QBism

The Program in Brief

QBism begins from an observation hard to argue with once stated plainly: every time a physicist writes down a quantum state, someone in particular is writing it down, on the basis of what they in particular know. The program takes that seriously all the way down. A quantum state is not a property of a system; it is an agent’s personal probability assignment — a degree of belief in the strict de Finetti sense, subject to the ordinary coherence requirements of Bayesian probability and revisable in the ordinary Bayesian way.

From that single commitment, the traditional puzzles lose their grip. Collapse is belief updating, not a physical process, because there was never a physical object to collapse. The Bell correlations are correlations among one agent’s expectations for outcomes at two places, not a mechanism reaching across space.

What keeps QBism from being merely deflationary is where it looks for quantum theory’s content. If the state is a belief, the Born rule cannot be a law of nature — it must be a norm, an addition to the calculus of probability telling a coherent agent how their gambles on one experiment must relate to their gambles on another. Fuchs and Schack sharpen this using symmetric informationally-complete measurements: relative to a SIC reference, the Born rule becomes a deformation of the law of total probability,

$q(j) = \sum_{i=1}^{d^2}\left[(d+1)\,p(i) - \tfrac{1}{d}\right] r(j|i)$

and the empirical content of quantum theory is compressed into that deformation and the integer $d$. The research program is then a genuinely mathematical one: find the principle that singles out this equation.

The charge of solipsism is the one QBists have answered most carefully, and it should not be leveled casually. Nothing in QBism denies that the world exists or that it is shared; what it denies is that any agent’s account can be completed into a view from nowhere. Fuchs calls the position participatory realism: the agent’s actions are real interventions in a world that pushes back. The theory is single-user; the world is not.

The Status of the Quantum State

This is where the two agree most, and where they are most often confused. The framework also holds that there is no observer-free state: a description is always relative to some observer’s relational invariants. Measurement makes the consequence explicit — after $\mathcal{O}$ measures $S$, a third party $\mathcal{P}$ who has not interacted with $S$ still describes it in superposition, and both descriptions are simultaneously correct. Neither is the “real” one.

But the framework refuses the next step. In Observer-Relative Objectivity the blunt subjective/objective binary is replaced by a trichotomy, and the interesting category is the middle one. Level 2 facts are genuinely observer-dependent and genuinely constrained: which observer holds the fact varies, what fact can be held does not. The non-fabrication theorem supplies the teeth — given an observer’s relational invariants with $S$, the state $|\psi_S\rangle_\mathcal{O}$ is determined and the probabilities follow with no freedom whatever. An observer cannot revise its assignment; it never had one to revise.

Flanking Level 2 are two categories QBism does not obviously have room for. Level 1 holds facts identical for every observer in a coherence-connected network: total coherence $C_0$, the conservation laws, and whether two observers share a relational invariant at all. Level 3 holds propositions no observer can assign a truth value to — the framework’s version of the QBist insistence that unperformed measurements have no outcomes. On Level 3 the two are close allies; on Level 1 they part.

Who Counts as an Agent

QBism is deliberately reticent here, and the reticence is principled rather than evasive: the agent is where the theory gets used, not what the theory describes, so a criterion for agenthood is not quantum theory’s business. In practice the agent is characterized by capacities — taking actions, placing bets, experiencing consequences. QBists do not require the agent to be human, but the vocabulary is unmistakably the vocabulary of deliberation.

The framework cannot afford that reticence, and does not want it. An observer is the triple $(\Sigma, I, \mathcal{B})$ of Observer Definition: a state space, a conserved invariant constituting its identity, and a partition of transformations into those that preserve that identity and those that threaten it, subject to non-triviality conditions ensuring the symmetry group is proper and the invariant carries information. Nothing there mentions knowledge, belief, decision, or awareness. A fundamental particle satisfies it with its internal phase space, its charge and spin quantum numbers, and its gauge symmetry; so does a cell, with a biochemical state space and a membrane. The derivation says as much in its own caveat: the formal object is closer to a persistent self-distinguishing system than to anything colloquial, and calling it an observer is a choice about emphasis.

This is not a refutation — QBism owes no such definition for its own purposes. It is a statement of what each program is buying. The framework needs observers to be structural objects because everything downstream depends on it: the three interaction types, the reverse Noether mechanism, and the bootstrap to particle content are theorems about $(\Sigma, I, \mathcal{B})$, not about anyone’s credences. A framework whose agents must be able to gamble cannot derive the spin-statistics theorem, and does not try to.

Where the Born Rule Comes From

Both programs refuse to accept $|\psi|^2$ as brute, and the shared refusal is more of an alliance than the slogans suggest: the SIC program is a search for a uniqueness statement, and so is Born Rule from Coherence Conservation. The difference is what does the uniqueness work.

