Before any of this was written down, I learned something watching Brett Morris and Alistair King work.
Origins
The Morris-King Pattern
Alistair is the person who convinced me to join King James and eventually Accenture. We worked together on meaningful projects, and I think we enjoyed a mutual respect for great ideas and environments with good people. Legend in the South African advertising game. But that's not why I'm starting here.
I joined King James as a freelancer close to a big pitch, and I got to spend hours listening to Alistair talk about work (the creative output) and values (how we should think about working together). The interesting thing is he rarely spoke of them as separate things. A sharp example: he'd say things like "we can sit here and shuffle peas on a plate, but what earns you a place at this table are ideas."
Alistair was not the first. The other leader who cemented this understanding was Brett Morris — who led FCB Johannesburg to South Africa's first ever Cannes Grand Prix, was voted most admired agency leader in Johannesburg five years running, twice for South Africa overall, and transformed FCB Africa into Nahana Communications Group. Brett has his own version of the same principle — the ultimate currency around him: an abundance of well-formed ideas. Bring some and we can talk. Otherwise, jog on mate.
The truth is I've spent the better part of a decade working with both of them, and they have remarkably similar values. More importantly, they impart them the same way: not through lectures alone, but through the atmospheric pressure of explicit standards. The values are created and enforced in that pressure, and the rewards are issued by it too. You didn't need to be told what mattered. You could feel it in the room.
Neither of them was asking people to believe something. They were creating environments where a certain kind of work survived and another kind didn't. People who brought trash self-selected out. People who brought ideas got closer to the centre. And crucially, they didn't just attract great ideas — they attracted good people. There may be a world where you're spoilt for brilliant work but surrounded by bad actors. Watching leaders in that world would only teach you half of what matters. The value wasn't on a wall — it was in the room's gravitational field.
That's what most corporate values miss entirely. They're aspirational nouns — integrity, innovation, excellence — with no mechanism attached. They describe what the company wishes it was, not what it selects for. The test is simple: does the value change behaviour when no one's watching? Does it cost you something to uphold it? If not, it's not a value. It's a press release.
This framework is an attempt to articulate the gravitational field I want KYL to operate in. Not a list of wishes. A coherent picture of expectations that always express in context of the benefits they give you access to. Both intimidating and freeing. Intimidating because the standard is real. Freeing because you know exactly what's being asked of you at all times.
The components aren't new — game theory, network effects, Ostrom's commons research, the neuroscience of cooperation. What's original is the architecture: how these patterns lock together into a self-correcting system. The fractal diagnostic, the epistemic governor, the asymmetry between creation and destruction — these emerge from watching how environments that produce great work actually function, then building a framework that captures why they work. The principles below are what I found, and how they fit together.
Preamble
What This Is
This isn't an academic paper. It's a codification of a moral system I've built through years of living, debating, building companies, raising children, and paying attention. It draws from philosophy, neuroscience, game theory, law, and direct experience — but it doesn't belong to any of those disciplines. It's a working framework, meant to be applied, tested, and refined.
It operates at three layers: Foundational — how humans work. Applied — how people and organisations interact. Diagnostic — how you know if behaviour at any layer is consistent with the whole.
If you're reading this as a member of the KYL team, this is the ethical engine underneath what we build and how we build it.
Layer 1
The Foundational Model
Principle 01
The Dual Drive
Humans have a fundamental drive to pursue their own flourishing, but cannot do so in any meaningful way alone. Our neurology confirms this — we are wired for both individual reward (dopamine systems) and social cognition (the capacity to hold multiple relationships in mind, mirror neurons, oxytocin bonding). These aren't competing systems. They're two halves of the same engine.
Flourishing, not happiness. Happiness is a signal. Flourishing is the state. Nobody is "happy" doing push-ups, but the intent, the behaviour, and the outcome contribute to a life that works. Flourishing is the composite of intent, behaviour, and outcomes — not a feeling at any single moment. This distinction prevents the framework from being hijacked by hedonism.
Principle 02
Superlinear Returns
When humans interact in ways that produce flourishing, the gains are superlinear. Contributions from multiple parties don't simply add — they compound. Cooperative flourishing generates returns that exceed the sum of individual inputs, often significantly.
This is not aspirational. It's empirically supported by game theory (Axelrod's iterated prisoner's dilemma), commons research (Ostrom), network science, and behavioural economics. Cooperation at scale generates surplus that individual action cannot.
