Many-Worlds (Everett–Deutsch) and UHM

Who this chapter is for

David Deutsch (Oxford) is the strongest living advocate of Everett's many-worlds interpretation (MWI): The Fabric of Reality (1997), The Beginning of Infinity (2011), the decision-theoretic derivation of the Born rule (1999), and — with Chiara Marletto — constructor theory. This chapter states precisely where UHM agrees with, where it strictly diverges from, and what it contributes to the many-worlds programme. Every claim carries a status marker and a link to the corpus theorem it rests on. Nothing here is rhetorical.

1. The formal core of MWI (Deutsch's form)

To compare rigorously, fix MWI as Deutsch defends it:

#	MWI commitment	Formal content
M1	Universal state	One pure state $\lvert\Psi\rangle$ of everything; no external observer
M2	Dynamics	Only unitary evolution $U(t) = e^{-iHt}$; no collapse, ever
M3	Branching	Decoherence splits $\lvert\Psi\rangle$ into quasi-classical branches ("worlds")
M4	Preferred basis	Supplied by environment-induced einselection (Zurek)
M5	Probability	Born weights recovered from rational-agent decision theory (Deutsch 1999)
M6	Evidence claim	Quantum computation: "where was Shor's factorisation computed, if not across worlds?"

2. Point-by-point rigorous comparison

2.1 The ontic carrier: $\lvert\Psi\rangle$ vs $\Gamma$ — strict divergence [Т]

UHM's fundamental object is not a universal pure vector but a density operator $\Gamma \in \mathcal{D}(\mathbb{C}^7)$ on the ∞-topos site (Axiom 1). This is not a notational choice; it carries a theorem-level consequence for the very definability of worlds:

No-branching argument [Т, linear algebra]

A mixed state admits no canonical convex decomposition: for any non-extremal $\Gamma$ there are uncountably many ensembles $\{(p_k, \lvert\psi_k\rangle)\}$ with $\Gamma = \sum_k p_k \lvert\psi_k\rangle\langle\psi_k\rvert$, and no internal fact of $\Gamma$ selects one (classical theorem; the Schrödinger–HJW purification freedom). Therefore, if the fundamental object is $\Gamma$, "worlds" as components of the state are not invariantly definable. Branch-talk in UHM can only ever be basis-relative bookkeeping, never ontology.

Everett's move (M1) is only available if the fundamental state is pure. UHM's No-Zombie / viability structure requires mixedness ($P < 1$ for every viable holon, with the conscious window $P \in (2/7, 3/7]$ — T-124): a pure-state holon is non-viable. So the divergence at M1 is not aesthetic — UHM's living systems are constitutively mixed, and mixedness kills canonical branching.

2.2 Dynamics: unitarity-only vs the triad — strict divergence [Т]

MWI's M2 says the generator is $-i[H,\cdot]$ and nothing else. UHM proves (T-57, LGKS completeness) that the axioms force exactly three dynamical contributions:

$$\mathcal{L}_\Omega = \underbrace{-i[H,\cdot]}_{\text{Aut}} + \underbrace{\mathcal{D}_\Omega}_{\text{Fano dissipator}} + \underbrace{\mathcal{R}}_{\text{regeneration}}$$

with $\mathcal{D}_\Omega$'s decoherence rate an exact number, $\lambda_{\text{deco}} = 5\gamma/21$ (T-59, machine-verified 2026-08-05 by the 49×49 superoperator: spectrum exactly $\{0, -5\gamma/21\}$), and $\mathcal{R}$ the non-unitary, non-linear replacement flow toward $\rho_*$ (39f–h). In MWI decoherence is descriptive (a consequence of $U$ acting on system⊗environment); in UHM dissipation and regeneration are generative — part of the law itself. A holon does not branch; it metabolises coherence.

2.3 Preferred basis: UHM's contribution to the MWI programme [Т]

MWI's weakest joint (M4) is well known: einselection answers "which basis survives decoherence" by pointing at the interaction Hamiltonian — and leaves "why that interaction" open. UHM closes this joint structurally:

• the measurement basis is the atomic structure of the classifier $\Omega$ (Theorem 6.1) — logic, not environment;
• the frame itself is rigid up to $G_2 = \mathrm{Aut}(\mathbb{O})$ (T-42a, Stone–von-Neumann analogue);
• and the seven atoms are not anonymous: they are the derived dimension set $\{A,S,D,L,E,O,U\}$ with 7/7 functional uniqueness [Т].

Read against Deutsch: this is the exact structure MWI would need to earn its quasi-classical domains rather than borrow them from the environment. It is exportable: even a committed Everettian can take the $\Omega$-atomic answer to M4 as a standalone result.

