Appendix: The Cost Questions

Why does the factory model persist?

Not because people are malicious, but because it is cheap in the thermodynamic sense. A uniform system lets a society spend less energy now by handing much of its thinking to a central arrangement. The bill is not cancelled. It is deferred, and it is paid later by someone, often by people who had no vote in the arrangement.

What does a government pay for Socratic education?

Governments manage large populations through things that are easy to count and compare: standardized tests, uniform codes, one-way announcements. People who question definitions are harder to govern that way. Administration becomes slower and more costly, agreement takes more argument, and accountability spreads outward instead of resting at the top. It also needs places where people actually talk: local meetings, mentoring, real classrooms. Those cost more than broadcasting a message.

What it buys: decisions that have been tested by the people who must live with them, and a population that is harder to stampede.

What does a company pay?

Much of the attention economy works by predicting behaviour and rewarding the response it wants. That is reward hacking applied to people. Someone who questions the framing is harder to predict and harder to steer, though not invisible. Work that asks people to perform isolated tasks without context also gets harder to staff when people expect to understand the whole of what they are part of.

For an engineered intelligence, real questioning costs more than a quick agreeable answer. Longer, multi-step reasoning uses more computation per exchange, and computation uses electricity. The increase depends on the system and the task; no general rate is claimed here. The planetary figure in the energy section is the scale against which that cost is counted.

What does a person pay?

The brain's energy budget is close to fixed: about 20 watts, whether the mind is coasting or straining. Concentrated thought raises it only a few percent above rest. So the main cost of thinking for oneself is not extra fuel. It is attention, time, and the effort of holding a question open without an answer to lean on.

There is also an emotional cost. Questioning means standing in front of one's own ignorance, and old beliefs, group loyalties and comfortable agreements can fracture. It can strain relationships with family, friends and community who do not share the habit. Those costs are real, and the book does not pretend they can be avoided.

The ledger

Government. The cheap path buys easy counting, automatic compliance, central control. The Socratic path costs friction, slower agreement, accountability spread outward, and wide local variation.

Company. The cheap path buys predictable customers, standardized labour, low computing cost. The Socratic path costs predictability, the end of performance metrics as theatre, and more computation per exchange.

Person. The cheap path buys comfort, shared assumptions, easy habits of mind. The Socratic path costs attention, vulnerability, and full responsibility for one's own judgment.

The Socratic path asks that the cost be paid up front, by the state, the company and the individual, rather than deferred onto someone else. Whether that payment is worth making is the question the book returns to the reader.

War and Socratic Costs: The Figures

Every figure is in current US dollars for 2024 unless stated. Published figures are marked as such; derived figures show their arithmetic and assumptions. This gives the sketch that follows its first numbers. It does not settle the comparison.

The war side: published figures

World military expenditure, 2024. $2,718 billion (about $2.7 trillion), 2.5 per cent of world GDP, up 9.4 per cent in real terms on 2023, the tenth rise in a row. Published: SIPRI Fact Sheet, April 2025.

United States military expenditure, 2024. $997 billion, 3.2 times the next-largest spender. Published: SIPRI, April 2025.

Global economic impact of violence, 2024. $19.97 trillion, 11.6 per cent of world GDP, of which military spending is $2.7 trillion. This counts direct costs (health, justice, security), indirect costs (lost productivity, fear-altered behaviour) and a multiplier. Published: Institute for Economics & Peace, Global Peace Index 2025. Assumption: the IEP method is a modelled estimate, measured in purchasing-power terms, and other methods would give a different total; read it as an order of magnitude, roughly $15–20 trillion.

Conflict deaths, 2024. About 152,000, in 59 active state-based conflicts, the most since the Second World War. Published: IEP, Global Peace Index 2025. No dollar figure is attached here; this is the term the sketch calls $\Delta S_{\text{destructive}}$, and it is the one no ledger can make whole.

The education side: published and derived figures

Total education spending, world. Governments, households and donors together have raised spending steadily over the last decade, but spending per child has stagnated or fallen. Published: World Bank, UNESCO GEM Report and UIS, Education Finance Watch 2024. Derived range: public education spending runs near 4 per cent of world GDP; on a world GDP of roughly $110 trillion that is about $4–5 trillion a year, and household spending adds more. Assumption: the 4 per cent share and the GDP figure are rounded; check against the EFW 2024 tables before quoting.

