guardian.
The model's guidance is explicitly addressed to institutional review boards, regulatory authorities, healthcare institutions, resource-constrained settings, and future policy development. That breadth signals its ambition: not a checklist for one trial, but a template for how oversight bodies everywhere can keep pace with AI that changes under their feet.
Overcoming Practical Barriers to Pediatric AI Implementation
Even a perfectly designed framework runs into hard material constraints. A systematic review of ethical and practical considerations of AI in pediatric medicine, published through PubMed Central, identifies impediments "beyond ethical considerations": implementing AI-driven technologies requires substantial financial investment, advanced technical infrastructure, and specialized training for healthcare professionals.
The review also flags an equity dimension that frameworks alone cannot solve. Disparities in access to AI technologies between high- and low-resource settings, it warns, "may exacerbate existing healthcare inequalities." A model validated in a well-resourced children's hospital may be unusable—or actively harmful—in a setting without the infrastructure to run, monitor, and maintain it. Any deployment plan that ignores this gradient risks widening the very gaps pediatric care aims to close.
These barriers explain why the frameworks are written the way they are. PEARL-AI addresses administrators as well as clinicians; the interim review model explicitly includes resource-constrained settings among its intended users. The literature has absorbed the lesson that ethics divorced from infrastructure is aspiration, not practice.
Navigating the Future of Pediatric AI: A Call to Action
Developers, clinicians, and policymakers should proactively integrate specialized ethical frameworks and oversight models into all stages of pediatric AI development and deployment. The evidence base now exists: ACCEPT AI for bias-aware design, PEARL-AI for child-centered clinical practice, and the interim review model for governance of evolving trials. None of this constitutes legal advice or a regulatory compliance manual, and none of it supplies model-training instructions—but together they form a coherent ethical architecture that adult frameworks cannot replicate.
The measurable indicator of progress is concrete: increased adoption and documented application of frameworks like PEARL-AI and ACCEPT AI in pediatric AI project proposals and institutional review processes. Until those citations start appearing routinely in protocols and ethics submissions, the frameworks remain proposals rather than practice—and the children they were written to protect remain served by systems designed, at root, for someone else.
Sources
- Ethical considerations in AI for child health and recommendations for child-centered medical AI - npj Digital Medicine — Nature
- Frontiers | Ethical oversight of AI-driven paediatric trials: a proactive, risk-sensitive interim review model
- Clinical and Experimental Pediatrics — E CEP
- Ethical and Practical Considerations of Artificial Intelligence in Pediatric Medicine: A Systematic Review - PMC