2.4 The Born rule: two derivations, different premisses [Т]/[С]

Deutsch (1999) recovers Born weights from decision-theoretic axioms (rational preference over quantum games); critics (Barnum et al.) press circularity. UHM derives measurement probability from the Bures geometry forced by the axioms: the unique monotone metric characterisation (T-187) plus the classifier-projection reading of measurement (§3, Born rule from UHM). Statuses differ: Deutsch's route imports rationality axioms [П] in our vocabulary; UHM's route stays inside A1–A2 [Т] with the interpretive bridge confined to what "measurement" names [И].

2.5 Page–Wootters: the point of maximal kinship [Т]

Here the two programmes genuinely touch. Everett's original name was the relative-state formulation; Deutsch has always emphasised this. UHM's Axiom 5 — now a theorem (T-87: A5 derivable from A1–A4) — is precisely a relative-state construction: a global constrained state $\hat{C}\Gamma_{\text{total}} = 0$ whose conditional states relative to the clock sector are the experienced instants (T-38b, emergent time).

So UHM does host a rigorous plurality: the $\mathbb{Z}_7$ family of clock-conditioned states co-present in one timeless $\Gamma$. The strict difference from MWI: this plurality is temporal (seven phases of one holon), not modal (alternative outcomes), and its cardinality is fixed by the algebra (7), not exponentiating with every interaction. Call it what it is [И]: co-present moments, not parallel worlds.

2.6 "Where was Shor computed?" — UHM's answer [Т]/[И]

Deutsch's M6 argument takes exponential quantum speed-up as direct evidence of parallel worlds. UHM answers without multiplying ontology: the computation lives in the off-diagonal sector of one $\Gamma$ — the same coherences whose Frobenius weight defines $\Phi$ (integration measure) and whose cooperative surplus is exactly $P(\rho^{(12)}_*) - P(\rho_{\text{diag}}) = 2\lVert\gamma_{\text{cross}}\rVert_F^2$ (T-77). Amplitude paths interfering within one state are resources, not places. The quantitative UHM-native measure of "how much parallel room" a configuration has is $\mathrm{Freedom}(\Gamma) = \dim\ker(\mathcal{H}_\Gamma) + 1$ (finite-dimensional theorem [Т]) — flat directions of one landscape, bounded by 7, never $2^n$ worlds.

2.7 The falsifiable fork [П]

The comparison yields one experiment-class divergence, stated honestly:

Fork (macro-interference of living matter)

MWI (pure unitarity): no principled ceiling on interference visibility for arbitrarily complex systems — with sufficient isolation, a virus, a cell, a holon interferes.

UHM: any viable system carries the internal Fano dissipator with rate $\lambda_{\text{deco}} = 5\gamma/21 > 0$ [T-59] by law, not by environment. Perfect isolation does not switch it off (it is part of $\mathcal{L}_\Omega$, [T-57]). Hence a hard ceiling: interference visibility of a living (viable, $P > 2/7$, $\mathcal{D}_\Omega \neq 0$) system decays at least at $\lambda_{\text{deco}}$ regardless of shielding.

Discriminator: matter-wave interferometry approaching the living threshold (large biomolecules → viruses → viable cells). Persistent full-visibility interference of a demonstrably viable system would falsify the UHM triad; a shielding-independent visibility floor matching $5\gamma/21$ scaling would falsify unitarity-only MWI. Status: research programme [П] — the numbers ($\gamma$ per system) need the $\omega_0$-calibration layer.

2.8 The Fabric's four strands — structural concordance [И]

Deutsch's Fabric of Reality weaves four strands and claims they form one explanatory fabric. UHM, built independently, arrives at a four-strand structure with exact counterparts — evidence that the fabric intuition tracks something real:

Deutsch's strand	UHM counterpart	Where
Quantum theory (Everett)	$\Gamma$, Bures topos, triad dynamics	A1–A2, T-57
Epistemology (Popper)	Reflection $R$, self-model $\varphi$, falsifiability registry	Self-observation, Falsifiability
Computation (Turing)	Classification of computations; $\Phi$-classes	Consequences §Computation
Evolution (Darwin)	Autopoiesis (AP), viability V, regeneration $\mathcal{R}$	Septicity

And constructor theory's counterfactual primitives ("possible / impossible transformations") have a working UHM analogue: the threshold catalogue ($P_{\text{crit}}$, $R_{\text{th}}$, $\Phi_{\text{th}}$, the conscious window) is a catalogue of possible-versus-impossible configurations, maintained with registry discipline.

