10 Civilization 12 min read
The Higher Planes, Made Concrete
The phrase “redesigning civilization” is where thinking about this subject usually goes to die. The discipline that saves it: every claim must name a mechanism.
This is the page where writing about AI usually becomes unfalsifiable. Words like flourishing, consciousness and the next stage of humanity arrive, and the reader is left with a warm feeling and nothing to do on Monday. So there is one rule here: no civilizational claim without a mechanism. If it cannot be stated as something specific that someone could adjust, it does not belong on the page.
Four things are genuinely adjustable. They are boring compared to the usual register, which is the point.
Lever one: what we measure
GDP counts production. It has no opinion about whether the producing was done by people, and no way to represent whether those people retained any agency in the process. A society that doubled its output while its citizens lost the capacity to direct their own lives would register an unambiguous boom.
This is not a new complaint — Kuznets, who built the accounts, warned about it in 1934 — but it becomes urgent when output and human involvement decouple. If the dashboard cannot see the thing that is being lost, the loss is invisible to policy until it is irreversible.
The serious alternative is Amartya Sen and Martha Nussbaum’s capabilities approach: measure not what is produced, nor even what people have, but what people are actually able to do and be. It is already operationalised — the UN Human Development Index descends from it, as does New Zealand’s wellbeing budgeting and Bhutan’s much-mocked and much-more-rigorous-than-you-think GNH framework.
What an AI-era capability dashboard would have to track, that current ones do not:
- Time sovereignty. What share of waking hours are self-directed? Abundance that arrives as more monitored work is not abundance.
- Agency in consequential decisions. Can people meaningfully contest a decision made about them — a loan, a diagnosis, a sentence, a hiring? Contest, not be informed of.
- Skill retention at population level. Are people becoming more or less able to act without assistance? Nobody measures this. It is the most important indicator on the list and there is no series for it.
- Depth of social connection. Not headcount. Number of people who would notice your absence within a week.
Lever two: how we distribute
If capital produces an increasing share of output, income concentrates among capital owners by default. Nothing about this is mysterious and nothing about it is self-correcting.
But the meaning page constrains the solution space in a way that most distribution debates ignore. Marienthal showed that the deprivations of worklessness are only partly financial: time structure, social contact, collective purpose and status are separate goods that money does not automatically buy. A pure cash transfer addresses one of five deprivations.
| Mechanism | Addresses income | Addresses structure & status | Honest weakness |
|---|---|---|---|
| Universal basic income | Yes | No | Trials show it reduces hardship and stress; they are too small and short to tell us anything about a whole society with no work in it. |
| Sovereign wealth fund / citizen dividend | Yes | No | Broadens ownership rather than just transferring income — Alaska and Norway are real precedents. Same structural gap. |
| Job guarantee in care, repair, restoration | Yes | Yes | Enormous administrative burden; risk of make-work, which corrodes the status it was meant to supply. |
| Shorter standard week | Partly | Retains both | Historically the main way productivity gains were taken as leisure. Currently the least discussed and arguably the most proven. |
| Broad capital ownership (co-ops, employee equity, pension stakes) | Yes | Partly | Slow, unglamorous, and it changes who is on which side of the split rather than redistributing after the fact. |
The honest conclusion is that no single mechanism does the whole job, and the debate’s fixation on UBI-versus-nothing is a failure of imagination about which deprivation is being addressed.
Lever three: how we decide
Two pressures arrive together and point in opposite directions.
Decisions are being made faster than deliberation can track — by automated systems, at machine speed, in trading, moderation, targeting, logistics and increasingly in administration. Meanwhile the quality of public deliberation is under direct attack from cheap, personalised, industrial-scale persuasion.
Three mechanisms with actual track records are worth more than any amount of speculation:
- Sortition and citizens’ assemblies. Randomly selected citizens, given time, expert briefing and skilled facilitation, deliberating on one hard question. Ireland used this to break a thirty-year deadlock on constitutional questions that its parliament could not touch. It works because random selection cannot be lobbied and cannot be targeted by a persuasion campaign that does not know who will be picked.
- Mandatory contestability. Not “explainability” in the abstract, but a concrete right: any consequential automated decision about a person must come with a route to human review that has the power to overturn it. The responsibility page’s three-question test is the design specification — information, time, standing.