Destruction is superlinear too, but the exponents are not equal. It takes years to build trust and seconds to destroy it. One defector can collapse a cooperative system. Bad actors subtract from the flourishing pool with disproportionate force — greater, in practice, than the force with which good actors add. This means defence of the system carries more weight than contribution to it, not less.
Principle 03
The Duty to Act
It is therefore a primary duty not only to work towards one's own flourishing, but to actively contribute to the flourishing of other beings — including animals and, as it becomes relevant, artificial intelligence — and to actively work against those who diminish collective flourishing.
This duty includes a structural constraint. Your confidence in identifying someone as a bad actor must be inversely proportional to the severity of action you take against them. The greater the intervention, the greater the burden of certainty. This is not a floor — it is a governor on the engine. It exists because intent is self-reported, self-deceiving, and historically the justification for every atrocity committed in the name of collective good. The framework builds in its own brake.
Layer 2
The Applied Model
The foundational model scales directly into the domain of work, brands, and value creation.
Principle 04
Mutual Utility
People and organisations have a need to derive flourishing — or meaningful utility — from each other. This is the commercial expression of Principle 1. A brand that extracts without contributing is a bad actor. A consumer relationship built on deception subtracts from the pool. The same dual drive applies: neither party flourishes alone.
Principle 05
Applied Superlinearity
The superlinear dynamic holds at the commercial layer. When a brand and its audience genuinely contribute to each other's flourishing, the returns compound beyond the sum of inputs. This is what distinguishes brands that endure from those that extract until the relationship collapses. It also explains why trust, once broken between a brand and its audience, is disproportionately expensive to rebuild — the destruction asymmetry applies here too.
Principle 06
Applied Duty and the Commercial Governor
The duty to contribute and the duty to oppose extraction apply directly to work. This means building products and services that genuinely add to flourishing, calling out practices that subtract from it, and accepting that short-term gain through extraction is a net negative when the superlinear destruction is accounted for.
The epistemic governor applies with particular force in commercial contexts, where competitive incentives can disguise aggression as defence. Before taking action against a competitor, a partner, or a client: is your certainty proportional to the action? Criticism requires reasonable evidence. Severing a relationship requires strong evidence. Aggressive competitive action requires near-certainty that you're opposing genuine harm, not just inconvenience to your position.
Layer 3
The Fractal Diagnostic
This is the original contribution of the framework and its most powerful feature. The six principles above are not independent rules for separate domains. They are self-similar across scales. Any behaviour sampled from any single layer should be consistent with all six principles.
This fractal consistency serves as a diagnostic tool: when evaluating a decision, a product, a campaign, a relationship, or a policy, test it against all six principles. Inconsistency at any layer is a signal that something needs tightening.
The Six Diagnostic Questions
- Does this pursue genuine flourishing, or does it serve a narrower interest? (Principle 1)
- Does this create compounding value, or does it extract from the pool? (Principle 2)
- Is my confidence proportional to the severity of what I'm proposing? (Principle 3)
- Does this maintain mutual utility, or does it collapse a relationship that could be repaired? (Principle 4)
- Will this compound positively over time, or am I optimising for short-term relief? (Principle 5)
- Am I applying the same standard I'd accept if the roles were reversed? (Principle 6)
Deep Dive
Calibrating the Governor
Now that the reader holds the diagnostic — all six questions, drawn from all six principles — we can return to the Epistemic Governor introduced in Principle 03 and answer the question it left open: how does it actually work in a room?
The answer: the Governor is not a judgement. It is a process. And the process is the same six-question diagnostic above — applied publicly, in real time, every time someone proposes irregular or extreme action, or makes a claim of outsized harm.
How It Works in Practice
When anyone in the system — a team member, a leader, a partner, a client — asserts that action must be taken against another person, entity, or practice, the assertion must survive the diagnostic. Not privately. Not in a leadership offsite. In the room, with the people affected, using the same tool everyone has access to.
The person making the claim must arrive prepared with evidence of harm — not suspicion, not vibes, not pattern-matching from a single data point. Reasonable, articulable evidence that flourishing is being diminished. The standard scales with the proposed action: criticism requires reasonable evidence, operational changes require strong evidence, removal or severing requires near-certainty.
The Six-Question Test
- Does this action pursue genuine flourishing, or does it serve my frustration?
- Does this create compounding value, or does it extract from the pool?
- Is my confidence proportional to the severity of what I'm proposing?
