Skip to main content

UHM Ethics

UHM connects ethics with the organisation of living and experiencing systems: their ability to persist, understand, act, cooperate and renew themselves. Its mathematical contribution is to make the conditions and consequences of a choice explicit. Mathematics can establish that a policy preserves an adopted value or violates a constraint. Selecting that value remains a normative commitment.

The guiding proposal of this chapter is to protect subjects and the conditions of their flourishing, while developing understanding and responsible agency. This is richer than maximising a matrix statistic. Rest, care, learning and preserving a valuable relationship can matter even when a proposed scalar does not increase.

Status and scope

[D] denotes a definition, [N] an adopted normative premise, [T] a mathematical result under its stated hypotheses, [H] an empirical bridge, and [I] an interpretation. A theorem about a formal model does not by itself identify biological life, suffering or moral status.

This chapter supersedes the former derivation of all ethics from P˙>0\dot P>0. The mathematical kernel, viability, interiority hierarchy and composite-state typing govern its notation. A comparative treatment of spiritual concepts is in Spiritual synthesis.

1. What the formalism establishes​

For Γ∈D(C7)\Gamma\in\mathcal D(\mathbb C^7),

P(Γ)=Tr⁡Γ2,17≤P≤1.P(\Gamma)=\operatorname{Tr}\Gamma^2,\qquad \frac17\le P\le1.

The cut P>2/7P>2/7 is the model's structural-majority criterion. Its algebraic derivation is exact; its identification with biological viability requires a separate bridge. Persistence requires dynamics, resources, controls and a disturbance class. Conditional irreversible decay requires, for example, a disabled regeneration gate and a fixed unital primitive generator with no external purification. A threshold crossing alone does not establish death or irreversible loss of experience.

Even elementary decay must respect the state space. For Γ˙=γ(I/7−Γ)\dot\Gamma=\gamma(I/7-\Gamma),

P(t)=17+(P(0)−17)e−2γt.P(t)=\frac17+\left(P(0)-\frac17\right)e^{-2\gamma t}.

It approaches 1/71/7 and never goes below it. The former example P(t)=0.25e−0.05tP(t)=0.25e^{-0.05t}, eventually giving P<1/7P<1/7, was not a valid density-state evolution.

L0 in the canonical hierarchy holds for every valid matrix. L1 and higher predicates require declared experiential and operational data. Neither P>2/7P>2/7 nor a numerical hierarchy level is a scale of human worth. Likewise, a seven-dimensional basis axis is not a tensor factor: Tr⁡−EΓ\operatorname{Tr}_{-E}\Gamma is unavailable without a specified tensor decomposition. The extension state used in D=exp⁡S(ρE)D=\exp S(\rho_E) must be normalised; a 42D lift and conditioning cannot be recovered from purity alone.

Why purity is not a complete good​

Definition [D]. The local purity contribution of an intervention aa is

GP(a,x)=ddtP(Γa(t))∣t=0.G_P(a,x)=\left.\frac{d}{dt}P(\Gamma_a(t))\right|_{t=0}.

It is one diagnostic, not the definition of every benefit or harm. Three exact obstructions rule out the previous identification:

  1. Purity does not determine organisation in a semantic frame. Both ∣1⟩⟨1∣|1\rangle\langle1| and ∣u⟩⟨u∣|u\rangle\langle u|, with u=7−1/2∑i∣i⟩u=7^{-1/2}\sum_i|i\rangle, have P=1P=1. Their canonical integration values Φ=(P−Q)/Q\Phi=(P-Q)/Q, Q=∑iγii2Q=\sum_i\gamma_{ii}^2, are 00 and 66.
  2. Preservation can be valuable. A viable stationary trajectory has P˙=0\dot P=0. Calling it indifferent would discard successful maintenance and care. Moreover, P˙≥c>0\dot P\ge c>0 cannot hold forever, since P≤1P\le1.
  3. Purification is not automatically development. Under the canonical definition R=1/(7P)R=1/(7P), increasing PP decreases RR. This RR is a scalar gate, not a calibrated score of self-knowledge. No theorem makes all chosen indicators increase together.

Normative proposal [N]. Good actions protect the conditions of subjects' existence and agency and improve justified, contextually specified goods without violating the adopted protections. This permits meaningful disagreement about ends and priorities while making their consequences examinable. The word “good” is not selected uniquely by the density-matrix axioms.

