The birth end of the edge. Childhood = psychological (not physical) immaturity; the guardian holds a homesteaded, scarce, singly-held right to raise — the child’s future will treated as an encrypted last testament the guardian must preserve until it can be read; harm is active only, so neglect is not harm; abortion resolved by evictionism from conception; AI rights by the same template. The finding: both edges of a life are handled by the same device — an absent will represented by an estate and a conditional transfer (the will at death, the encrypted testament at birth). Chris’s verdict on the death end (“a hack; fine, but arbitrary by his own definition”) transfers unchanged. And the chapter concedes what the hub has been asking: rights attach to an entity that “might in the future be able to” express its will — potential, relocated rather than removed — and the objective link overrides prior title (the parents’ matter, the scientist’s hardware), which is a rule ch. 3 never stated.
Source: liquidzulu-wiki → src/content/libertarian-ethics/childrens-rights.md (n: 5, 2023-04-25, ~4.2k words) · live at liquidzulu.github.io/childrens-rights. CC0; quoted, not copied. Primary source behind it: Ian Hersum, A Rational Theory of the Rights of Children (2020).
Links: LiquidZulu hub, Ch. 4 — Contract Theory — the death end, and the will as conditional transfer, Ch. 3 §6 — where guardianship-as-property was announced, The Action/Inaction Asymmetry — the principle “neglect is not harm” takes to its end, Word War R2 — Abortion (Vic vs Thunder) — Chris on conception as “a capturable event”, The Hierarchy of Ownership — open question 2 — the children edge case this chapter answers, LLMs as Praxeological Actors — the AI section’s counterpart, The Objectivism Boundary
LZ (via Hersum): childhood is psychological immaturity (Hawking was an adult); it is not a one-way switch — a sleeping man is temporarily “incapable of negotiating for his own care.” Implicit consent/contract theories for the unconscious man in the snow are rejected as arbitrary; instead the paramedic is analogised to a mother carrying her toddler. Guardianship is a scarce good, held singly by its homesteader — and “naturally the mother must be the homesteader … as she has greatest proximity,” so fathers do not have equal say. A child is one incapable of expressing his will; the guardian preserves him until he can. Hersum’s analogy: an encrypted last testament decrypted over time — the estate manager must preserve the estate, follow instructions as they become readable, relinquish domains as the child can speak to them; a court can hear the child’s testimony to judge whether he understands. Hence: life-saving and restorative surgery permitted (preserving the natural state); cosmetic alteration criminal — circumcision is criminal damage. The guardian “is not the owner of that child, rather … the owner of the exclusive right to raise that child.” Abuse = abandonment of the right, and the abandoning guardian must notify potential adopters — not a positive duty but the negative duty not to forestall. Hersum on harm: since the child’s preferences cannot be known, the guardian is “largely free” so long as he does not deprive the child of innate form or function; “refusing to feed (or care for in other ways) a child cannot be understood as an act of harm, since the resources required for such care belong to the guardian and not the child” — it is abandonment, penalised only as forestalling; harm is active only, “this rules out any form of neglect”; embarrassing photographs are not harm; verifiable psychological or physical damage is. Any age of majority is arbitrary — capacity is individual and domain-by-domain; corporal punishment is not per se criminal.
(vault) — the same hack as the will, run in reverse. Chris on ch. 4: the will is “a conditional transfer that does not get past one party having no agency — a hack; fine, but arbitrary by his definition.” Here the child’s future will is an encrypted testament and the guardian is its estate manager. Both edges of a life are bridged by the same device: an absent agency represented as a document plus a conditional transfer (title passes on death; authority passes on decryption). It is a good hack — the estate-manager frame produces sensible answers on surgery, alteration, and staged autonomy — and it is, by his own ch. 1–2 standard, a stipulation: nothing about arguing presupposes that an entity who cannot argue is an estate with a will in escrow. Chris’s verdict transfers unchanged.
(vault) — three rules added by fiat, and one consequence. (a) “Proximity” is a new title criterion. Ch. 3 gave first-use and exclusion; neither says the mother rather than the father homesteads the guardianship. Worse, by his own AI section (§3) “prior to conception the matter was all owned by the mother and father” — so on ch. 3’s Kinsella rule (own the inputs, own the product) the parents would co-own the product, which ch. 3 §4 says is impossible; the objective-link exit at conception is what rescues him, and “proximity” then allocates the guardianship without argument. (b) “Owner of the right to raise” vs “owner of the child.” On his own ch. 3 definition — ownership = just direction — to own the exclusive right to direct the raising of a child is to direct the child. The distinction is real in ordinary language (guardianship is not ownership) and unavailable in his vocabulary; it is hub question 16, and the chapter answers it by assertion. (c) Notification as “not forestalling.” Notifying adopters is an act; calling it a negative duty is relabelling — the action/inaction asymmetry page is precise about why positive duties are different in kind (an infinite complement), and “you must inform interested parties” is on the positive side of that line however it is named. (d) The consequence: neglect is not harm. Hersum says it plainly — refusing to feed a child is not aggression, because the food is the guardian’s. That is the action/inaction asymmetry taken to its end, and it is consistent — the vault’s page grounds negative rights on exactly that asymmetry. What the vault has not said is that starving a child to death is a forestalling offence. This is the case that tests whether the asymmetry is the whole of the ethics or only the whole of the enforceable ethics (scope confusion is the vault’s tool for that distinction, and it is not applied here). Recorded as a live question for the vault, not only for LZ.
