Agency in the Emergent Multiverse

Where the laws meet the branches

Force the coin. By the accounting of the kybit, fixing a fair coin’s outcome costs exactly one kybit of control, and physics bills you for it: a floor of \(k_B T \ln 2\) in dissipated work, paid through whatever machinery — fingers, solenoids, neurons — does the forcing. That was Part I’s result, condensed into three laws: agency is real, agency is costly, agency is limited.

Now run the same intervention in the Quantum Branching Universe (QBU) — the branching ontology mapped in its own chapter. Somewhere in the branchcone ahead of you, the coin lands tails anyway. Not as a possibility you failed to prevent — as an actual outcome, witnessed by an actual version of you, in a branch exactly as real as this one. Every future you paid to steer away from still happens somewhere. So what did the joules buy?

This is the point where the volume’s two halves — the thermodynamics of control and the physics of branching — either collide or weld. I am going to argue that they weld, and that the cleanest way to see it is through the most rigorous statement of the branching picture on offer: David Wallace’s The Emergent Multiverse. Wallace presents the Many-Worlds Interpretation as a coherent emergent ontology — branches objectively split through decoherence, worlds are patterns rather than fundamental furniture — and, crucially, he frames decision-making as intrinsically quantum, giving rational choice a naturalistic foundation inside the physics rather than on top of it. Set his machinery beside the three laws and the fit is not loose analogy. It is law-by-law alignment.

Steering Is Quantum Work

Start with the first law: exercising intentional control requires physical work proportional to the kybits exerted.

In Wallace’s picture, a decision-point is a branching-point. When an agent chooses, it does not select which outcomes exist — decoherence has already seen to it that the alternatives all occur — it shapes the quantum Measure of the branches that follow. Choosing, in the QBU, just is exerting control over branch weights: influencing quantum states and their amplitudes is physical work, performed by a physical system, paid for in the currency the first law names. The kybit’s definitional machinery carries over without modification. The \(P_{\text{initial}}\) of the kybit’s definition — the distribution drift would deliver — becomes a Measure distribution over the branchcone forward of the agent’s Vantage; the \(P_{\text{final}}\) is the distribution the agent’s intervention achieves; the kybits exerted are the divergence between them. The unit of control was always, without knowing it, a unit of Measure redistribution.

Wallace’s quantum decision theory says what the rational version of this steering looks like: maximize Measure-weighted expected utility. And the Deutsch–Wallace theorem says that a rational agent’s subjective probabilities must align with the objective Measure — the decision-theoretic route to the Born rule, whose epistemological twin, Credence tracking Measure, is developed in Probability Without Collapse. Read through the first law, the theorem has a thermodynamic moral: rational agency is not arbitrary work but predictive work. An agent whose Credences track Measure minimizes predictive error, and an agent that minimizes predictive error wastes the fewest joules per kybit. Effective quantum agency is the art of buying Measure where it is cheapest.

Decoherence Underwrites Decay

The second law: in a closed system without external energy input, the capacity to exercise agency inevitably decreases.

Here the alignment runs deeper than compatibility — the two frameworks share a mechanism. Decoherence, the process at the center of Wallace’s emergent multiverse, is thermodynamically irreversible: it increases entropy, and that irreversibility is precisely what makes quantum outcomes effectively classical and the branches objectively distinct. The engine that manufactures worlds and the clock that runs agency down are the same engine. Every branching event exports entropy; every exported bit of entropy is free energy the agent no longer has; and free energy, by the first law’s exchange rate, is exactly what control capacity is purchased from.

So the emergent multiverse does not merely permit agency decay — it enforces it. Cut an agent off from fresh energy gradients and its steering does not degrade because of some separate biological accident; it degrades because the very process that generates the branching landscape is entropic, and the agent is embedded in it. Drift, translated into QBU vocabulary, is what happens at every branch-point where no one is paying: Measure spreads across outcomes according to the bare physics, thinly and indifferently. An unsteered future is not an empty future. It is a future whose weights nobody bothered to bias.

No Frictionless Branches

The third law: perfect, frictionless control is physically impossible.

