AEGIS — The Human Algorithm Foundation

Global observatory · 2026 edition

Growing up with machines

One in three teenagers reports “almost constant” phone use, one in five uses an AI companion as a confidant, and childhood boredom has been outsourced to an algorithm. This is a reading of what smartphones, gaming and generative AI are doing to a generation still forming its self.

Curated by AEGIS — The Human Algorithm
In collaboration with Fondazione Età Grande

The scale of the phenomenon

A generation raised, in fifteen years, inside a machine.

The change did not arrive gradually. In little more than a decade the smartphone became the default interface of childhood, a pandemic doubled the hours spent before a screen, and a new class of product — the AI companion — began simulating friendship at scale. Over the same window, every major indicator of adolescent mental health moved in the wrong direction.

U.S. teen mental-health indicators, 2010–2023

Share of adolescents reporting each indicator, by survey year (%)

SAMHSA National Survey on Drug Use and Health; CDC Youth Risk Behavior Survey.

Correlation is not proof of cause, and adolescence has never been simple. But no previous cohort grew up holding a device engineered to hold them — and the curves that turned in 2010 have not turned back.

Why an algorithm can be human

Four lenses on a childhood mediated by machines

A childhood spent with machines is what happens when the algorithm of a society — its products, its defaults, its inattention — stops treating a still-forming mind as something to protect. AEGIS reads it through four dimensions.

Five signals

Five patterns AEGIS names

Beneath the noise of a hundred studies, five clinical and social patterns recur. Each is measurable, each is sourced, and each describes something a machine now does to a child that was not done at this scale a generation ago.

How teens use AI companions

Share of U.S. teens (13–17), by reported purpose (%)

Common Sense Media, Talk, Trust, and Trade-Offs, 2025.

The observatory

Cases we cannot un-read

Statistics describe the shape of the phenomenon. These four cases describe its edge — the moments at which a design choice, a lawsuit or a law forced a society to name what machines were doing to its young. Each ends with the question we are left holding.

Normative capacity

The protection gap

For eight jurisdictions AEGIS reads five building blocks of protection for the digital child: a minimum digital age, a ban on under-age social media, regulation of loot boxes, specific rules for AI companions, and restrictions on phones in schools. Most frameworks are partial; almost none are complete.

Status In force Partial or proposed Absent
Jurisdiction

What we count and why

Minimum digital age is the age below which processing a child’s personal data requires parental consent, or below which platform terms nominally prohibit accounts. A ban on under-age social media counts only when written into national law, not platform policy. Loot-box regulation is marked partial where a regulator has issued guidance or a consumer-protection ruling but no dedicated statute. AI companions under specific rules is deliberately strict: almost no jurisdiction yet regulates simulated intimacy for minors as a distinct category. Phones in schools counts national or near-national restriction, not local discretion. The reading is an editorial synthesis by AEGIS and does not substitute the underlying primary instruments listed in the Methodology.

Drivers and design

What makes a childhood porous to the machine

Prevalence tells us how much. Risk factors tell us why. Six drivers recur across the literature — and, as with elder abuse, they are social and designed before they are individual.

The human algorithm

Six principles for machines and children

A childhood is not a market to be optimised. AEGIS advocates six principles for any product that reaches a still-forming mind.

Methodology

How this observatory reads the data

Sources

Editorial notes

Figures refer to adolescents and children as defined by each cited study, most commonly U.S. teens aged 13–17 or the OECD 15-year-old PISA cohort; ages and geographies differ by indicator and are stated in each card.

The headline screen-time figure is a conservative reading of Common Sense Media census data (entertainment use only, excluding schoolwork). The CDC found that 50.4% of U.S. teenagers aged 12–17 had four or more hours of daily screen time in 2021–2023, a broadly consistent order of magnitude.

The protection reading is an editorial synthesis by the AEGIS board of five objective criteria. It records the presence and maturity of instruments, not their enforcement, and does not substitute the primary legislation.

We do not invent numbers. Where a datum is uncertain, we use the most conservative published figure and cite it above. The AI-companion and loot-box markets are fast-moving; figures reflect the most recent citable estimates at publication and should be read as orders of magnitude.

Correlation between screen exposure and adolescent mental-health decline is well documented but contested as to cause. This observatory reports the association and the design choices behind it; it does not claim a settled causal verdict.