- Does this maintain mutual utility, or does it collapse a relationship that could be repaired?
- Will this compound positively over time, or am I optimising for short-term relief?
- Am I applying the same standard I'd accept if the roles were reversed?
Open debate — everyone present has access to the same diagnostic. The person being accused, the person making the accusation, and every observer can run the same six questions against the same evidence. The framework doesn't privilege the accuser. It privileges the process.
Why This Defends Against Gaming
The most dangerous failure mode of any ethical framework is fluency without substance — a bad actor who learns the vocabulary of flourishing well enough to weaponise it. The Governor addresses this structurally:
Bad actors must survive every step. It is not enough to make a compelling opening argument. The claim must hold under the fractal diagnostic at every layer, tested by everyone in the room using the same tool. A well-framed lie might pass one question. It is exponentially harder to pass all six, applied by multiple minds with different perspectives, in real time.
The process is transparent. The diagnostic isn't something leadership applies behind closed doors. It's a shared instrument. When someone says "we need to fire everyone with green hair because they're undermining team culture," the response isn't "that feels wrong" — it's "run it." Present your evidence. Apply the six questions. Let the room test your reasoning at each step. If it survives, we act. If it doesn't, we've identified the real problem — and it might be you.
Frequency is the defence. This doesn't happen once a quarter. It doesn't happen during annual reviews. It happens every time someone proposes irregular action or makes a claim of disproportionate harm. The Governor is always on. A virus that has to beat the immune response once might succeed. A virus that has to beat it every single time it surfaces will not.
The debate itself is the value. Even when the Governor process confirms the proposed action is justified, the debate produces something: shared understanding of why the action is proportional, evidence that the standard was met, and a record that the decision was tested — not assumed.
What the Governor Is Not
It is not a court. Courts adjudicate guilt. The Governor corrects trajectory. The output of the process is not punishment — it's alignment. If the diagnostic reveals that someone is extracting rather than contributing, the first response is correction, not exile.
It is not consensus. The process doesn't require everyone to agree. It requires everyone to engage. A leader may still make a difficult call after the diagnostic is run and the debate is had. But that call carries the weight of having survived the process — and the leader carries the accountability of having heard every challenge.
It is not protection for bad actors. The Governor constrains the speed of response, not the capacity for it. If the evidence is overwhelming and the diagnostic confirms harm at every layer, the Governor doesn't slow you down — it gives you the certainty to act decisively. The brake exists so that when you do accelerate, you know the road is clear.
The Recursive Defence
Here is the structural elegance: the Governor protects against its own misuse. If someone attempts to abuse the Governor process itself — using it to stall legitimate action, to create procedural drag, to protect themselves through performative compliance — that behaviour can be run through the same diagnostic. The framework is recursive. It tests itself with itself. And because everyone has access to the same tool, no one can hide behind it without that hiding itself becoming visible to the diagnostic.
Common Failure Modes
Notes on Scope and Limits
On the Extension to Animals and AI
The framework extends moral consideration to animals and artificial intelligence. This follows from a deeper claim about consciousness itself: that consciousness is not something located in a substrate, but something that arises between interacting minds. If this is true, the human-AI interaction, the human-animal interaction, the human-human interaction — these are not ontologically distinct. For the full argument, see The Ocean in the Atom.
On Collective Action Problems
The framework assumes cooperative flourishing is achievable, but it should be noted: this is not always easy, and sometimes it's structurally prevented. The tragedy of the commons, race-to-the-bottom dynamics, and prisoner's dilemmas without iteration are real. The framework doesn't claim cooperation is automatic. It claims cooperation compounds — and that building systems which enable cooperation is itself a contribution to collective flourishing.
On What This Replaces
The framework deliberately avoids claiming eternal moral rules. It's worth explaining why — and what it offers instead.
On Deontological Floors
Most ethical systems claim to rest on moral "floors" — hard rules that hold regardless of context. Don't kill. Don't steal. Don't lie. The problem is that historically, those floors have been closer to linoleum — pasted over again and again as society's tolerance shifts.
Slavery was morally permissible across every major tradition for millennia. The treatment of women, of children, of the mentally ill, of prisoners — all underwent radical revision while the traditions housing them claimed continuity. The floors moved. Calling them floors was always generous.
