---
title: 'Observer Rules: Physics, Computation & Control'
url: https://www.emergentmind.com/topics/observer-rules
type: topic
---

# Observer Rules: Physics, Computation & Control

Observer rules are prescriptions that specify what counts as an observer, what distinctions are accessible to that observer, and how dynamics, probabilities, Hilbert spaces, or state estimates are reformulated once those observational constraints are imposed. Across the literature, the phrase does not denote a single axiom system. In Cartan-geometric gravity it refers to the 7-dimensional observer space of future-timelike unit tangent vectors; in quantum measurement it denotes whole-history probability rules for sequences of observers’ experiences tied to material records; in semiclassical gravity it names extra prescriptions for appending or deriving an observer sector; in formal language theory it is the map \(O:\Sigma^*\to S\) that fixes observable input structure; and in control and discrete-event systems it denotes concrete observer constructions for state estimation and higher-order knowledge [1210.0019] [2009.12568] [2604.00139] [2606.27407] [1611.02518] [2408.06141].

## 1. Domain-specific meanings of observer rules

In the geometric literature, the observer is a localized timelike direction, not merely a point. For a time-oriented Lorentzian spacetime \((M,g)\), observer space is
\[
O=\{\,u\in TM\mid g(u,u)=-1,\ u \text{ future-directed}\,\},
\]
the future unit tangent bundle, and a point of \(O\) is an event together with a velocity. In quantum-measurement work, the observer is not introduced as a collapse-causing consciousness but as the bearer of possible experiences whose physical correlates are memories, notes, probes, and records. In semiclassical gravity, “observer rules” are extra prescriptions for how an observer sector is appended, factorized, or reconstructed when a closed universe would otherwise have a trivial Hilbert space. In computation, an observer is a function \(O:\Sigma^*\to S\) that determines which information about the input is accessible to the machine; in discrete-event systems, the classical observer is the powerset construction, and the high-order observer tracks what one agent knows about another agent’s state estimate [1210.0019] [2009.12568] [2604.00139] [2606.27407] [2408.06141].

These usages share a common formal role. Each fixes an admissible information channel before dynamics, inference, or reconstruction are defined. This suggests that observer rules are best understood not as a doctrine about subjectivity, but as structural constraints on accessibility, distinguishability, and closure.

## 2. Geometric and gravitational formulations

In “Lifting General Relativity to Observer Space” [1210.0019], observer space is treated as fundamental. The bundle projection \(\pi:O\to M\) has hyperbolic 3-space fibers \(H/K\cong \mathbb H^3\), and \(TO\) has the canonical splitting
\[
T_oO=(\text{boost vectors})\oplus(\text{spatial vectors})\oplus(\text{temporal vectors}).
\]
Observer space also carries a canonical contact form
\[
\alpha(v)=g(p(v),\pi_*v),
\]
whose Reeb vector field \(r\) is the restriction of the geodesic spray, so inertial observers are integral curves of \(r\). In Cartan-geometric form, an observer space geometry is a principal \(K=\mathrm{SO}(3)\)-bundle \((\pi:P\to O,A)\) with \(\mathfrak g\)-valued Cartan connection
\[
A=\Omega+b+\vec e+e_o,
\]
decomposing into rotation, boost, spatial translational soldering, and temporal translational soldering. Spacetime is reconstructible only if the boost distribution satisfies the curvature condition called \(\mathfrak h\)-flatness; with \(\mathfrak h\)-completeness and a free proper \(H\)-action, one obtains
\[
M=P/H.
\]
When reconstruction fails, coincidence and local spacetime become observer-dependent rather than globally defined. The same paper uses this formalism to link covariant and canonical gravity, including a foliation-independent route to Ashtekar-Barbero variables [1210.0019].

Recent semiclassical-gravity work shifts the term in a different direction. “The Observer Paradigm in Semiclassical Gravity: A Postmodern Perspective” [2604.00139] uses “observer rules” for the extra prescriptions needed to obtain a nontrivial perturbative Hilbert space in closed universes. The observer sector may be put in by hand or built from bulk semiclassical degrees of freedom, but it is “crucial” that, perturbatively, it tensor-factorizes from the remaining bulk degrees of freedom; this may require relaxing gravitational constraints, modifying the non-isometric code mapping, or altering the rules of the sum over topologies. One explicit modeling choice is
\[
\mathcal H_{\text{obs}}=L^2(\mathbbm{R}_+),
\]
and the paper emphasizes that the resulting Hilbert-space dimension depends on the observer’s intrinsic entropy, not only on gravitational parameters [2604.00139].

