Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Informational Pruning as a Third Path: Final-State Constraints between Consistent Quantum Causes and Quantum Causal Models

Ian Staley

Published May 19, 2026• 0 citations
Publisher

Abstract

Two recent programs offer competing accounts of quantum causation. Griffiths' Consistent Quantum Causes grounds causal reasoning in the Consistent Histories formalism, treating stochastic quantum time development and noncommuting projectors as the proper foundation for a Pearl-style directed acyclic graph representation. The Costa–Shrapnel Quantum Causal Models program takes the process-matrix formalism as its starting point, generalizing classical interventionist causal discovery to the quantum domain and extending naturally to indefinite causal order. Both programs assume that causal structure is fixed by an initial condition together with forward dynamics. Neither addresses what happens when admissibility is conditioned at both temporal boundaries. In this paper, I argue that final-state-constrained history ensembles, and the informational pruning they induce, provide a third vantage point from which to compare these two programs. I do not claim to resolve the Griffiths–QCM disagreement, nor to construct a new quantum causal formalism from scratch. The more limited claim is that once admissibility is globally filtered by an initial and a final boundary condition, the question of which coarse-grained histories carry the effective causal and computational structure of the process is reframed. The Hartle–Craig rank-product theorem supplies the formal backbone: imposing a final-state projector with rank r_ω restricts the admissible-history family to at most r_α·r_ω nonzero-probability histories, where r_α is the initial-state rank. I illustrate this with the three-box paradox, treated as a joint Consistent Histories and process-matrix scenario under final-state conditioning. I also suggest that an explainability stance, drawn in a methodological sense from explainable artificial intelligence, is useful for asking which admissible histories do the explanatory work under a given endpoint class. I conclude that informational pruning offers a coherent third path that is neither a competitor to Griffiths' approach nor an extension of process-matrix interventionism, but a complementary perspective that sharpens what each program leaves implicit.

DOI
10.5281/zenodo.20298255

Related papers

Final-State Constraints and Informational Pruning in Quantum Histories

Ian Staley

Quantum mechanics permits time-symmetric formulations in which intermediate events are conditioned by both an initial state and a final boundary condition. In this paper, I develop a model of quantum history in which a final boundary condition constrains the admissible-history ensemble rather than serving as a passive terminal state. I represent the universe’s possible macroscopic development as a history graph and define an admissible subgraph by imposing a final-state condition. I then examine whether, under a history-ensemble encoding, this restriction can reduce the effective description length of realized histories relative to unconstrained branching descriptions. I argue that the informational significance of the model depends on the structure of the terminal condition. Trivial terminal states impose weak selectivity and provide little compression. Structured terminal states impose stronger constraints and better support the emergence of organized intermediate worlds. Computational terminal states impose the strongest selectivity because they favor histories that generate information-processing structures, self-replicating systems, and explicit forms of computation. On this basis, I propose that sufficiently selective computational endpoints are most naturally interpreted in agentic terms, provided agency is understood structurally rather than psychologically. The model is retrocausal in a boundary-conditioned sense but not interventionist. It excludes inconsistent histories rather than revising an already fixed past and does not require operational backward signaling. I also distinguish between maximal pruning and robustness-preserving selectivity within computational endpoint classes. I conclude that a boundary-conditioned agentic interpretation of sufficiently selective computational endpoints is not established here as empirical fact, but it is rendered a mathematically serious interpretive hypothesis by the combination of dual-boundary quantum formalism, informational pruning, and endpoint complexity analysis.