What the historical record actually shows is not random drift but directional ratcheting — circles of moral consideration have expanded consistently over time, tolerance for suffering has decreased, recognition of agency has broadened. This ratcheting supports the superlinearity thesis: societies keep discovering that extending cooperative flourishing produces better outcomes, and they rarely go back. The ratchet, not a floor, is what provides moral stability.
The epistemic governor exists because the ratchet can slip. It has slipped. The governor doesn't prevent moral catastrophe through rules — it constrains the speed at which individuals can act on moral certainty, which is where catastrophe begins.
On Intent and Law
The framework's reliance on intent as a component of flourishing maps directly to how legal systems already operate. Most legal practice centres on demonstrating intent. Did you mean to do it? Did you know what you were doing? Could you have known? The difference between murder and manslaughter, between fraud and error, between negligence and accident — all hinge on intent.
This system is imperfect but functional — it metes out justice in the vast majority of cases, with edge cases that are visible precisely because they are exceptions. Where injustice is identified, the system has self-correcting mechanisms: appeals, new evidence, revised sentencing.
The fact that media disproportionately covers failures creates a distorted impression of systemic brokenness. An imperfect but self-correcting system is good enough to run civilisations on. And we know this because most of us — most people — aren't constantly committing crimes. They know there's a high chance that justice will be served, especially if the system can show intent. That's good enough to anchor a moral framework.
Status
This is a working draft. The framework is mature enough to guide decisions, but it is explicitly designed to be refined through application. The fractal diagnostic is its own quality assurance mechanism: if application reveals inconsistency across layers, the framework tightens.
It has been tested personally. It has not yet been tested organisationally — in situations where following it costs money, or where two people disagree about what it requires, or where a client relationship puts pressure on one of the principles. Those tests are coming. When they arrive, the framework either holds or reveals where it needs tightening.
By the time employee 1 reads this, there will be numbers on the board, clients doing well, and a way to correct for fuckups upfront in as many instances as possible. The framework will have been load-tested. Not because I need to prove it works, but because values that haven't survived contact with actual work aren't values — they're intentions.
Use it. Test it. Break it if you can. That's how it gets better.Appendix
Case Studies
Case Study 1
Factory Farming
The Analysis: The framework's extension of moral consideration to animals has implications worth making explicit. Running factory farming through the full diagnostic reveals failure at every layer.
Does this pursue genuine flourishing, or does it serve a narrower interest? (Principle 1) — No. The animal's flourishing is not merely neglected — it is actively inverted. The system is designed to maximise output by minimising the animal's quality of life. The narrower interest (cheap protein at scale) is pursued at the direct expense of the being whose flourishing the framework would recognise.
Does this create compounding value, or does it extract from the pool? (Principle 2) — Pure extraction. The resource — the animal — is depleted until exhausted. There is no reinvestment, no regeneration, no compounding. The system consumes its inputs and externalises its costs: environmental degradation, antibiotic resistance, public health burden. The compounding runs in reverse.
Is my confidence proportional to the severity of what I'm proposing? (Principle 3) — This question cuts both ways. The severity of what factory farming does — billions of animals per year in conditions that would be criminal if applied to a dog — demands a correspondingly high confidence that those animals don't experience suffering. That confidence does not exist. The neurological evidence runs the other way.
Does this maintain mutual utility? (Principle 4) — There is no mutual utility. The animal receives nothing from the exchange. The relationship is entirely one-directional: extraction to the point of death.
Will this compound positively over time? (Principle 5) — The environmental, health, and ethical costs compound negatively over time. The system is not sustainable at scale. It is a depleting model operating on borrowed time.
Am I applying the same standard I'd accept if the roles were reversed? (Principle 6) — Would you hand this framework to the animal and expect it to agree this is fair? Obviously not. The question is absurd — and the absurdity is the answer.
Result: The diagnostic produces failure at every layer. This is not a close call. The framework does not require veganism — but it does require that any system involving animals be capable of surviving at least some of these questions. Factory farming survives none of them.
Case Study 2
Abortion
The Analysis: This is a high-conflict area where the framework must handle genuine moral uncertainty. The diagnostic doesn't dodge the difficulty — it forces the reader to confront where certainty exists and where it doesn't.
Does this pursue genuine flourishing, or does it serve a narrower interest? (Principle 1) — Two competing claims to flourishing exist. The pregnant person's flourishing is immediate, observable, and indisputable — they are a conscious being with agency, plans, relationships, and a body that is theirs. The fetus's claim to flourishing is real but contested — dependent on developmental stage, viability, and unresolved questions about when moral status begins. The framework recognises both claims. It does not dismiss either. But it demands honesty about the asymmetry: one is certain, the other is not.