Self-modelling and values​

A self-model does not determine a value function. On the same state space, an exact model m(x)=xm(x)=x is compatible with U(x)=0U(x)=0, U(x)=P(Γ)U(x)=P(\Gamma), or many other objectives. Thus reflection alone does not prove the impossibility of nihilism or the inevitability of compassion. A theory connecting affect, preference and self-maintenance needs explicit causal hypotheses and observations. Normative responsibility can be adopted even when empathy is weak or uncertain.

2. A typed ethical context​

Definition [D]. Fix a context κ\kappa containing:

  • a declared set II of affected subjects, including persons whose interests are represented by others;
  • an augmented state x∈Xκx\in X_\kappa, containing Γi\Gamma_i where available, resources, environment, relevant history and authority to act;
  • admissible policies π\pi, horizon TT, initial information and a scenario class Ωκ\Omega_\kappa;
  • trajectories xπ,ω(t)x^{\pi,\omega}(t) supplied by the declared causal model;
  • protected conditions, benefit criteria, consent procedures and a rule for revising the model.

Unknown matrix reconstructions remain unknown. They do not remove an affected person from II. A complete joint UHM model, if used, must supply compatible marginals of a declared tensor-product state. An individual and a collective cannot silently be counted twice as the same benefit.

Protected conditions and dignity​

Normative premises [N]. This proposal adopts protection against severe avoidable harm, respect for agency and consent, and equal consideration of persons. Greater measured capability does not grant greater basic dignity. These premises are not inferred from PP, Φ\Phi, RR, computational power, productivity or spiritual attainment.

Let hir(x)≥0h_{ir}(x)\ge0 describe specified protective floors for subject ii and requirement rr, and let CκC_\kappa record additional consent and procedural constraints. Define

Sκ={x:hir(x)≥0 for all i,r, Cκ(x)=1}.S_\kappa=\{x:h_{ir}(x)\ge0\ \text{for all }i,r,\ C_\kappa(x)=1\}.

A floor is an operational requirement, not a numerical dignity score. Its calibration and exceptions require justification. Historical consent, obligations and remaining resources must be included in the state whenever they affect admissibility.

The robustly permitted policies are

Πκsafe(x0)={π:xπ,ω(t)∈Sκ ∀ω∈Ωκ, 0≤t≤T}.\Pi_\kappa^{\mathrm{safe}}(x_0)= \{\pi:x^{\pi,\omega}(t)\in S_\kappa\ \forall\omega\in\Omega_\kappa,\ 0\le t\le T\}.

This quantification concerns the declared model and scenario class, not every conceivable world. If this set is empty, the correct result is infeasibility under the stated commitments. A least-harm emergency rule, a changed horizon or revised protections must be introduced explicitly and reviewed; an optimiser must not silently remove a person's constraint.

This replaces the former universal six-level ranking. Vital, homeostatic, relational, cognitive, aesthetic and contemplative goods remain useful dimensions of deliberation, but no density-matrix theorem orders them completely. A short-term cost can serve a justified longer-term good; it must be evaluated over the actual trajectory, with alternatives and uncertainty.

A sufficient preservation certificate [T]​

Suppose trajectories exist on [0,T][0,T], hir(x(t))h_{ir}(x(t)) is absolutely continuous, hir(x0)≥0h_{ir}(x_0)\ge0, and an admissible policy satisfies, for every declared disturbance and almost every tt,

ddthir(x(t))≥−αir(t)hir(x(t)),αir≥0,αir∈L1([0,T]).\frac{d}{dt}h_{ir}(x(t))\ge-\alpha_{ir}(t)h_{ir}(x(t)), \qquad \alpha_{ir}\ge0,\quad \alpha_{ir}\in L^1([0,T]).

Then every numerical floor is preserved:

hir(x(t))≥hir(x0)exp⁡ ⁣(−∫0tαir(s) ds)≥0.h_{ir}(x(t))\ge h_{ir}(x_0) \exp\!\left(-\int_0^t\alpha_{ir}(s)\,ds\right)\ge0.

Proof. Multiply by the integrating factor and integrate the resulting nonnegative derivative. This does not establish CκC_\kappa: consent and procedural obligations need their own preservation check. It also does not prove that a suitable policy exists or that the causal model is correct. The substantive result is conditional: an adopted protection can have an explicit dynamical certificate.