Quantum mechanics does not just tolerate this limit; it supplies the enforcement at the most fundamental level available. There is no such thing as a frictionless quantum manipulation. Infinite precision is unavailable in principle; every interaction that steers an outcome also entangles, decoheres, and dissipates; every act of quantum choice inherently involves uncertainty and thermodynamic loss. You cannot pick a branch the way you would pick a card from a deck — cleanly, exhaustively, at no cost. What an agent can do is amplify preferred branches at a metered price, and the meter never reads zero.

This is worth pausing on, because it kills a fantasy that branching universes tend to invite: the fantasy of the agent as branch-selector, surfing the multiverse, choosing worlds like channels. Nothing in the physics supports it. The third law’s asymptote — control approaching completeness only as cost grows without bound — is native to the quantum substrate, not imposed on it.

Meaning Is Found in Measure

Now return the opening question to the table. If all possible outcomes occur, does your agency matter?

Yes. Absolutely. Not because you select one exclusive future — you don’t, and no honest reading of the physics says you do — but because your actions influence the Measure of outcomes across branches. You don’t eliminate futures; you guide how prominently “you” inhabit them.

Each branch carries its Measure — a weight, a thickness, fixed by the squared amplitude and defined precisely in Measure, Vantage, Branchcone. A higher-Measure branch is not more real than a lower-Measure one, but it counts for more: more weight behind its observers, more of what your expectations and your evaluations ought to track. Agency is about the distribution of that weight. Intentional, model-driven action concentrates Measure on preferred outcomes; passive drift and reflex spread it thinly and randomly. When you build predictive models, anticipate outcomes, and act to favor some of them, you are engaging environmental variables in ways that shape branching Measure — redistributing weight among possibilities rather than deleting any of them.

In a deterministic multiverse where every outcome exists, meaning is found in Measure. You don’t control what exists; you control which outcomes flourish — which versions of you become prominent, how much of the weight of the world flows through futures you endorse. That is what the first law’s energy expenditure purchases, and it is why the expenditure is an investment rather than a ritual. It is also the load-bearing wall under everything this framework builds downstream: if agency directs Measure, then agents are answerable for where they direct it, and measure responsibility has physical content rather than metaphorical decoration.

Sharing a World

Everything so far treats the agent alone against the branching landscape. But steering is rarely solitary. We persuade, coordinate, promise, compete — and all of that presupposes something the multiverse makes suddenly non-trivial: that the agent you are dealing with is in your world. The multiverse contains observers beyond counting. What makes a particular one share yours?

Observers in the QBU are not abstract Cartesian minds; they are physical patterns — instantiated agents embedded in decohered branches, who join branches rather than create them. Model each observer as a Strong Pattern Identifier (PI): a reproducible, high-fidelity structure — a neural connectome, a cognitive-functional profile — persisting across timelines that share a common ancestor:

\[ \text{PI}_{\text{observer}} = \{\, P : P \text{ matches the defining pattern of a given observer across branches} \,\} \]

These PIs pick out the set of observers with enough structural continuity to count as “the same” observer across nearby branches.

Observers do not enjoy arbitrary access to all possible bases. They operate within decohered frames that select preferred ones — spatial location, object permanence, classical causality. Two observers A and B share a basis at a Vantage \(V\) if and only if they descend from a common ancestor PI within the branchcone of \(V\); they have not decohered into orthogonal states, so they remain mutually observable; and their internal measurement and interpretive models are functionally compatible. That defines the Observer Class at \(V\):

\[ \mathcal{O}(V) = \{\, \text{PI}_i : \text{PI}_i \text{ is coherent and communicative from } V \text{ forward} \,\} \]

— the set of observers sharing an effective basis for communication, interaction, and perception. And sharing comes in degrees. Define Observer Class Alignment (OCA):

\[ \text{OCA}(\text{PI}_A, \text{PI}_B, V) = \text{degree of effective basis overlap at vantage } V \]

quantifiable in terms of mutual predictability of measurement outcomes, shared semantic encoding (language, perception, object identity), a common entanglement environment — the ambient decoherence field — and functional synchrony of memory and causal models. High OCA is what shared experience is made of: a common direction of time read off shared entropy gradients, compatible decision models within which value and agency register, and channels over which communication can actually carry information.

