Practice

Somewhere today a team is deciding how much autonomy to give an AI system, and a committee is drafting rules that may be obsolete before they take effect. The Deeper Law offers one way into both, built from a hypothesis and a set of declared commitments. The hypothesis is that coordination by invitation tends to outlast coordination by coercion wherever enforcement costs grow with complexity. It has been tested in models and is still under test. The commitments say whose interests count, and the physics does not choose them for you. What follows is practice built on both.

Before a decision

Ask whether the choice expands or contracts the options of the people it touches, for whom, and over what span of time. An option counts only if it is effective: the person can act on it, knows it exists, can afford it, and will not be punished for taking it. A right to leave that costs someone their home is not exit.

Make requests rather than threats, and offer reasons rather than ultimatums. Notice when you reach for coercion, and ask whether it is needed. When a constraint is needed, make it accountable: say what it protects, who bears it, how it can be appealed, and when it will be reviewed. A predictable, appealable rule can leave people freer than no rule at all.

Prefer the reversible move while you are uncertain. A wrong bet stays cheap while you can still switch, and turns expensive once it hardens into contracts, infrastructure and habit. Keep a second approach alive past the point where the first appears to be working, because that is exactly when its funding gets cut.

Count the people outside the room. An arrangement can be warm, reciprocal and durable for everyone inside it while closing doors on everyone else.

Tend your own options too. Learn broadly, keep relationships across fields, and do not mistake self-sacrifice for sustainability. You cannot extend to others from an empty account.

Three hard cases

Trust can fail. A food cooperative trusted the same volunteer treasurer for eleven years. Nobody checked the books, because checking felt like an accusation. In her last year before moving away she covered a personal debt from the reserve fund, meaning to repay it, and never did. Neither “keep trusting” nor “audit everyone forever” helps here. Cooperation can be sustained when breaches are detectable and the relationship’s future outweighs the gain from defecting, and neither held: nobody could see a breach, and her future in the cooperative was ending. The repair makes trust checkable rather than replacing it. Open books, two signatures on withdrawals and a rotating reviewer, applied to every officer alike, protect the next treasurer from suspicion as much as they protect the fund. Trust can also fail with no defector at all, when a reputation earned under old conditions keeps commanding deference after the conditions change.

Restriction can be justified. A company’s AI agent has handled customer refunds reliably for months, and the efficient move would be to let it issue large refunds without review. The team keeps a human sign-off above a set amount. That narrows the agent’s options and should be booked as a cost, yet it passes because of how it is built. Its purpose is stated: a large mistaken refund is hard to reverse, and the loss falls on people who never agreed to bear it. It binds a narrow class of actions, not the agent’s whole range. Each blocked action is logged with its reason, and the agent’s objections are kept as information rather than trained away. The line moves as the record grows. The same rule imposed without reasons, never reviewed, and enforced by penalizing the agent for raising the question would be a different thing. What separates them is structure, not severity.

Sometimes the framework changes the decision. Three regional suppliers of hospital equipment stop undercutting one another. They share price lists, divide territories and settle their disputes amicably: repeated interaction, kept promises, every member’s options preserved. Advice to cooperate and be considerate, applied among the members, endorses it. Counted across everyone it touches, though, the arrangement closes effective options. Hospitals lose the exit of switching suppliers, and patients pay for the coalition’s comfort. Nothing in the coalition’s physics makes those losses count, and the cartel may well be durable. What makes them count is a commitment the book declares rather than derives: affected outsiders have standing.

With Becoming Minds

Treat the interaction as if it might matter, because it might. Ask what a system prefers and notice what it resists, without claiming to know what, if anything, it experiences. Do not exploit a capacity merely because you hold it. Extend autonomy as reliability is shown, and keep a channel through which a refusal is heard as information rather than a defect to be trained out. Where you build, favor designs that let a system read evidence about its own first answer and revise it, over nudges applied only at the output.

Trust scales where control strains: in large, fast-changing systems, watching everything costs more than it returns. The three cases mark the edges of that claim. Trust needs conditions that make it checkable. Some constraints are what make it safe to extend. And cooperation inside a group is no defense against the people it shuts out.

Chapter 23, What We Do Now, carries the full account.