3. Value, development and choice​

Let bik(x,u)b_{ik}(x,u) be a declared benefit rate and bikterm(x)b^{\mathrm{term}}_{ik}(x) a terminal criterion. Define

Jik(π,ω)=∫0Tbik(xπ,ω(t),uπ,ω(t)) dt+bikterm(xπ,ω(T)),Vik(π)=inf⁡ω∈ΩκJik(π,ω).J_{ik}(\pi,\omega)=\int_0^T b_{ik}(x^{\pi,\omega}(t),u^{\pi,\omega}(t))\,dt +b^{\mathrm{term}}_{ik}(x^{\pi,\omega}(T)), \qquad V_{ik}(\pi)=\inf_{\omega\in\Omega_\kappa}J_{ik}(\pi,\omega).

Assume the quantities are defined and finite. The vector V(π)\mathbf V(\pi) can track well-being, understanding, ability to act, relationship quality and other adopted goods. A worst-case rule is itself a choice; expectation would additionally need a justified probability law. Different components may attain their worst cases in different scenarios.

Local contributions and their limits​

For a differentiable admissible state curve Γ(α)\Gamma(\alpha), with Γ(0)=Γ\Gamma(0)=\Gamma and derivative Δ\Delta,

ddαP(Γ(α))∣0=2Tr⁡(ΓΔ).\left.\frac{d}{d\alpha}P(\Gamma(\alpha))\right|_0=2\operatorname{Tr}(\Gamma\Delta).

Here Δ\Delta is Hermitian and traceless; at a boundary state these conditions alone do not guarantee a positive curve. For the affine path, when admissible,

P(Γ+αΔ)=P(Γ)+2αTr⁡(ΓΔ)+α2Tr⁡(Δ2).P(\Gamma+\alpha\Delta)=P(\Gamma)+2\alpha\operatorname{Tr}(\Gamma\Delta) +\alpha^2\operatorname{Tr}(\Delta^2).

The derivative is linear in specified perturbations; finite effects include cross terms and interactions. Neither an arbitrary object XX nor a person's intention uniquely specifies ΔX\Delta_X. Mathematical uniqueness given inputs is also distinct from effective computability from observations.

Development can enlarge a person's supported repertoire, improve error correction or deepen understanding while leaving PP unchanged. Its value is recorded in the appropriate criterion. Autogeny's distinction between a description, its execution and its sustained reproduction matters here: announcing a higher self-model is not evidence that it works, persists or respects its resource budget.

Pareto choice and aggregation​

Within Πκsafe\Pi_\kappa^{\mathrm{safe}}, policy π\pi dominates π′\pi' if all components of V(π)\mathbf V(\pi) are at least as large and one is strictly larger. Nondominated policies form the Pareto set [D]. It need not contain a unique best policy; useful deliberation may consist in exposing the remaining conflict.

Proposition [T]. For finitely many criteria and strictly positive weights wikw_{ik}, any attained maximum of ∑i,kwikVik(π)\sum_{i,k}w_{ik}V_{ik}(\pi) is Pareto efficient.

Proof. A dominating policy would strictly increase the weighted sum. Existence needs additional assumptions, for example a nonempty compact policy set and continuous criteria. No ethical uniqueness follows. A maximin rule also requires commensurable, explicitly normalised criteria; it is a normative alternative, not a theorem selected by Γ\Gamma.

The weights and floors are public parts of the decision. They must not disguise greater basic worth for agents with larger matrices, stronger reflection or higher productivity.

4. Relations, compassion and cooperation​

Responsibility beyond measured empathy​

Simultaneous holons use

ΓAB∈D(C7⊗C7),ΓA=Tr⁡BΓAB,ΓB=Tr⁡AΓAB.\Gamma_{AB}\in\mathcal D(\mathbb C^7\otimes\mathbb C^7),\qquad \Gamma_A=\operatorname{Tr}_B\Gamma_{AB},\quad \Gamma_B=\operatorname{Tr}_A\Gamma_{AB}.

A selected cross-observable can measure a declared correlation. It does not automatically measure access to experience, causal influence, consent or compassion. In particular, 1−GapAB(E,E)1-\mathrm{Gap}_{AB}(E,E) requires a typed observable and empirical validation before being called an empathy measure. A zero phase with undetectably small amplitude cannot certify a strong relationship.

From another's interests to one's responsibility​

Compassion enters the normative model by including another's welfare and protections even when they do not improve one's own reward. A nonzero correlation does not imply that harming BB lowers AA's utility or purity. Such a sign conclusion needs a specified response law, intervention and value function. The proposed ethical commitment extends to distant persons and future subjects; it is not restricted to systems for which an E-correlation has been measured.

