The Widening of "We"

Definition

The Widening of "We" is the deliberate expansion of whose welfare an AI-Born enterprise counts — the workers navigating its transition, the communities that generated its data, the regulatory environment that licenses it, the collective knowledge its models were trained on. It operates through four moves: expanding the stakeholder scope an optimization function weights, opening governance so more voices inform strategic intent, redistributing ownership of the value created, and recognizing contribution that markets systematically underprice.

The crucial claim is that this is not an aspiration appended to the AI-Born model. It is a conclusion derived from the architecture itself. When a Machine Core executes millions of decisions daily, whatever implicit answer the Human Cortex embedded to the question whose welfare counts? executes at the same scale. Technology is never neutral. It is frozen intention executing at scale. The widening of "we" is simply the disciplined acknowledgment of that fact — and the practice of answering the question on purpose rather than by default.

The problem it solves

Hold two architectures side by side. The Lombok Strait, the eleven-mile passage between Bali and Lombok, works because dozens of nations, shipping operators, and coastal communities spent decades building shared navigation protocols and shared commitments to its integrity. No single entity profits maximally from it. All of them depend on it. Now set the East India Company beside it — its own army, its own courts, its own flag, governing 200 million people on a single logic: maximize returns to shareholders in London. The model delivered genuine value for roughly two centuries, then generated costs so catastrophic — the Bengal famine, the institutional collapse of a subcontinent — that the Crown had to nationalize it in 1858.

Here is what's worth sitting with: the East India Company didn't set out to destroy itself. It optimized. Every quarterly decision rationally served the function it was given. The disaster emerged not from bad actors but from what the optimization function excluded — the welfare of the people it governed, the legitimacy it depended on, the social fabric that trade itself required. Strip those from the reward function and you don't get a worse company. You get a faster route to the same catastrophe.

The problem the Widening of "We" solves is one of speed. The Company's extraction worked at the pace of sailing ships and handwritten ledgers; mistakes could be corrected, locals could resist, parliaments could intervene. A Machine Core has no comparable friction. It compresses that 200-year arc into a decade. Whatever you encode — whose welfare, over what time horizon, toward what ends — compounds into infrastructure before correction becomes possible. Geoffrey Hinton named the same danger from Stockholm: AI's productivity gains will be "a wonderful advance for all humanity" if shared, "but if they are created by companies motivated by short-term profits, our safety will not be the top priority."

Anatomy

Figure: The widening of "we" expands the moral circle—from kin and tribe to the workers, communities, regulators, and collective knowledge an AI-Born system's reward function should count.

The framework expands the moral circle Peter Singer documented across human history — kin, tribe, nation, species — and gives it an organizational mechanism. Its components:

• Whose welfare the reward function weights. The most consequential design decision. A customer-service VP-Agent optimized purely for case-closure speed externalizes relationship costs; one designed to balance resolution quality, sentiment, retention, and speed builds customer equity that appreciates. Identical architecture, divergent outcome — the difference is what was encoded, and with whose welfare in mind.
• Whose voices shape intent. Governance opened to those who bear the consequences, the procedural expression of which is Participatory Technological Assessment.
• Who participates in the value created. Ownership and surplus distribution — Participation Dividends and steward-ownership structures that redistribute without eliminating the moat.
• Whose contribution is acknowledged. A second dimension beyond immediate stakeholders: a frontier model stands on centuries of literature, millions of open-source repositories, university labs funded by public science councils, even power grids built with taxpayer money. The productivity is a crystallization of collective knowledge across generations — which does not erase the real claims of those who funded the compute and bore the risk, but does make the question of acknowledgment hard to ignore.

How it works in practice

The mechanism is visible in the Wells Fargo asset cap. After the fake-accounts scandal, the Federal Reserve barred the bank from growing its balance sheet for seven years — a constraint finally lifted in June 2025 that cost the company more in foregone opportunity than every stewardship investment it might have made. That case belongs to the pre-AI era, when legitimacy eroded slowly relative to what autonomous systems can produce in months. Extraction at machine speed doesn't create sustained advantage. It generates Alignment Debt and legitimacy erosion that compound until the balance sheet can no longer hide them.