The implications reach further than the formalism suggests. Only high-OCA observers form mutually relevant agent classes — the natural population over which the coordination, ethics, and rivalry of agency play out. Ethical and epistemic structures are therefore frame-local, not universal: they bind within an observer class because that is where mutual observability lives, a locality that Volume 5’s agent binding will make normative. Even the perception of time reflects incomplete alignment — the residue of not being in a perfectly co-moving basis with your world. And the boundary is real: outside your OCA-defined class, other observers exist in the multiverse in full ontological standing, but they do not share your world. No common basis, no communication, no shared stakes. Observation is a relational, basis-bound phenomenon rooted in shared branchcones — and “the world,” for any agent, is exactly as big as its observer class.

The Honest Ledger

The objections chapter promised that this chapter would keep the ledger on the framework’s known gap. Here it is, stated without cushioning.

The claim that intentional kybit expenditure shifts Measure toward preferred branches is asserted by this framework, not derived within it. The thermodynamics tells me what steering costs. The QBU tells me what the arena is. The bridge between them — that the paid work actually redistributes branch weight relative to what drift would have delivered — is a commitment of the framework, the same one flagged in the three laws’ objections, and I do not currently know how to prove it. In the spirit of the ledgers Volume 2 keeps on its own open problems, here is what a genuine derivation would need to supply:

• A defined baseline. “Shifts Measure” is a comparison, and the comparison class is a counterfactual: the Measure distribution over the branchcone forward of the agent’s Vantage had the agent’s predictive machinery been inert — the kybit’s \(P_{\text{initial}}\), restated as a fact about branch weights. That counterfactual must be constructed with the branch-based semantics of Causality and Counterfactuals and shown to be well-defined in a branching ontology, not smuggled in from single-world intuitions.

• A dynamical bound. From decoherence dynamics, a demonstration that a dissipative, predictive process changes conditional branch weights relative to that baseline — and by how much. The natural target is a fluctuation-theorem-style result bounding the divergence between achieved and baseline Measure distributions by the work dissipated. That would promote Landauer’s floor from analogy to exchange rate: so many joules, at minimum, per kybit of Measure actually moved.

• A no-conjuring proof. Unitarity conserves total Measure from any Vantage. Whatever “shifting” turns out to be, it must be redistribution by correlation — the agent making itself the kind of physical structure whose actions co-occur, branch by branch, with preferred outcomes — never the creation of weight. The derivation must show that this correlational sense of shift is strong enough to underwrite the counterfactuals the framework asserts, and no stronger than unitarity permits.

• A triviality check. The hardest entry. In a deterministic wavefunction, every Measure distribution is simply what unitary evolution delivers, agents included. A derivation must locate a principled difference between an agent shifting Measure and Measure merely flowing through an agent-shaped channel — or else concede that the difference is perspectival and rebuild the claim rigorously from within the agent’s Vantage. Until then, the objection that steering is bookkeeping rather than achievement remains live, and I list it here as live.

These are honest debts, entered in the ledger rather than swept under the formalism. Everything in this chapter that is welded — steering as control work, decoherence as the engine of decay, the impossibility of frictionless branch-selection, the OCA account of shared worlds — stands independently of them. What the debts cover is the bridge that makes the weld matter: the claim that when you spend the joules, the weights move. I believe it. I have argued it is the only reading on which agency in a branching universe is worth the name. But I will not pretend the weld is a theorem.

The three laws survive translation into the emergent multiverse not just intact but sharpened: the multiverse is a quantum-thermodynamic landscape, and agents navigate it by optimizing predictive outcomes under fundamental physical constraint. What the translation has not yet delivered is the view from inside — what it is like, and what it can possibly mean, to choose when the demon bookkeeping your future sees not one trajectory but a branching structure of realized outcomes. That is where Part III begins, with Everett’s Demon.