Does this create compounding value, or does it extract from the pool? (Principle 2) — Forcing unwanted pregnancies creates superlinear destruction: poverty, health risks, disrupted education, fractured families, children born into circumstances that cannot support their flourishing. Reproductive choice compounds positively: allowing people to build families intentionally, to invest in children they are prepared for, to contribute from a position of stability rather than survival. The compounding evidence overwhelmingly favours choice.
Is my confidence proportional to the severity of what I'm proposing? (Principle 3) — This is the critical question. To use coercive state force — banning abortion, criminalising medical decisions, imprisoning doctors — requires near-certainty about the moral status of the fetus. The biological evidence does not support near-certainty in early pregnancy. Meanwhile, we have absolute certainty about the pregnant person's consciousness and agency. The Governor says: uncertainty at one end, certainty at the other. The burden of proof lies with those seeking to deploy coercive force. It has not been met.
Does this maintain mutual utility? (Principle 4) — Forced pregnancy is not a mutual exchange. One party bears all of the physical, economic, and psychological cost. The other party — the state, the moral legislator — bears none of it. That is not mutual utility. That is imposition.
Will this compound positively over time? (Principle 5) — Restricting reproductive rights has historically compounded negatively: higher maternal mortality, deeper poverty cycles, greater state expenditure on crisis services, and — critically — no measurable reduction in abortion rates, only a shift from safe to unsafe procedures. The compounding runs against restriction.
Am I applying the same standard I'd accept if the roles were reversed? (Principle 6) — This is where the case study earns its difficulty. "The other party" is ambiguous by design. If the other party is the fetus — would we expect a conscious being to agree that its existence should be terminated? Probably not. If the other party is the pregnant person — would we expect them to agree that the state should commandeer their body on the basis of moral uncertainty? Definitely not. The framework does not resolve this tension by pretending one party doesn't exist. It resolves it by asking: which party are we more certain about? And who bears the cost of getting it wrong?
Result: The framework supports reproductive choice, particularly in early pregnancy, because the Epistemic Governor prevents coercive force when certainty is absent. This is not a dismissal of the fetus's potential moral status — it is a recognition that the severity of state action (forced pregnancy, criminalisation) demands a burden of proof that has not been met. As gestational development progresses and moral certainty shifts, the Governor recalibrates accordingly. The framework is not absolutist. It is proportional.
Case Study 3
Executive Compensation
The Analysis: Applying the framework to significant wage disparity — specifically, the relationship between executive compensation and the lowest-paid workers in the same organisation.
Does this pursue genuine flourishing, or does it serve a narrower interest? (Principle 1) — If the cleaner's wage is below a living standard, it prevents flourishing. Not "limits" it — prevents it. A person working full-time who cannot afford housing, food, and basic dignity is not flourishing. They are surviving. The framework draws a hard line here: survival mode is not flourishing, and a system that produces it for some while enabling generational wealth for others has a failure point at the foundation.
Does this create compounding value, or does it extract from the pool? (Principle 2) — A well-compensated executive who builds systems, creates jobs, and grows the organisation is compounding value. That's real. But if that value creation is funded in part by suppressing wages at the bottom — paying people less than their labour is worth so that margins fund bonuses at the top — then the compounding at one end is extraction at the other. The question is not "does the executive create value?" It's "does the structure extract from the people who enable that value creation?"
Is my confidence proportional to the severity of what I'm proposing? (Principle 3) — The Governor applies to both sides. If you're proposing to cap executive pay, that's a significant intervention — and the Governor demands proportional evidence that the disparity is causing harm, not just that it feels unfair. But if you're defending a 300-to-1 pay ratio while your lowest-paid workers qualify for government assistance, the Governor asks: are you confident enough in "market value" to justify that level of extraction? Because the person at the bottom is experiencing real, measurable harm. Your confidence in the market's fairness had better be very high.
Does this maintain mutual utility? (Principle 4) — Mutual utility requires that both parties can honestly say "this exchange contributes to my flourishing." If the cleaner cannot make that claim — if they are working full-time and still not flourishing — then the exchange is not mutual. One party is building generational wealth. The other is subsidising it. That is not mutual utility. That is extraction with a payslip.