2.9 The learning-theoretic bridge: UHM fixes Vanchurin's free function [Т]

Deutsch's multiverse and Vanchurin's self-learning universe are the two live programmes that derive physics from something more primitive (branching worlds; learning dynamics). UHM engages the second one quantitatively, and the result is a set of exact theorems (T-293–295; the full derivation, with proofs, the falsification protocol and the reproducibility table, is The Learning Algorithm of a Holon):

Vanchurin parametrises the learning algorithm by the functional relation $g=\kappa^{a}$ between metric and noise covariance (Geometric framework for biological evolution, arXiv:2603.15198, Eq. 6.8), with $a=0$ (stochastic gradient), $a=\tfrac12$ (efficient learning — Adam-like, conjectured to underlie biological complexity) and $a=1$ (natural gradient, the "quantum regime"), and states that direct estimates of $\kappa$ remain unavailable. In UHM neither factor is free: $g$ is Bures (unique monotone metric, T-187) and $\kappa$ is the Kraus-increment covariance of the canonical dissipator (T-41/T-59). Computing both gives

$$\kappa^{\uparrow\uparrow}_{\text{at}}=\frac{\gamma}{4N}\,g^{-1} \qquad\Longleftrightarrow\qquad \Pi\,g\,\kappa_{\text{at}}\,\Pi=\frac{\gamma}{4N}\,\Pi ,$$

which is verbatim his own criterion for $a=1$: substituting $a=1$ into his Eq. 7.5 yields exactly $g^{-1}=g^{-1}\kappa g^{-1}$. Hence $a=1$ exactly. (The constant is $\gamma/4N$ because the Bures metric is one quarter of Fisher–Rao on commuting perturbations, not one half; in the SLD-QFI normalisation used in the natural-gradient literature the identity reads $\gamma\Pi/N$. The exponent and all ratios below are independent of that choice.)

Three companions, all state-independent. The universal Fano factor $\operatorname{Tr}(g\kappa_{\text{full}})/\operatorname{Tr}(g\kappa_{\text{at}})=11/9$ at matched per-channel rate — equivalently, the block layer carries exactly $2/11$ of the noise. The noise–purity law $\operatorname{Tr}\kappa=\frac{\gamma}{N}(1-P)$, so the conscious window becomes a learning-noise band $[4\gamma/49,5\gamma/49)$. And a sector split: $\kappa^{\text{coh}}=0$ on the decohered manifold, where coherences instead contract deterministically at $5\gamma/21$ — stochastic learning noise does not reach them. Three consequences deserve emphasis:

1. What is a modelling choice for him is a theorem here. The relation $g^{-1}\propto\kappa^{\uparrow\uparrow}$ is his $a=1$ condition, one option among three; UHM derives it. It is the dynamical counterpart of the static identity $g^{-1}=c$ that his §4 obtains from a maximum-entropy argument, where $c$ is the population covariance. Granting that static identity too, the two together predict $\kappa^{\uparrow\uparrow}=\frac{\gamma}{4N}c$ — the covariance of temporal changes has the same shape as the static covariance, differing by a single scalar that is a pure rate. One side is routinely measured; the other, he notes, never has been.
2. A falsifiable disagreement: UHM predicts that a viable learner sits at $a=1$ with Fano-induced anisotropy, not at the $a=\tfrac12$ regime conjectured for biological complexity.
3. The test needs no metric. His Eq. 7.5 degenerates at $a=\tfrac12$, where $\kappa^{\uparrow\uparrow}=I$ carries no information about $g$ at all. So the two hypotheses make opposite shape claims about one measured object: $a=\tfrac12$ makes $\kappa^{\uparrow\uparrow}$ spherical, while $a=1$ makes it proportional to $\operatorname{diag}\lambda-\lambda\lambda^{\top}$, computable from the state alone. A sphericity test on about a hundred aggregated time windows separates them. Its resolving power vanishes at the maximally mixed state and grows away from it — so the theory predicts that the systems it calls holons, sitting in the window $P\in(2/7,3/7]$, are exactly the systems in which its own central claim is measurable.

Read together with §2.6: Deutsch asks where the computation happens and answers "in other worlds"; Vanchurin asks what algorithm the world runs; UHM answers the second question with numbers and, in doing so, removes the need for the first — the computation happens in the coherence sector of one state, and the learning happens in its complement.

3. Summary verdict

What UHM strictly says about parallel universes:

1. No ontological branching [Т-level argument]: the fundamental state is mixed (viability forces $P < 1$), mixed states have no canonical decomposition, and the law itself is the non-unitary triad — three independent blockers, any one of which suffices. The topos is one; there is no outside for worlds to sit in (T-55/T-56, No Outside).
2. Maximal kinship at Page–Wootters [Т]: UHM is a relative-state theory of time — seven co-present clock-conditioned states of one $\Gamma$; Everett's machinery, capped cardinality, temporal not modal.
3. A genuine contribution to MWI's own programme [Т]: the $\Omega$-atomic, $G_2$-rigid preferred basis — the missing answer to M4 — stands independently of whether one accepts branching.
4. The quantum-computation argument is answered without worlds [Т]/[И]: coherences of one state are the resource; Freedom counts the flat directions and is bounded by 7.
5. One honest experimental fork [П]: a law-level decoherence floor for living matter versus unitarity's unbounded interference — a real, eventually decidable disagreement.

Related pages: Measurement · QM reduction · Emergent time · Consciousness theories compared