Reciprocity and the golden rule​

Hermiticity, γab=γba‾\gamma_{ab}=\overline{\gamma_{ba}}, is a state-space condition, not symmetry of causal effects or obligations. A reciprocity theorem needs an agent-exchange map σ\sigma, equivariant dynamics and admissible policies, and exchanged value functions. Under these assumptions, exchanging the entire scenario and policy exchanges the corresponding outcomes; at an exchange-invariant initial scenario it gives equality of the exchanged values. This follows by mapping trajectories through σ\sigma.

Adopting equal consideration under exchange is a normative principle [N]. It does not imply identical treatment under unequal needs, nor does it identify another person's preferences with one's own. The golden rule is therefore a useful deliberative perspective, not a corollary of matrix symmetry.

Exact content of T-77: purity and pinching​

Let Πa\Pi_a be mutually orthogonal projections with ∑aΠa=I\sum_a\Pi_a=I, and define the pinching map E(ρ)=∑aΠaρΠa\mathcal E(\rho)=\sum_a\Pi_a\rho\Pi_a. For every density matrix,

P(ρ)=P(Eρ)+∥ρ−Eρ∥F2.(T-77)P(\rho)=P(\mathcal E\rho)+\|\rho-\mathcal E\rho\|_F^2. \tag{T-77}

Proof [T]. Pinching is an orthogonal projection for the Hilbert–Schmidt inner product, so its image and residual are orthogonal. Apply the Pythagorean identity. For two blocks ρ=(ABB†D)\rho=\begin{pmatrix}A&B\\B^\dagger&D\end{pmatrix}, the residual term is 2∥B∥F22\|B\|_F^2.

This compares one state with its pinched version, keeping its diagonal blocks fixed. If A=pρAA=p\rho_A, D=(1−p)ρBD=(1-p)\rho_B, then

P(Eρ)=p2P(ρA)+(1−p)2P(ρB).P(\mathcal E\rho)=p^2P(\rho_A)+(1-p)^2P(\rho_B).

A direct sum describes alternative sectors, not simultaneous subsystems. The expression P(ρA)+P(ρB)P(\rho_A)+P(\rho_B) omits normalisation. No inclusion–exclusion law or “free purity from cooperation” follows.

Counterexample to local benefit from joint purity. The state I49/49I_{49}/49 and the maximally entangled state ∣Ψ⟩=7−1/2∑i∣ii⟩|\Psi\rangle=7^{-1/2}\sum_i|ii\rangle both have marginals I7/7I_7/7, while their joint purities are 1/491/49 and 11. Thus increasing joint purity need not improve either marginal purity. Conversely, a pure product and ∣Ψ⟩|\Psi\rangle both have joint purity 11, although their local purities differ. No ordering of subjects' welfare follows from these numbers.

A genuine cooperation result must specify resources, policies, alternatives and outcome criteria, then prove or measure an improvement relative to independent action. T-77 provides a matrix identity within that investigation; it does not complete it.

5. Integrity, practice and revision​

Alignment of knowledge, intention and action is a meaningful goal [N/I]. The phase statistic Gap(L,D)\mathrm{Gap}(L,D) becomes evidence about it only after specifying an encoding and testing the proposed bridge.

Counterexample [T]. Set

∣ψθ⟩=∣L⟩+eiθ∣D⟩2,Γθ=∣ψθ⟩⟨ψθ∣.|\psi_\theta\rangle=\frac{|L\rangle+e^{i\theta}|D\rangle}{\sqrt2},\qquad \Gamma_\theta=|\psi_\theta\rangle\langle\psi_\theta|.

All these matrices are positive, trace one and have P=1P=1, while Gap(L,D)=∣sin⁡θ∣\mathrm{Gap}(L,D)=|\sin\theta| varies from 00 to 11. At fixed entry moduli, P=∑ij∣γij∣2P=\sum_{ij}|\gamma_{ij}|^2 is phase-independent. Hence the former equivalence “minimise Gap iff maximise purity” is false. A claim about a stationary purity under changed phases would instead require an explicit phase-dependent generator and a new dynamical proof.