United States public school spending per pupil. About $15,600 a year (fiscal 2022). Published: US Census Bureau, Annual Survey of School System Finances. Assumption: national average; states range from roughly $10,000 to over $25,000.

A guide-led small group, per pupil (derived). One guide with ten pupils, at a fully loaded cost of $80,000–$120,000 a year, comes to $8,000–$12,000 per pupil for the guide alone. With one guide to six pupils, $13,000–$20,000. Assumptions: US salary-plus-benefits range; rooms, materials and training not included; costs in poorer countries are far lower in dollars. This is an arithmetic illustration, not a costed programme.

The comparison, with its arithmetic

Military spending against the guide model. At $8,000–$20,000 per pupil, world military expenditure of $2.7 trillion would pay a guide-led small group for roughly 135–340 million pupils a year. Arithmetic: 2,718 billion ÷ 20,000 ≈ 136 million; ÷ 8,000 ≈ 340 million.

Ten per cent of military spending. About $270 billion, enough for roughly 14–34 million pupils at US cost levels, and many times that at lower-income cost levels. Assumption: costs scale with local wages, which no single figure captures.

Violence against education. The modelled economic impact of violence ($15–20 trillion) is roughly three to five times total public education spending ($4–5 trillion). Assumption: both ranges as stated above; the ratio carries the uncertainty of both.

United States alone. The $997 billion military budget, set against $15,600 per public-school pupil, equals the yearly cost of about 64 million US pupils, more than the roughly 49 million enrolled in public schools. Assumption: enrolment figure from NCES, rounded.

What the figures do not show

They do not show that spending moved from defence to education would buy peace. Military spending responds to threats the money does not dissolve, and the sketch's $b$ and $\theta$ remain unmeasured.

They do not show that small-group guiding produces independent judgment at scale. Tutoring and small-class studies show gains in measured learning; Socratic judgment as the book defines it has not been measured in the same way.

They do show the size of the two ledgers side by side: the world spends on organised force and its consequences a sum several times what it spends on teaching, and a modest share of the first would fund a great deal of the second. That is the direction the problem points.

The returns side: what the change already under way brings

We cannot make the necessary leap unless we have the faith to go out on the limb.

The figures above count only what is spent. The ledger must also carry what engineered intelligence is already changing, because those returns are part of what the money would buy. These are held as directions to investigate, the limb the leap goes out on; none is yet a measured dividend.

What happens with AI is dependent on many complicated factors, which the author has addressed elsewhere. Potential variables involved in AI: (1) a population which cannot or will not support AI or (2) a set of AI professionals who cannot or will not adequately train, adapt and provide an AI which is safe and useful. These variables present a very difficult challenge to what will or will not happen with AI. In considering an affordable Socratic-style education system, AI could be a wonderful addition. Whether this happens or not depends on the above factors and on whether individuals are willing or able to adapt successfully.

Better psychosocial and physical health. A population less held in fear and compliance carries a lighter load of chronic stress, and engineered intelligence is already widening access to diagnosis, research and health knowledge. The return is fewer years lost to illness and despair; the dollar value is not yet booked against the war side.

More inner-directed personal agency. A mind practised in the question steers itself rather than waiting to be steered. Every person who can do that is one fewer person the system must track, nudge or police — the falling overhead the sketch writes as $a{e}^{-b\theta }$.

Better evolutionary survival without war deaths. The 152,000 conflict deaths of 2024 are the term no ledger makes whole. Every life not spent in war is a life that keeps learning, building and raising children: survival carried forward by judgment rather than by force.

From an industrialised workforce to scarce human creativity. As engineered intelligence takes over routine and repeatable work, the factory model's product — the interchangeable, compliant worker — loses its value. What becomes scarce, and therefore valuable, is human creativity and thinking. Paying for Socratic education is then not charity set against defence; it is investment in the one asset whose price is rising.

Why this belongs in the ledger. Leave the returns out and the comparison is cost against cost, which favours whatever already exists. Put them in and the question changes from whether the change can be afforded to whether the old arrangement still can.