Relational Quantum Mechanics

The Program in Brief

Relational Quantum Mechanics begins from a diagnosis: the measurement problem is an artifact of an unexamined assumption. Standard quantum mechanics quietly assumes that there is a single, absolute, observer-independent set of facts about what happened — and then discovers that the formalism will not supply one. Rovelli’s move, in the 1996 paper and everything since, is to drop the assumption rather than patch the formalism.

In RQM, a quantum state is not a property of a system. It is a property of the relation between a system and another system. When an electron interacts with a detector, the electron acquires a definite value relative to the detector, and to nothing else. A third system that has not interacted with either still correctly describes the detector-plus-electron as entangled. Both descriptions are complete and both are correct, because there is no third description — no view from nowhere — against which they could be judged.

Crucially, RQM refuses to privilege any system. There is no classical/quantum cut, no threshold of complexity, no role for consciousness. Any physical system can play the role Copenhagen reserved for the apparatus. The theory is democratic about observers precisely because “observer” is not a special kind of thing but a role any system occupies in any interaction. Rovelli also proposed that quantum mechanics itself could be reconstructed from informational postulates — that the information extractable from a system is finite, and that new information is always obtainable — an early and influential instance of the reconstruction program.

RQM is also deliberately modest about scope. It is an interpretation of quantum mechanics: it takes the Hilbert space formalism, the unitary dynamics, and the Born rule as given, and explains what they are about. It does not try to derive them, and it makes no attempt on the particle spectrum or on gravity. That modesty is a feature, not an oversight — Rovelli’s claim is that the conceptual problem is solved by dropping absolute facts, and that nothing further is owed.

Of every framework compared on this site, RQM is the closest kin. The framework’s own derivation pages say so explicitly: measurement is “structurally closest to Rovelli’s relational quantum mechanics”. Where the two differ is not mostly in what they say about quantum facts. It is in what they think relationality is for.

What Counts as an Observer

RQM’s answer is: everything, equally. Any system can serve as the reference for another system’s state, and no system is intrinsically an observer. This is a liberating move and the framework fully endorses its spirit — the observer definition is explicitly functional, applies to an electron and a galaxy alike, and requires nothing about awareness.

But the framework does not stop at “any system.” It gives the role a structure. An observer is a triple $(\Sigma, I, \mathcal{B})$: a state space, a conserved invariant that constitutes its identity through interactions, and a self/non-self boundary partitioning transformations into those that preserve that identity and those that threaten it. Non-triviality conditions rule out the degenerate cases — the symmetry group cannot be trivial, threats must exist, the invariant must carry information. Observers so defined form a category, so they can be composed, compared, and classified.

This is a real addition, and also a real commitment. RQM’s “any system” is maximally permissive and therefore cannot be wrong; the framework’s triple can be wrong, because it says which systems count and how they compose. The payoff is that downstream results — the three interaction types, the requirement that at least three observers exist (Multiplicity), the loop closure that gives every observer a period and a phase — all reference that structure. RQM has no such downstream, because it was not built to have one.

There is a fair objection running the other way. RQM’s permissiveness is not laziness; it is the position that relations are all there is, with no self-standing relata underneath. The framework’s triple sits uncomfortably close to a relatum: an observer has an intrinsic state space and an intrinsic invariant before any relation obtains. A strict relationalist can reasonably say the framework has smuggled substance back in.

Facts Without a View From Nowhere

Both frameworks say that after a measurement, the friend has a definite outcome and Wigner does not, and that this is not a contradiction. The framework’s measurement derivation reconstructs this carefully: measurement is a Type III interaction generating a new relational invariant $I_{\mathcal{O}S}$ on the joint state space; “collapse” is the creation of that invariant, unitary on the joint space and non-unitary only in the conditional description relative to one party. Wigner, holding no invariant with the system, correctly describes the lab as entangled.

The interesting divergence is over what guarantees that observers agree when they finally compare notes. This is a live question inside RQM itself — the concern that purely relative facts risk a kind of solipsism has led to the addition of a cross-perspective-links postulate, ensuring that stable records held by different systems do not disagree. That RQM found it necessary to add such a postulate is worth taking seriously; the framework’s position is that this consistency should not need postulating.