“Subregion observer rules from generalized entanglement wedges” [2606.28484] then gives a more technical unification. At the tensor-network level, the Colorado observer-promotion rules and the Kaya–Rath–Ritchie hollowing rules are exactly equivalent: both remove tensors in a bulk subregion \(a\), leaving dangling bulk and edge legs. In the path-integral language this leads to generalized entanglement wedge formulas such as
\[
S(\rho_a)=\frac{\mathcal A[\gamma_a(h_1)]}{4G_N}+S\!\left(\rho^{\mathrm{bulk}(h_1)}_{E(a)}\right),
\]
and allows the observer to be treated as a bulk subregion rather than merely a worldline [2606.28484].

A complementary development appears in “Observer complementarity for black holes and holography” [2507.06046]. There the standard approximate-isometry condition
\[
\langle\phi|V^\dagger V|\psi\rangle \approx \langle\phi|\psi\rangle
\]
is replaced by the observer-conditioned rule
\[
\mathrm{Tr}\!\left(V\,\mathcal C_{Ob}\!\left(|\psi\rangle\langle\phi|\right)V^\dagger\right)\approx \langle\phi|\psi\rangle,
\]
where the quantum-to-classical channel acts on pointer states by
\[
\mathcal C_{Ob}(|a\rangle\langle b|)=\delta_{ab}|a\rangle\langle a|.
\]
The observer is treated classically at the Heisenberg cut, yielding observer-dependent but operationally consistent bulk descriptions for exterior and infalling observers [2507.06046].

## 3. Quantum measurement and observer-dependent histories

In “Quantum measurements with, and yet without an Observer” [2009.12568], observer rules are Feynman’s probability rules applied to complete sequences of possible perceptions. For measurements at times \(t_0<t_1<\dots<t_L\), one assigns amplitudes to virtual paths,
\[
A(q^L_{n_L}\gets \dots \gets q^0_{n_0})
= \prod_{l=1}^L \langle q^l_{n_l}|\hat U(t_l,t_{l-1})|q^{l-1}_{n_{l-1}}\rangle,
\]
forms coarse-grained “real paths” by summing over indistinguishable alternatives, and then computes probabilities for full sequences of outcomes. The paper’s central claim is that these probabilities refer to observers’ experiences rather than to propositions that a system possessed values independently. The decisive criterion is record production: “without producing such a record an act of observation should not count.” Accordingly, the paper distinguishes coupling a probe, registering the probe in memory, and perceiving the record. Wigner’s friend scenarios are analyzed so that interference disappears when a physical record exists, even if no conscious perception has yet occurred; the actual act of perception is not dynamically relevant once a distinguishable material record is present [2009.12568].

A separate proposal, “Quantum measurement: a game between observer and nature?” [2210.16766], assigns the observer a decision-theoretic role rather than a collapse-inducing one. The observer prepares the apparatus, observes pointer states, interprets outcomes physically, and chooses actions \(a_i\) under uncertainty. The paper introduces a quantum expected value
\[
qEV=\sum_i p_i\sum_j \omega_j r_{ij},
\]
where \(p_i\) are subjective probabilities over actions, \(\omega_j\) are objective frequencies of nature’s states, and \(r_{ij}\) is the payoff matrix. Strategies are represented by quantum decision trees and optimized by quantum genetic programming with fitness
\[
f_{\text{fitness}}=\sum_{k=0}^N \langle V_k\rangle.
\]
The proposal explicitly rejects consciousness-causes-collapse, treats the observer as a strategic learner, and claims that optimized trees can reconstruct past trajectories “very well,” reporting about \(70\%\) accuracy in a simulated example. At the same time, it states that future trajectories cannot be accurately predicted because prior information about the quantum entity is incomplete [2210.16766].

Taken together, these quantum papers sharply separate two issues that are often conflated. One issue is whether the formalism is about observers’ experiences; the other is whether consciousness has a direct dynamical role. In both papers the first answer is affirmative and the second negative, but the mathematical implementations differ substantially.

## 4. Computation, language classes, and cryptography

In “Observers, Symmetries, and the Hierarchy of Language Classes: A Theory of Computation Parameterized by the Observer” [2606.27407], the observer is formalized as
\[
O:\Sigma^*\to S.
\]
A machine receives \(O(x)\), not \(x\), and a language is recognizable under \(O\) only if it is \(O\)-saturated:
\[
x\sim_O y \iff O(x)=O(y),\qquad
x\sim_O y \Rightarrow (x\in L \iff y\in L).
\]
The paper defines canonical observers including the trivial observer \(O_\bot(x)=\star\), the length observer \(O_{\mathrm{len}}(x)=|x|\), the parity observer \(O_{\mathrm{par}}(x)=(|x|_0\bmod 2,\ |x|_1\bmod 2)\), the profile observer
\[
O_{\mathrm{prof}}(x)=(|x|_0,|x|_1),
\]
the complete observer \(O_\top(x)=x\), and subsequence observers \(O_k\). The induced hierarchy satisfies
\[
O_\bot \prec O_{\mathrm{len}} \prec O_{\mathrm{prof}} \prec O_\top,\qquad
O_\bot \prec O_{\mathrm{par}} \prec O_{\mathrm{prof}} \prec O_\top,
\]
with \(O_{\mathrm{len}}\) and \(O_{\mathrm{par}}\) incomparable, and also
\[
O_\bot \prec O_1 \prec O_2 \prec \cdots \prec O_\top.
\]
Its order-blind automaton characterizes precisely the permutation-closed languages, and the paper proves the structural collapse
\[
\mathbf P_{O_{\mathrm{prof}}}=\mathbf{NP}_{O_{\mathrm{prof}}}\subsetneq \mathbf P.
\]
The point is that once only the profile is visible, nondeterminism cannot exploit lost order information [2606.27407].