QBism’s constraints are normative — an agent’s gambles must be Dutch-book coherent, and the Born rule is the additional norm relating gambles on a reference measurement to gambles on any other. The framework’s are conservation-theoretic. Normalization comes from coherence conservation, phase covariance from the $U(1)$ loop structure, composition from the interaction graph; the three force $f(x) = x$ and nothing else, checked against Aczél’s functional-equation theorem and confirmed independently by Gleason for $d \geq 3$. The derivation then goes further than it needs to: the coherence functional itself is pinned by a Cauchy multiplicative equation to $\mathcal{C}(|\psi\rangle) = \langle\psi|\psi\rangle$, and the Hilbert space structure is reconstructed rather than assumed. The interpretive payload is stated without hedging: the Born probability is the fraction of total coherence flowing through the $k$-th outcome channel, “a statement about coherence distribution, not about subjective ignorance.”

The same split shows up in information geometry. The framework’s Fisher metric derivation runs Čencov’s uniqueness theorem and fixes the scale constant at $\hbar$, so that action is $\hbar$ times information-geometric arc length. A Bayesian reads Fisher information as a fact about an inferential agent’s power to discriminate hypotheses; the framework reads identical mathematics as the metric coherence conservation induces on state space. Same equations, different ontological weight — which is the whole comparison in miniature.

Is Physics About Experience?

QBism’s boldest claim is that a measurement outcome is an experience, and that quantum theory is a manual for an agent acting in a world that will not sit still to be described. The world is unfinished; each measurement adds to it.

The framework agrees with more of this than one might expect. Measurement is a real event, not the passive reading of a pre-existing fact — no universal definiteness rules out any single assignment of values valid for all observers, and entanglement monogamy is why. Its world is emphatically unfinished too: every Type III interaction generates a genuinely new conserved quantity on the joint state space, which by the reverse Noether mechanism generates a new symmetry, which licenses a new degree of freedom, which enables further interaction. Structure accumulates rather than being laid down in advance — close enough to the QBist picture of a world still in creation that the resemblance deserves acknowledging.

The divergence is over where the creativity lives. In QBism it lives in the agent’s act and registers as the agent’s experience. In the framework it lives in the interaction, and what gets created is a relational invariant, irreducible to either party — created identically whether the two parties are a physicist and a photomultiplier or two protons. Experience is never invoked, and there is no place in the formalism where it could be.

The Pattern: Indexed, Not Owned

Both frameworks say the quantum state carries an index — you cannot write $|\psi\rangle$ without saying whose. The whole disagreement is over whether indexing implies ownership. QBism says the state is yours: your belief, your gamble, yours to revise, answerable only to coherence. The framework says the state is indexed to you and owned by no one: it is fixed by your relational invariants and the conservation structure, and you have no more latitude in it than you have in the value of a conserved charge.

That is why the framework can be observer-relative without being subjectivist, and Wigner’s friend shows the difference in method. QBism dissolves it by making the two accounts two agents’ personal accounts, never meant to be reconciled. The framework dissolves it by removing the external clock: time is phase ordering on the dependency DAG, so there is no observer-independent “now” at which both descriptions must be jointly evaluated — and when Wigner does interact with the friend, strong subadditivity forces the accounts to agree. Consistency is not diplomacy between agents; it is a constraint.

Where They Genuinely Part Ways

Three differences are real rather than verbal. The first is modality: for QBism the Born rule is what a coherent agent ought to obey, and an incoherent agent is irrational rather than impossible; for the framework it is what coherence conservation permits, with no incoherent-observer option to be irrational in. The second is Level 1. The framework insists on agent-independent facts — total coherence, conservation laws, the topology of the interaction graph — where QBism is reluctant to grant that quantum theory delivers any facts about the world rather than norms for acting in it.

The third is scope and risk. QBism is an interpretation and changes no predictions, a design virtue that also means almost nothing can bear against it. The framework treats quantum mechanics as one layer and keeps going into spacetime, gauge structure, and particle content, which means it can be wrong in public. Both predict exact unitarity with no objective collapse — a test that separates them from much of the field but not from each other.

What the Framework Takes From It

QBism’s lasting service was to make “whose state is it?” an unavoidable question and to show how much of the apparent paradox evaporates once it is asked. The framework’s Level 2 category is a debt to that, as is the discipline of never writing a state without its observer index. The SIC program deserves credit as the closest thing in the interpretive literature to what the Born rule derivation attempts: an honest search for the constraint that makes $|\psi|^2$ the only option. Whether the framework’s three constraints imply the SIC deformation of total probability is an open and rather attractive question; nothing here settles it.

There is one place the framework should not claim the advantage. On the single-outcome question — why this result, this time — QBism’s answer is clean: it was the agent’s experience, and there is nothing further to say. The framework’s answer, that the outcome is fixed by the full coherence geometry of the interaction but is structurally inaccessible from inside a bounded observer, is flagged in its own derivation as an interpretive position rather than a theorem. It is a more ambitious answer, but not yet a better-supported one, and the comparison is fairer for saying so.