In Observer-Relative Objectivity, agreement is derived. Strong subadditivity of the coherence measure — property C5 of Coherence Conservation — gives $I(A{:}S|B) \geq 0$, which forbids two observers sharing coherence from holding independently arbitrary descriptions of a third system. Wigner’s outcome must correlate with the friend’s, because the existing invariant $I_{\mathcal{F}S}$ constrains which eigenvalues the new invariant $I_{\mathcal{W}\mathcal{F}}$ can take. The framework adds a second, sharper dissolution: the Wigner’s friend puzzle only bites if one assumes an external time parameter at which both descriptions must be jointly evaluated. Since time is phase ordering on a dependency DAG, there is no observer-independent “now,” and no slice on which the two descriptions collide before the observers actually interact.

The result is a three-level classification that RQM does not draw: facts that are observer-invariant (conservation laws, network topology), facts that are observer-relative but uniquely constrained, and propositions genuinely undefined for every observer in the network. The sheaf formulation makes this precise — coherence and probability data glue to unique global sections, outcome data does not, and the obstruction is Kochen–Specker contextuality.

Where the Born Rule Comes From

RQM takes the Born rule as part of the quantum formalism it interprets. Rovelli’s reconstruction program aimed to recover the formalism from informational postulates, but the postulates were about information capacity, chosen for their plausibility.

The framework’s Born rule derivation runs from three constraints it argues are forced: normalization from coherence conservation, phase covariance from the $U(1)$ loop structure, and composition from the interaction graph. These uniquely force $P(k) = |\psi_k|^2$ — no other function survives — with Gleason’s theorem confirming from an independent direction, and with the Hilbert space structure itself derived rather than assumed. This matters for the relational picture specifically: it is what makes observer-relative facts non-subjective. No observer can fabricate an outcome, because given its relational invariants the probabilities admit no freedom at all.

The Pattern: Relationality as Premise, Relationality as Consequence

Everything above is one difference wearing different clothes. RQM treats relationality as the premise and quantum mechanics as the thing being interpreted. The framework treats relationality as a consequence — of what it takes for anything to persist while distinguishing itself from its surroundings — and treats quantum mechanics as one of the things that consequence should produce, alongside dynamics, a preferred basis, and eventually a particle spectrum.

This is a difference in ambition, and ambition cuts both ways. RQM’s conclusions are safe because it claims less: it cannot be refuted by a failure in a derivation chain it never attempted. The framework’s derivation chain is long, and several of its quantum-layer results — measurement, observer-relative objectivity, the sheaf formulation — remain provisional because they depend on a structural postulate about the interaction-Hamiltonian origin of the preferred basis. Length is exposure.

Where They Genuinely Part Ways

Three divergences are real rather than terminological.

Whether relational facts are constrained or merely relative. RQM has no internal mechanism forcing observers’ descriptions to cohere; that is what the cross-perspective postulate supplies. The framework claims coherence conservation supplies it as a theorem. If the framework’s strong-subadditivity argument does not carry the weight placed on it, this advantage evaporates.

Whether there is any absolute structure. The framework’s total coherence $C_0$ is observer-invariant, conserved on every Cauchy slice, and the same for all observers. That is an absolute quantity in a theory otherwise about relations, and a consistent relationalist should object to it. The framework’s answer is that “objective” means “constrained by structure no observer can violate,” not “observer-independent” — but the absolute conservation law is doing real work, and it is a background element RQM does not carry.

Whether interpretation is the end of the job. RQM holds that once absolute facts are dropped, the conceptual problem is closed and physics resumes as normal. The framework holds that the same structure which dissolves the measurement problem should also generate the dynamics — and stakes falsifiable claims on it, including exact unitarity at all scales with no objective collapse and no nonlinear corrections. RQM is compatible with that prediction but does not make it.

What the Framework Takes From It

Rovelli got there first, and got the hardest part right. The insight that the measurement problem is generated by an assumption rather than by the formalism — and that dropping “absolute facts” costs nothing empirical — is RQM’s, dated 1996, and the framework’s account of measurement is a reconstruction of a position Rovelli had already staked out. On the core interpretive question the framework is not an alternative to RQM; it is an attempt to give RQM a foundation and a downstream.

RQM is also cleaner in ways worth naming. It achieves its result with no new mathematical machinery, inherits quantum mechanics’ entire empirical record without risk, and is more philosophically careful about what relationality commits you to — in particular about whether relations require relata, a question the framework’s observer triple answers by assumption rather than by argument. And RQM’s willingness to stop is a genuine intellectual position, not a failure of nerve: if the conceptual problem really is solved by dropping absolute facts, then a theory that keeps going is taking on risk it did not need to take.

The framework takes that risk deliberately. Whether it pays is not something an interferometer will settle; it is settled by whether the derivations hold.