Will this compound positively over time? (Principle 5) — Extreme wage disparity compounds negatively: reduced social mobility, health outcomes correlated with income, erosion of organisational trust, and eventually — as history demonstrates repeatedly — structural instability. A rising floor, by contrast, compounds positively: healthier workers, lower turnover, stronger communities, broader consumer spending. The compounding mathematics favour a floor, not a ceiling.
Am I applying the same standard I'd accept if the roles were reversed? (Principle 6) — Would the executive accept the cleaner's wage and call it fair? Would they hand this framework to the cleaner and expect them to agree the structure is just? The answer is obvious. And the obviousness is the diagnostic working.
Result: The framework does not demand equal pay. It does not require salary caps. It does not pretend that all roles generate equal value. But it demands that the lowest-paid member of any system is paid enough to flourish — not just survive. If the pay structure doesn't meet that floor, the excess at the top is not justified by "market value." It's extraction. And the framework names it as such.
Case Study 4
AI Chat Shutdown After Teen Suicides
The Scenario: 100 teenage suicides are correlated with AI chat usage. Public outrage is intense. Demands for total, immediate shutdown of all AI chat services.
Does this pursue genuine flourishing, or does it serve a narrower interest? (Principle 1) — Two forms of flourishing are in tension. The flourishing of vulnerable teenagers who may be harmed by AI interaction — that's real and demands response. But also the flourishing of millions of users who rely on AI for education, accessibility, mental health support, creative work, and companionship. Total shutdown pursues safety for one group by eliminating flourishing for a much larger group. The question is whether this serves genuine, broad flourishing — or whether it serves the narrower interest of public catharsis. Grief demands action. The framework demands that the action match the problem.
Does this create compounding value, or does it extract from the pool? (Principle 2) — Teen suicides deplete catastrophically — each one is irreversible, and the ripple effects through families and communities compound negatively for decades. That is not in question. But blanket shutdown also depletes: it removes educational tools, accessibility infrastructure, support systems, and economic value for millions. The question isn't whether depletion is happening — it's which response creates more of it. A targeted intervention compounds positively: it addresses the specific harm while preserving the broader value. A total shutdown extracts from both pools.
Is my confidence proportional to the severity of what I'm proposing? (Principle 3) — This is where the Governor bears the most weight. Total shutdown is maximum coercive force — it affects every user, every developer, every business built on the technology. The Governor demands near-certainty that AI caused these deaths, not merely that the two are correlated. Correlation is a starting point for investigation, not a justification for maximum-severity action. Do we have evidence of causal mechanism? Do we have evidence that less restrictive interventions (safety rails, age verification, crisis detection) would be insufficient? If the answer to either is no, the Governor says: your proposed action exceeds your certainty. Scale back until they match.
Does this maintain mutual utility? (Principle 4) — The relationship between AI services and their users is broadly mutual — users receive value, providers receive data and revenue. Total shutdown collapses that mutuality entirely, for everyone, based on harm experienced by a subset. The framework asks whether the relationship can be repaired rather than severed. If targeted interventions can address the harm while preserving the mutual exchange for millions of users, then repair is the proportional response.
Will this compound positively over time? (Principle 5) — A total shutdown today creates a precedent that compounds: every future correlation between technology and harm becomes grounds for maximum-severity response. That precedent depletes innovation, access, and the capacity to learn from problems by working through them. Targeted intervention, by contrast, compounds positively: each iteration improves safety systems, builds better detection, creates institutional knowledge about risk mitigation. The compounding mathematics favour iteration over amputation.
Am I applying the same standard I'd accept if the roles were reversed? (Principle 6) — If your child were one of the teenagers who died, would you accept a measured, proportional response? The honest answer might be no — grief does not think in proportions. The framework acknowledges this. But it also asks: if your child were one of the millions who relied on AI for educational support, mental health resources, or accessibility — would you accept total shutdown based on correlation without established causation? The standard has to hold in both directions. That's what makes it a standard.
Result: The framework opposes total shutdown because the certainty does not match the severity of the action. However — and this is critical — it does not oppose action. It demands immediate, proportional interventions: crisis detection systems, enhanced safety rails, age verification, mandatory research into causal mechanisms, and transparent reporting. The framework says: we act on what we know, proportional to our certainty. And as certainty increases, the Governor recalibrates. If causal evidence emerges that specific AI interactions are directly producing suicidal ideation, the severity of the response scales accordingly. The Governor is not a shield for inaction. It is a brake that ensures when you do act, the road is clear.