- Audit before deployment, in public. The model aviation and pharmaceuticals arrived at after enough disasters: independent pre-deployment testing, mandatory incident reporting, published investigations. It is expensive and slow and it is the only regulatory pattern that has ever demonstrably worked on a complex technical system.
Lever four: what we honour
This is the most underrated lever because it looks unserious. It is not. Status is the currency human societies actually run on, and status hierarchies are designed — usually implicitly, occasionally on purpose, and they do move. Duelling was high-status and now looks insane. Smoking was sophisticated. Drunk driving was a joke. Each shifted within a generation, through deliberate campaigns, and the shift changed behaviour far more than the law did.
We currently award status overwhelmingly for economic productivity. That allocation was reasonable when production was the binding constraint. If production stops being the constraint and human agency becomes it, then continuing to award status for output is a design error with severe consequences — it means the highest-status activity available is one people are being progressively excluded from.
The activities worth deliberately elevating are not hard to identify. They are the ones that are already essential, already unautomatable, and already low-status: care, teaching, repair, maintenance, stewardship, and the restraint to not do something you could do. Every one of them is a stewardship role. Every one is undervalued precisely because it produces no artefact.
The Illich test
Ivan Illich, in Tools for Conviviality (1973), proposed the sharpest single question available for evaluating a technology, and it is more useful now than when he wrote it.
A convivial tool expands the range of what people can do for themselves. A manipulative one makes them dependent on a system they do not control and cannot inspect.
After Illich, Tools for Conviviality
The same technology can be either, depending on how it is deployed. A model that teaches you to reason is convivial. The same model, delivered as an oracle whose reasoning you cannot inspect and whose availability you do not control, is not. The difference is not in the weights. It is in the interface, the ownership structure and the incentives of whoever runs it — all of which are choices, and all of which are currently being made by a very small number of organisations with no obligation to consider the question.
Three failure modes, named plainly
- Capture. The productivity gains accrue to a narrow group. Historically the default outcome of a general-purpose technology, and it takes decades of political conflict to broaden. This one is not a risk; it is the baseline you have to actively deviate from.
- Comfortable irrelevance. Material abundance combined with mass agency loss. Nothing hurts, nothing depends on anyone, and the population declines in a way that no economic indicator can see. This is the specifically novel risk — every previous civilisational failure mode involved scarcity.
- Deliberative collapse. Decision speed and persuasion capability outrun collective reasoning, so governance loses the ability to steer at exactly the moment steering matters most.
The historical record supports the objection about comprehensive design and not at all about specific institutional mechanisms. Nobody designed the industrial economy. People absolutely did design the factory inspectorate, the eight-hour day, universal schooling, the pension, the antibiotic approval process and the air-accident investigation bureau — each a targeted response to an identified failure, each contested, each transformative. That is the register this page is in. Not a blueprint for civilization. A short list of adjustable things, argued one at a time.
Where civilization sits on the ladder
- 1
Cede Machine
The machine does it. You never see the work.
Algorithmic governance without contestability. Efficient, and it terminates self-government as a practice rather than as a formality.
- 2
Approve Machine-led
The machine drafts. You review and sign.
- 3
Collaborate Shared
You and the machine work the problem together.
Machine-assisted deliberation — modelling consequences, surfacing trade-offs, translating between publics. Genuinely promising and barely explored.
- 4
Critique Human-led
You do the work first, unaided. Then ask the machine to attack it.
- 5
Reserve Human
You do it alone, on purpose, knowing help exists.
What a society chooses to honour, and what it refuses to do though it could. Definitionally ours.
In your workplace, your professional body, your school board: any consequential automated decision needs a human review route with real power to overturn. Ask who has it. The question alone changes deployments.
Population-level skill retention and agency are not measured by anyone. What is not measured will not be defended. This is unglamorous, cheap, and the highest-leverage item on the page.
You allocate status every day, in what you praise and to whom you defer. Aim some of it at care, maintenance, teaching and repair. Status economies are made of exactly these small acts, which is why they can be changed.
A committee, an assembly, a board, a union. Deliberation is a skill that decays at both individual and civic scale, and the venues where it is practised are closing.
Does this tool expand what I can do for myself, or deepen my dependence on a system I cannot inspect? Ask it about every tool, including the ones recommended on this site.