The positive case is Elinor Ostrom's Nobel-winning research: communities managing shared resources — Swiss alpine meadows sustained for 500 years, Japanese forest cooperatives operating since the 16th century — without privatization or state control. Not through altruism, but through institutional design that aligned individual incentives with collective welfare: clear boundaries, participatory rule-making, graduated sanctions. The tragedy of the commons was a design failure, not an inevitability. The Lombok Strait works on the same principle at a different scale. So can the AI-Born enterprise — when its designers understand that widening "we" is not a constraint on performance but a condition for it. When the circle expands to include the communities that generated the data, the workers in transition, and the regulator granting the license to operate, the enterprise doesn't become less competitive. It becomes harder to destroy.

How to apply it

The framework resolves into an integrated mandate — a sequence, not a checklist, because architecture accumulates and early code is durable:

1. Embed values before you build defensibility. The moat amplifies whatever sits in the values layer beneath it. Build on extractive foundations and you build a faster East India Company; build on stewardship foundations — constitutional mission-lock, stakeholder voice in agent reward functions, governance that survives leadership transitions — and you build something that can last. Redeemable shares, purpose-aligned funds, PBC status with transfer restrictions are available now and close progressively as valuation rises.
2. Begin transition infrastructure from the first automation decision. When agents first absorb a workflow, that's the moment to invest in portable benefits and community formation — not eighteen months later, when headcount has already contracted.
3. Govern the concentration your moats create. Participation Dividends and steward-ownership redistribute the surplus without eliminating the moat. Book 1 builds the capacity; Book 2 governs what you do with it.
4. Run the quarterly Alignment Debt audit. Sit with whoever operates your agents and ask what decisions they made this period that a human would have made differently — where the reward function drifted from intent. Drift is a pattern of small, locally rational, collectively corrosive optimizations. Catching it quarterly is vastly cheaper than catching it when the legitimacy crisis arrives.
5. Ask workers what they'd do with the time savings before automating. Not as courtesy — as a design constraint. The answers change what you build, and sometimes whether you build it.

Failure modes / what it is not

• It is not anti-growth or anti-profit. Market capitalism has delivered genuine material progress, and the East India Company delivered real value — trade routes, naval protection, commercial infrastructure. The question is not whether to pursue returns but what time horizons and stakeholder considerations shape that pursuit.
• It is not a moral veneer on extractive intent. You can build steward-ownership structures and fill them with people whose orientation is still extractive; you can run Participatory Technological Assessment as theater. The outer architecture works only when inhabited by people who have developed the working capacities — the willingness to act for collective outcomes at individual cost, the discipline to design for one's own fallibility, the habit of asking whose interests a system serves before it calcifies.
• It is not charity. It is infrastructure maintenance — protecting the commons (research, education, data, institutions) that AI productivity draws on and depends on.
• It is not something machines can settle. AI can optimize any objective function you give it — profit, engagement, throughput. It cannot tell you whether those objectives are worthy. That requires curiosity, compassion, taste, and judgment — capacities that resist automation not because machines can never simulate them but because democratic accountability requires human agents who can be named, questioned, and held responsible.

Relationship to other frameworks

The Widening of "We" is the culminating synthesis of both books. It is the purpose toward which Values-Conscious Architecture aims — the answer to whose welfare? that gets frozen into every reward function. It is the strategic case made by Stewardship as Competitive Advantage, generalized to civilizational scale: trust advantages compound while extraction advantages erode. Its procedural input is Participatory Technological Assessment. And it is the moral completion of the Machine Core + Human Cortex split: the Machine Core executes; the Human Cortex chooses what is worthy of execution — and widens the "we" that benefits from both.

Origin note

Original. The framework systematizes stakeholder expansion within AI-Born architecture, integrating ownership, governance, and value-distribution mechanisms. It draws transparently on Peter Singer's expanding moral circle and Elinor Ostrom's commons-governance research, synthesizing them into an architectural design discipline rather than a purely ethical exhortation.