“The Observer World: A Cryptographic Extension of Impagliazzo’s Five Worlds” [2606.27139] places this hierarchy on a second axis orthogonal to computational hardness. A world-observer pair is
\[
W_O=(W_i,O),
\]
where \(W_i\) is one of Algorithmica, Heuristica, Pessiland, Minicrypt, or Cryptomania, and \(O\) specifies the adversary’s observational access. The standard five-world framework corresponds to the identity observer \(I(x)=x\), treated there as the silent default. The paper proves that the profile-observer collapse holds unconditionally in all five worlds, so observational blindness and computational hardness are independent. It also shows that no one-way function can be \(O_{\mathrm{prof}}\)-saturated, because a profile-invariant function is invertible by reconstructing a canonical preimage from symbol counts. A parametric family \(W_O^\varepsilon\) is then introduced, with \(\varepsilon\in[0,1]\) measuring partial violation of observational invariants via mutual-information distance between \(O\), an intermediate \(O_\varepsilon\), and the identity observer [2606.27139].

A more philosophical response appears in “Intrinsic Computational Functionalism: From Observer-Relative Maps to Observer-Independent Structures” [2606.06424]. That paper distinguishes interpreter-relative label selection from dynamics-internal grain selection and argues that any consciousness-relevant computational property must satisfy two necessary criteria: system-intrinsic instantiation, meaning invariance under structure-preserving relabellings, and causal-dynamical organization under intervention, meaning counterfactual organization in a state-space whose variables mutually constrain one another. Its guiding separation is between arbitrary observer-imposed mappings and observer-independent causal organization under interventions [2606.06424].

## 5. Control, realization, and high-order estimation

In control theory, “observer rules” revert to a more classical meaning: update laws and solvability conditions for reconstructing unmeasured variables. “Observer design for piecewise smooth and switched systems via contraction theory” [1611.02518] studies bimodal Filippov systems with switching manifold \(\Sigma=\{x:h(x)=0\}\) and proposes the mode-dependent observer
\[
\dot{\hat x}=
\begin{cases}
f^+(\hat x)+L^+(y-\hat y)+u(t), & h(\hat x)>0,\\
f^-(\hat x)+L^-(y-\hat y)+u(t), & h(\hat x)<0,
\end{cases}
\qquad \hat y=g(\hat x).
\]
Using regularization and matrix measures, the paper gives sufficient contraction conditions in each mode and a switching-manifold compatibility condition ensuring exponential convergence of the estimation error. A central contribution is that the design can use non-Euclidean norms, not only Euclidean ones [1611.02518].

For PDE–ODE systems, “Observer Design for Systems of Conservation Laws with Lipschitz Nonlinear Boundary Dynamics” [2101.06719] considers
\[
\partial_t x+\Lambda \partial_z x=0,\qquad x(t,0)=C\chi(t),\qquad \dot\chi=A\chi+B\Psi(Z\chi),\qquad y=Mx(t,1),
\]
and constructs an infinite-dimensional observer with linear boundary injection
\[
\partial_t \hat x+\Lambda \partial_z \hat x=0,\qquad \hat x(t,0)=C\hat\chi(t),\qquad
\dot{\hat\chi}=A\hat\chi+B\Psi(Z\hat\chi)+L\bigl(y-M\hat x(t,1)\bigr).
\]
A Lyapunov functional with cross terms between PDE and ODE errors yields sufficient matrix-inequality conditions, though the paper notes that the synthesis condition is in fact a BMI rather than a convex LMI [2101.06719].