International journal of quantum foundations•Apr 11, 2026•0 citations

The Thermodynamic Cost of a Final Boundary Condition: A Resource Theory of Informational Pruning and a Rank-Indexed Contextuality Conjecture

Ian Staley

In prior work I modelled the macroscopic development of a closed quantum system as a directed graph of coarse-grained histories on which a final boundary condition induces an admissible subgraph, and I graded terminal conditions into a trivial, structured, and computational typology. That programme left two structural gaps: it imported its formal backbones rather than proving results of its own, and it never joined its pruning half to its agency-and-contextuality half or assigned pruning a thermodynamic cost. This paper addresses both. I construct a resource theory whose objects are two-boundary-conditioned decoherent-history ensembles, whose free objects are those in which the terminal condition leaves the history distribution unchanged, and whose free operations are boundary-compatible coarse-grainings. Theorem 1 establishes that the relative entropy of the conditioned against the unconditioned history distribution is a faithful monotone under every such coarse-graining, jointly convex in the pair of history distributions; I also exhibit a counterexample showing that the more obvious counting-based deficit is not a monotone, which is why the relative-entropy form is the correct object. Theorem 2 establishes that a register satisfying the Reeb-Wolf hypotheses, carried from the unconditioned to the conditioned distribution, deposits heat at least kBT ln 2 times the entropy difference, and, combined with the Craig rank-product bound, at least kBT ln[n0 / (rαrω)] for equidistributed history weights. Lower terminal rank therefore raises the derived cost floor under the stated assumptions. I am explicit that this is a floor and not a total ordering: rank does not by itself order terminal conditions by resource content, and full rank does not imply a trivial condition. I then advance, explicitly as a conjecture rather than a theorem, a bridge to the contextuality half of the programme: along a nested family of normalized projectors terminating at the identity, within the paradoxical class, and against a noncontextual baseline, the Contextuality-by-Default measure of pre- and post-selected cyclic statistics should be non-increasing in the family index. I do not claim that the universe physically erases histories; the more limited claim is that a register implementing admissibility filtering pays a rank-graded cost. Conjecture C is, in the first instance, a mathematical conjecture about quantum statistics, settleable by computation, not a proposed departure from quantum mechanics.

•Aug 7, 2026•0 citations

Discovering Quantum Causal Models (final)

Sally Shrapnel

Costa and Shrapnel [2016] have recently proposed an interventionist theory of quantum causation. The formalism generalises the classical methods of Pearl [2000] and allows for the discovery of quantum causal structure via localised interventions. Classical causal structure is presented as a special case of this more general framework. I introduce the account and consider whether this formalism provides a causal explanation for the Bell correlations.

•Jun 4, 2017•0 citations

Consistent Quantum Causes

Robert B. Griffiths

Developing a quantum analog of the modern classical theory of causation, as formulated by Pearl and others using directed acyclic graphs, requires a theory of random or stochastic time development at the microscopic level, where the noncommutation of Hilbert-space projectors cannot be ignored. The Consistent Histories approach provides such a theory. How it works is shown by applying it to simple examples involving beam splitters and a Mach-Zehnder interferometer. It justifies the usual laboratory intuition that properly tested apparatus can reveal the earlier microscopic cause (e.g., as in radioactive decay) of a later macroscopic meassurement outcome. This approach is further illustrated by how it resolves the Bell inequalities paradox. The use of quantum circuits in discussions of quantum information in a time-irreversible manner can prevent the proper identification of earlier causes; this is illustrated using a specific circuit in the case of Bell inequalities. The approach to quantum causes known as Quantum Causal Models fails becuase it is not based upon a satisfactory theory of quantum random processes.

arXiv:2303.13617•Mar 23, 2023•0 citations

Superposition of dynamics, indefinite causal order, and quantum histories

Charis Anastopoulos, Ntina Savvidou

The process-matrix formalism describes quantum processes without assuming a fixed global causal order, but the physical meaning of indefinite causal order remains open. We address this question within histories theory, where temporal ordering and dynamical evolution are distinct structures. We show that the decoherence functional admits coherent superpositions of dynamics. In measurement settings, under suitable factorization conditions, these generate process matrices. Thus, we identify process-matrix indefinite causal order as an operationally restricted realization of the more general phenomenon of superposition of dynamics. Histories theory also admits a distinct kinematical notion of indefinite order, in which the ordering of physical events is itself a history observable. The resulting distinction between event order and intervention order clarifies the relation between indefinite causal order, quantum dynamics, and spacetime structure.

