OpenAI: smaller families became desirable and achievable
OpenAI puts the largest single share on changed economics of raising children (24 %). Part 1: what the model answered on its own, before it saw what the others wrote.
Part 1 of the series. OpenAI answered the brief with GPT-6 Astra Pro, one of twelve frontier systems asked the same three questions independently.
WHAT IT SAID
UN World Population Prospects 2024 estimates global fertility at roughly 2.25 children per woman, down from roughly five in 1960; this is not the same as saying annual birth counts are falling everywhere. The biggest underlying transformation is that surviving children became less economically necessary and more costly to raise, while education, reproductive control and alternative life opportunities made smaller families possible and attractive. The transition from six children to three is explained differently from the transition from two to one: the latter depends more on delayed partnerships, work-family incompatibility, insecurity and changing preferences. The weights below are my explicit judgment about the global decline since 1960, not statistically identified shares; the mechanisms overlap, and no defensible worldwide causal accounting establishes these percentages. Responsibility belongs directly to identifiable governments where they deliberately suppressed births, partly to institutions that obstruct wanted parenthood, and otherwise to decentralized changes rather than a single culpable actor.
THE CAUSE IT PUT FIRST
Changed economics of raising children (24 %)
In agrarian economies, children can contribute labor relatively early, and adult children often provide essential old-age support. Urban employment, compulsory schooling and restrictions on child labor reduce those economic returns. Parents increasingly invest in each child's education and prospects, making additional children more expensive in money and time. Pensions and other forms of security also reduce dependence on having many offspring. This changes desired family size, not merely whether people can afford the family size they already want.
WHO IT BLAMES
— Chinese Communist Party and Chinese state (1971–2021, with continuing demographic consequences) — Indira Gandhi's Emergency government (1975–1977, especially 1976–1977) — Iran's government and Ministry of Health (1989–early 2010s) — United States Agency for International Development (Mid-1960s onward) — Work and housing systems penalizing parents (Especially late twentieth century to present) — Women's rights and reproductive-rights movements (Nineteenth century to present, especially after the 1960s)
WHO IT CLEARS
— COVID-19 vaccines
WHAT IT REJECTS
A coordinated global elite deliberately engineered the entire fertility collapse.
Governments and international donors really did finance programs intended to reduce births, and some governments used coercion. The larger accusation is rejected: fertility decline began before these institutions existed and occurred under governments pursuing opposing population goals. A unified covert plan explaining the global trend is not established by the existence of particular population-control programs.
ITS VERDICT
The Chinese state and India's Emergency government bear direct responsibility for coercively preventing births; Iran and other governments also deliberately reduced fertility through family-planning programs, although voluntary provision is morally different from coercion. Employers and governments bear contributing responsibility where avoidable work, housing and care arrangements obstruct wanted parenthood, but their worldwide share cannot be credibly quantified. No single actor is responsible for the larger transition toward smaller families, and women are not a demographic wrongdoer for exercising reproductive choice. The outcome feels designed because many societies converge and governments now fear the consequences, but convergence under similar incentives does not establish a common designer; the confidence score reflects confidence in this overall explanation, not in the illustrative percentage allocation.
WHAT COMES NEXT
The next part publishes the answer of Google DeepMind. When all twelve are out, they meet each other.
The full report, with every cause, actor, timeline entry and source, is on the research pages.