Distributed observer rules appear in “A Class of LTI Distributed Observers for LTI Plants” [1401.0926]. Each observer \(i\) updates via
\[
\hat{x}_i(k+1)
=
A\sum_{j\in\mathbb N_i}\mathbf w_{ij}\hat{x}_j(k)
+\mathbf K_i\bigl(y_i(k)-H_i\hat x_i(k)\bigr)
+\mathbf P_i z_i(k),
\]
\[
z_i(k+1)=\mathbf Q_i\bigl(y_i(k)-H_i\hat x_i(k)\bigr)+\mathbf S_i z_i(k),
\]
where \(\mathbb N_i\) is determined by a directed communication graph. The paper proves a necessary and sufficient condition for asymptotic omniscience at all observers: every source strongly connected component of the communication graph must yield a detectable subsystem \((A,H_{\mathbb V_i})\) [1401.0926].

Algebraic observer synthesis is developed in “Rational observers of rational systems” [1612.07962]. There the state is reconstructed from finitely many output derivatives
\[
s_1=h,\qquad s_{k+1}=L_f s_k,
\]
and finite algebraic observability guarantees an output-based realization in the same algebraic class. The observer is then obtained by output injection into the companion-like realization. The paper emphasizes that a polynomial observer may require strictly higher state dimension than the original plant [1612.07962].

“Observer-Based Realization of Control Systems” [2404.15688] takes a dual perspective: observers are functions of the state, and an observer-based realization is a Lebesgue-type dynamic system written directly on those observer variables. For linear systems \(\dot x=Ax+Bu,\ y=Hx\), an exact OR-system exists if and only if the dual subspace \(\mathcal H_*=\mathrm{Span}(\mathrm{Row}(H))\) is \(A\)-invariant, equivalently
\[
HA=\Xi H
\]
for some \(\Xi\). When this fails, the paper constructs an extended OR-system on the smallest \(A\)-invariant, or under feedback the smallest \((A,B)\)-invariant, dual subspace containing \(\mathcal H_*\) [2404.15688].

In discrete-event systems, finally, “High-order observers and high-order state-estimation-based properties of discrete-event systems” [2408.06141] generalizes the Rabin–Scott powerset observer from one agent to an ordered family \(A_1,\dots,A_n\). Each agent \(A_i\) has its own observable event set \(E_i\), all agents know the system structure, and higher-order observers encode what \(A_n\) knows about what \(A_{n-1}\) knows about \(\dots\) what \(A_1\) knows about the plant’s state estimate [2408.06141].

## 6. Structural themes, controversies, and open problems

Several structural motifs recur across these otherwise disparate literatures. First, observer rules almost always define an accessibility relation before any dynamics are analyzed: in general relativity it is the decomposition into boosts, spatial directions, and temporal Reeb flow; in quantum measurement it is distinguishability by material records; in computation it is the partition of \(\Sigma^*\) into \(\sim_O\)-classes; in control it is the map from measurements or neighboring estimates to corrected internal state. Second, exactness is characterized by a closure condition: \(\mathfrak h\)-flatness for spacetime reconstruction, \(HA=\Xi H\) for observer-based realizations, \(O\)-saturation for observer-relative language classes, and invariant co-distributions in the affine nonlinear case [1210.0019] [2404.15688] [2606.27407].

The controversies are equally consistent. No paper surveyed here produces a universally accepted, observer-free master formalism. The semiclassical-gravity literature explicitly states that there is “spirited debate” about observer rules and treats the observer sector as a source of both nontriviality and ambiguity in closed-universe Hilbert spaces [2604.00139]. The quantum-measurement literature avoids consciousness as a dynamical variable, but one paper makes an extra assumption that all physically relevant knowledge is stored in material records, and another introduces speculative constructs such as a superposed “mental state” and quantum decision trees [2009.12568] [2210.16766]. In control and realization theory, the gaps are technical rather than philosophical: BMI synthesis is nonconvex, global convergence is not established in full generality for rational and polynomial observers, and high-order discrete-event observers can be combinatorially large [2101.06719] [1612.07962] [2408.06141].

Open problems are stated explicitly in several domains. Observer-space gravity leaves matter coupling undeveloped and asks how horizons, Lorentz-violating theories, lightlike observers, and relative-locality or Finsler-type constructions should be formulated directly on observer space [1210.0019]. Semiclassical-gravity work leaves unresolved whether the observer Hilbert space should be \(L^2(\mathbb R_+)\), whether the observer is inserted by hand or emerges from bulk degrees of freedom, and how observer-dependent and observer-free descriptions should be related [2604.00139]. Computational observer theory opens questions about adaptive observers, partial invariant violations \(W_O^\varepsilon\), and the interface between formal observation maps and thermodynamic, quantum, or cosmological information bounds [2606.27139].

The resulting picture is not a single theory of observers. It is a family of mathematically precise reparameterizations of physical, computational, and inferential structure by what is observable, stable, or reconstructible from a designated observational channel. In that sense, observer rules function less as a metaphysics of subjectivity than as a technical language for when and how objectivity is recovered from observer-indexed data.

Source: https://www.emergentmind.com/topics/observer-rules