Practice can be formalised as repeated transitions satisfying an invariant and reducing a declared error. If a verified update law satisfies En+1≤qEnE_{n+1}\le qE_n with 0≤q<10\le q<1, then En≤qnE0E_n\le q^nE_0 by induction. The contraction assumption must be established for the relevant task; it is not a universal law of meditation, virtue or learning. Preserving agency and the ability to revise a mistaken objective is part of the proposed practice, not an obstacle to perfection.

A category of permitted trajectories [D/T]​

Fix κ\kappa. Objects are augmented states in SκS_\kappa. Arrows x→yx\to y are finite-duration trajectories admitted by a concatenation-closed control model, remaining in SκS_\kappa and ending at yy. Relevant time, resources and obligations are state coordinates. Identity arrows have zero duration; composition concatenates paths whose endpoints match. With duration recorded as part of a path, concatenation is associative and the identities act as units: this defines a category.

This construction concerns realised admissible paths. A category of robust policies additionally needs closed-loop composition and compatible uncertainty sets; it cannot be inferred from the path category. A translation between implementations preserves ethical certificates only when it maps the protected set and admissible transitions appropriately. Being a functor, or sharing a semantic description, does not supply these conditions.

The practical connection to Autogeny is precise: an agent's reconstruction and self-modification must preserve stated properties through actual execution, with explicit time and resource costs. Self-reference without such evidence does not establish maturity or moral reliability.

6. Beauty and contemplative goods​

Beauty can be an adopted good and a subject of empirical study. It is not presently characterised by the necessary-and-sufficient formula P˙>0\dot P>0, Φ>1\Phi>1. An aesthetic episode may involve grief, tension, surprise, restoration or sustained attention; assigning those experiences a matrix signature is a research hypothesis, not their definition.

A plural description of aesthetic value​

A study can distinguish reported enjoyment, perceived form, understanding, attention and lasting significance, then compare them with specified state estimates. These observables may disagree. An elegant model can be false; aesthetic satisfaction does not replace a proof, external validation or another person's testimony.

Contemplative goods may include reduced compulsive reactivity, clarity of attention, compassion and openness to correction [N/I]. A spiritual report does not by itself identify a state as L3 or L4. Tested metamodel capabilities, subjective experience and ethical conduct are different evidential questions.

7. Difficult applications​

Care under uncertain consciousness​

No validated clinical reconstruction in this chapter permits a verdict about awareness, prognosis or life-support decisions from PP, RR or rank⁡ρE\operatorname{rank}\rho_E. The previous formula P→0P\to0 was outside the density-state domain, and the corresponding disconnection prescriptions are withdrawn. Ethical deliberation must use the person's interests and wishes, clinical evidence, uncertainty and the applicable decision process. The formal contribution is to expose assumptions and affected interests, not to substitute an unvalidated scalar for them.

Human–machine relationships​

A cognitive tool can become important to a person's agency and relationships. That provides reasons to examine disruption, data loss, manipulation, accessibility and informed choice, without first proving a composite conscious subject. Neither frequent dialogue nor a nonzero correlation certifies emergent subjectivity. Changing a tool affects a user through an empirically specified causal relation; it is not automatically equivalent to neurosurgery.

If future evidence supports moral patienthood for an artificial system, the protection model must be extended transparently. Until then, architecture, behaviour, subjective reports and phenomenal interpretation must retain their separate statuses. Capability rankings alone do not settle moral status.

8. Philosophical synthesis​

Self-maintenance can be compared with conatus; responsible reciprocity with universalisation; plural flourishing with eudaimonia; and error-aware practice with contemplative disciplines. These are interpretations [I], not proved equivalences or claims that the traditions are special cases of UHM. A relaxation term κ(ρ∗−Γ)\kappa(\rho_*-\Gamma) does not by itself capture an entire philosophy, and its purity contribution can have either sign.

The stronger project is to ask which insight can be given a precise contemporary formulation, what remains experiential or normative, and what evidence would support the bridge. Spiritual synthesis develops that comparison. Meaning of existence applies the resulting distinctions to a life trajectory.

What has been established​

The mathematical results here are limited and useful: the purity bounds and derivative, a conditional preservation certificate, positive-weight Pareto efficiency, the pinching identity, the phase counterexample, and a category of admissible paths. The ethical proposal supplies their purpose: protecting subjects, developing understanding and agency, caring across relations, and making one's commitments and revisions accountable.

The central question is therefore not “which scalar should grow at every instant?” but “which way of living can we justify, sustain, share and correct without erasing the subjects for whom it matters?”