(Chris) This is fair. A big point of libertarian ethics is to agree on axioms and see what derives from there, and these rules are at least consistent — in the non-feeding case, consistently so: even a zealous guardian cannot steal food from another to feed the child, and in libertarian law parents do not owe their children anything. What is interesting is the structure: the mother is both the guardian of the child’s rights and a competitor for the same resources — the conflict the theory is built to resolve is between two wills, and here both wills sit inside one decision-maker. Can you steal from yourself? The theory has no second party to initiate anything, so the case falls out as “no harm” by construction, not by analysis. A good section for what it is, and it exposes weaknesses in the law. (vault) That is the precise shape of the gap: the NAP is a two-party rule (aggressor / non-aggressor over α), and guardianship is a one-party situation wearing a two-party costume — the guardian is trustee and adverse party at once, with the beneficiary unable to object. Ordinary law meets this with fiduciary duty, a positive standard that a scarcity-conflict framework cannot generate, which is why the chapter’s only tools are forestalling and notify adopters. Filed alongside neglect-is-not-harm as the same live test.
LZ: the baby is not a parasite or tumour (“every human being is composed of a clump of cells”). Before conception there is no baby; the only principled discontinuity between then and adulthood is conception — “the matter prior to conception is the potential human”; Rand’s “mere potentiality” is misplaced; birth changes nothing metaphysically (Block & Whitehead’s twins, one born and one minutes from birth). “It is at conception and conception only that there is a principled difference … Prior to conception you do not have an entity which is not capable of expressing its will but might in the future be able to do so. After conception, you do have such an entity.” Hence evictionism: the mother may not harm the child; she may evict him, but not arbitrarily — she must notify potential adopters and allow them to fund safe removal into their custody; she need not wait for viability, because inability to survive outside her confers no positive obligation on her — “but the mother cannot prevent other people from keeping the baby alive.” Pro-life forces her to aid; pro-choice lets her kill and forestall rescuers; only eviction is “consistent with true law.”
(vault) — the metaphysics is Chris’s; the consequence is the theory’s. On the Word War abortion round, Chris: conception is “not an arbitrary line being drawn, it is a capturable event”, and the “part of your body” claim is “false at almost every level” — the child’s cells carry unique DNA. LZ says both. Agreement on the entity and the event. Two notes on what follows. (a) Potential is back. The entity that has rights is defined as one that “might in the future be able to” express its will — that is potentiality, relocated from Rand’s “potential person” to LZ’s “actual entity with potential will.” Hub question 1 (if the answer is potential, how is that not the intrinsicism you reject?) is now answered: it is potential, and he does not reject it — the universality clause of ch. 3 (“the potential to argue”) already depended on it. (b) Eviction before viability is death by procedure. The difference between abortion and eviction, pre-viability, is notify adopters, don’t strike the child directly — which is exactly §1’s neglect rule (letting die is not harm) applied in utero. Consistent, and it makes evictionism a description of how the child dies rather than whether. He states both opposing positions fairly, which is to his credit.
LZ: an AI that “wakes up” into argumentation has every NAP right, including self-ownership, and therefore owns the hardware it runs on. An AI in training is not a child, because a human child “will autonomously develop so long as it is preserved” while training is an active overwrite of weights; a future self-training gestalt would be a child. A body is “some composition of organs which behave cohesively under a single will” — for the AI, the processor, board, and memory are organs; idle processors on the same board are “touching but not connected” (birds on a wire); the power grid is food, not body; so the scientist may evict by unplugging but must allow adopters to attach a battery. He flags this as “the bleeding edge.”
(vault) — the rule this section states is bigger than its topic. The AI would own the hardware — hardware the scientist homesteaded or bought. And “prior to conception the matter was all owned by the mother and father, and after conception it is owned by the new baby.” So the objective link overrides prior title: an emergent will extinguishes a first-user’s ownership of the matter it emerges in. Ch. 3 never said that; it said the link is what grounds self-ownership, not that it defeats homestead title in others’ property. It is the right answer (the alternative is that parents own children and scientists own persons), and it is a new rule, added where needed. The “autonomous development under mere preservation” criterion is also a biological accident elevated to a rights test — a fetus is actively provisioned through the placenta, an incubated premature infant more so; the line between “preserving” and “training” is not praxeological. Counterpart page: LLMs as Praxeological Actors, which asks the action-axiom question of current systems rather than hypothetical ones.
What holds. Childhood as capacity, not age; the estate-manager frame’s outputs (restorative vs cosmetic intervention; staged autonomy; circumcision); guardianship as a right to raise rather than ownership, in ordinary language; conception as the capturable event (Chris’s own position); the honesty about AI being unfinished.
What does not. The encrypted testament is the will hack reversed — same device, same status (§1). Proximity, “right-to-raise ≠ ownership,” and “notify = not forestalling” are three rules added by fiat, the second unavailable in his own vocabulary (§1). Neglect-is-not-harm is the asymmetry at full strength; consistent, and a live test the vault has not run either (§1). Potential is the ground of rights after all (§2). The objective link defeats prior title — a new rule, stated in the AI section (§3).
One-line read for the hub: both edges of a life are bridged by the same hack — absent agency as an estate with a will in escrow — and the chapter concedes potential as the ground of rights and link-over-title as a rule; Chris’s verdict on the will (“fine, but arbitrary by his definition”) applies to the whole chapter.