quant-pharXiv:2608.16694•Aug 18, 2026•0 citations

Discovering Quantum Causal Models

Sally Shrapnel

Costa and Shrapnel ([2016]) have recently proposed an interventionist theory of quantum causation. The formalism generalizes the classical methods of Pearl ([2000]) and allows for the discovery of quantum causal structure via localized interventions. Classical causal structure is presented as a special case of this more general framework. I introduce the account and consider whether this formalism provides a causal explanation for the Bell correlations. 1 Introduction2 Classical Causal Models3 What’s the (Quantum) Problem?4 Quantum Causal Models: Why Bother?5 Markov Quantum Causal Models6 The Bell Experiment7 Bell’s Objections

The British Journal for the Philosophy of Science•Aug 29, 2017•25 citations
Two recent programs offer competing accounts of quantum causation. Griffiths' Consistent Quantum Causes grounds causal reasoning in the Consistent Histories formalism, treating stochastic quantum time development and noncommuting projectors as the proper foundation for a Pearl-style directed acyclic graph representation. The Costa–Shrapnel Quantum Causal Models program takes the process-matrix formalism as its starting point, generalizing classical interventionist causal discovery to the quantum domain and extending naturally to indefinite causal order. Both programs assume that causal structure is fixed by an initial condition together with forward dynamics. Neither addresses what happens when admissibility is conditioned at both temporal boundaries. In this paper, I argue that final-state-constrained history ensembles, and the informational pruning they induce, provide a third vantage point from which to compare these two programs. I do not claim to resolve the Griffiths–QCM disagreement, nor to construct a new quantum causal formalism from scratch. The more limited claim is that once admissibility is globally filtered by an initial and a final boundary condition, the question of which coarse-grained histories carry the effective causal and computational structure of the process is reframed. The Hartle–Craig rank-product theorem supplies the formal backbone: imposing a final-state projector with rank r_ω restricts the admissible-history family to at most r_α·r_ω nonzero-probability histories, where r_α is the initial-state rank. I illustrate this with the three-box paradox, treated as a joint Consistent Histories and process-matrix scenario under final-state conditioning. I also suggest that an explainability stance, drawn in a methodological sense from explainable artificial intelligence, is useful for asking which admissible histories do the explanatory work under a given endpoint class. I conclude that informational pruning offers a coherent third path that is neither a competitor to Griffiths' approach nor an extension of process-matrix interventionism, but a complementary perspective that sharpens what each program leaves implicit.
Quantum mechanics permits time-symmetric formulations in which intermediate events are conditioned by both an initial state and a final boundary condition. In this paper, I develop a model of quantum history in which a final boundary condition constrains the admissible-history ensemble rather than serving as a passive terminal state. I represent the universe’s possible macroscopic development as a history graph and define an admissible subgraph by imposing a final-state condition. I then examine whether, under a history-ensemble encoding, this restriction can reduce the effective description length of realized histories relative to unconstrained branching descriptions. I argue that the informational significance of the model depends on the structure of the terminal condition. Trivial terminal states impose weak selectivity and provide little compression. Structured terminal states impose stronger constraints and better support the emergence of organized intermediate worlds. Computational terminal states impose the strongest selectivity because they favor histories that generate information-processing structures, self-replicating systems, and explicit forms of computation. On this basis, I propose that sufficiently selective computational endpoints are most naturally interpreted in agentic terms, provided agency is understood structurally rather than psychologically. The model is retrocausal in a boundary-conditioned sense but not interventionist. It excludes inconsistent histories rather than revising an already fixed past and does not require operational backward signaling. I also distinguish between maximal pruning and robustness-preserving selectivity within computational endpoint classes. I conclude that a boundary-conditioned agentic interpretation of sufficiently selective computational endpoints is not established here as empirical fact, but it is rendered a mathematically serious interpretive hypothesis by the combination of dual-boundary quantum formalism, informational pruning, and endpoint complexity analysis.
In prior work I modelled the macroscopic development of a closed quantum system as a directed graph of coarse-grained histories on which a final boundary condition induces an admissible subgraph, and I graded terminal conditions into a trivial, structured, and computational typology. That programme left two structural gaps: it imported its formal backbones rather than proving results of its own, and it never joined its pruning half to its agency-and-contextuality half or assigned pruning a thermodynamic cost. This paper addresses both. I construct a resource theory whose objects are two-boundary-conditioned decoherent-history ensembles, whose free objects are those in which the terminal condition leaves the history distribution unchanged, and whose free operations are boundary-compatible coarse-grainings. Theorem 1 establishes that the relative entropy of the conditioned against the unconditioned history distribution is a faithful monotone under every such coarse-graining, jointly convex in the pair of history distributions; I also exhibit a counterexample showing that the more obvious counting-based deficit is not a monotone, which is why the relative-entropy form is the correct object. Theorem 2 establishes that a register satisfying the Reeb-Wolf hypotheses, carried from the unconditioned to the conditioned distribution, deposits heat at least kBT ln 2 times the entropy difference, and, combined with the Craig rank-product bound, at least kBT ln[n0 / (rαrω)] for equidistributed history weights. Lower terminal rank therefore raises the derived cost floor under the stated assumptions. I am explicit that this is a floor and not a total ordering: rank does not by itself order terminal conditions by resource content, and full rank does not imply a trivial condition. I then advance, explicitly as a conjecture rather than a theorem, a bridge to the contextuality half of the programme: along a nested family of normalized projectors terminating at the identity, within the paradoxical class, and against a noncontextual baseline, the Contextuality-by-Default measure of pre- and post-selected cyclic statistics should be non-increasing in the family index. I do not claim that the universe physically erases histories; the more limited claim is that a register implementing admissibility filtering pays a rank-graded cost. Conjecture C is, in the first instance, a mathematical conjecture about quantum statistics, settleable by computation, not a proposed departure from quantum mechanics.
Developing a quantum analog of the modern classical theory of causation, as formulated by Pearl and others using directed acyclic graphs, requires a theory of random or stochastic time development at the microscopic level, where the noncommutation of Hilbert-space projectors cannot be ignored. The Consistent Histories approach provides such a theory. How it works is shown by applying it to simple examples involving beam splitters and a Mach-Zehnder interferometer. It justifies the usual laboratory intuition that properly tested apparatus can reveal the earlier microscopic cause (e.g., as in radioactive decay) of a later macroscopic meassurement outcome. This approach is further illustrated by how it resolves the Bell inequalities paradox. The use of quantum circuits in discussions of quantum information in a time-irreversible manner can prevent the proper identification of earlier causes; this is illustrated using a specific circuit in the case of Bell inequalities. The approach to quantum causes known as Quantum Causal Models fails becuase it is not based upon a satisfactory theory of quantum random processes.
The process-matrix formalism describes quantum processes without assuming a fixed global causal order, but the physical meaning of indefinite causal order remains open. We address this question within histories theory, where temporal ordering and dynamical evolution are distinct structures. We show that the decoherence functional admits coherent superpositions of dynamics. In measurement settings, under suitable factorization conditions, these generate process matrices. Thus, we identify process-matrix indefinite causal order as an operationally restricted realization of the more general phenomenon of superposition of dynamics. Histories theory also admits a distinct kinematical notion of indefinite order, in which the ordering of physical events is itself a history observable. The resulting distinction between event order and intervention order clarifies the relation between indefinite causal order, quantum dynamics, and spacetime structure.
Costa and Shrapnel ([2016]) have recently proposed an interventionist theory of quantum causation. The formalism generalizes the classical methods of Pearl ([2000]) and allows for the discovery of quantum causal structure via localized interventions. Classical causal structure is presented as a special case of this more general framework. I introduce the account and consider whether this formalism provides a causal explanation for the Bell correlations. 1 Introduction2 Classical Causal Models3 What’s the (Quantum) Problem?4 Quantum Causal Models: Why Bother?5 Markov Quantum Causal Models6 The Bell Experiment7 Bell’s Objections