Pith. sign in

REVIEW 3 major objections 5 minor 44 references

The Observational Partial Order of Causal Structures with Latent Variables

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read The paper gives a complete observational dominance order for causal structures with latent variables on three visible variables and a partial order for four.

desk verdict A genuine advance on the 3-node observational partial order and two new facet-merging rules, but the 4-node counts rest on an external algorithm and one direct proof in Sec. 6.2 is wrong as written. read the letter →

arxiv 2502.07891 v2 pith:EKZ6YFTS submitted 2025-02-11 stat.ML cs.LGquant-ph

classification stat.MLcs.LGquant-ph MSC 62D2005C2068T37
keywords observationaldominancecausaldiscoverylatentvariablesmDAGsequivalenceinequalityconstraintspartialordersupportrealizability
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks which causal structures with hidden variables can be told apart from observational data alone, and in what direction their explanatory power runs. It defines observational dominance as containment of the sets of realizable distributions over the visible variables, so a structure that can explain more distributions dominates one that can explain fewer. The paper's main result is a complete map of this dominance order for all causal structures on three visible variables (fifteen equivalence classes) and a partial map for four visible variables (between 1,253 and 1,444 classes, with 1,156 fully identified). Along the way it proves two new rules for showing when adding a common-cause facet does not change the observational profile, and it shows that a brute-force support-comparison method subsumes all previously known graphical tests for non-dominance. This matters for causal discovery because it tells practitioners when conditional-independence tests are too weak: for four variables, fewer than 10 percent of classes can be singled out by conditional independence alone, and at least 85 percent of classes obey nontrivial inequality constraints.

What carries the argument

The workhorse is the mDAG (marginalized directed acyclic graph): a DAG whose visible nodes carry directed edges for direct causation together with a simplicial complex of faces recording which visible subsets share an unobserved common cause. The observational profile of an mDAG is the tuple, over all cardinalities of the visible variables, of the sets of distributions it can realize; dominance is elementwise set inclusion of these profiles. Dominance is established by structural edge and facet additions plus two new facet-merging rules (Moderate and Strong), which show when merging common-cause facets preserves or extends the realizable-distribution set. Nondominance is decided hierarchically: skeleton comparison, d-separation, e-separation, densely connected node pairs, the directed-edge-free rule, and finally comparison of unrealizable supports, where an enumeration of response functions decides which support sets are realizable. A key structural result is that the support-comparison step subsumes all the cheaper graphical nondominance rules, so the remaining gaps in the four-node case are purely about support sets at higher cardinalities.

What would settle it

Take the support set with three binary events that the paper uses to separate two of the three-node classes and independently search for a latent-variable model realizing it, allowing latent cardinalities beyond the enumeration's bound; a successful construction would invalidate the claimed three-node nondominance and with it the completeness of the fifteen-class order.

Watch

Extended reading notes

Core claim

The central claim is that the observational partial order of latent-permitting causal structures is accessible in full for three visible variables and in large part for four, when one fixes the temporal ordering of the visible nodes. For three variables, the 72 candidate graphs collapse into 15 observational equivalence classes, and every dominance relation among them is identified, yielding the first complete such order. For four variables, the paper brackets the number of equivalence classes between 1,253 and 1,444 and completely identifies 1,156 of them, leaving the rest as well-defined open cases. The paper further claims that nonalgebraic classes, meaning those whose realizable-distribution sets are cut out by inequalities rather than equalities alone, are the majority: at least 85.2 percent of four-variable classes, up from one third at three variables. It also claims that conditional-independence relations alone identify fewer than 10 percent of four-variable classes, establishing that richer constraints are needed for causal discovery.

Load-bearing premise

Everything the paper concludes about non-dominance presupposes that the combinatorial enumeration of realizable supports, including its bound on how large the hidden variables need to be, is correct and complete; if that enumeration misses a support, the claimed three-node classification and the four-node lower bounds would not be established.

Editorial extensions

If this is right

  • For three visible variables, the complete fifteen-class order means any new causal-compatibility constraint found for one class automatically transfers to its class, and realizability of a distribution can be propagated up or down the order.
  • For four variables, conditional-independence-based algorithms can single out fewer than 10 percent of classes, so nested constraints and inequality constraints are not optional extras for causal discovery.
  • With at least 85.2 percent of four-variable classes nonalgebraic, most causal structures can in principle show quantum-classical gaps, since such gaps require inequality constraints.
  • The two new facet-merging rules tighten the upper bound on four-node classes from 1,481 to 1,444, showing that previously known equivalence rules missed real equivalences.
  • Because support comparison subsumes all graphical nondominance rules, the unresolved four-node cases are isolated to finite computational searches over supports, not to missing graphical criteria.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the monotone trend in the fraction of nonalgebraic classes suggests that for five or more visible variables nearly every observational equivalence class will carry inequality constraints; this is an extrapolation, not a result in the paper.
  • The seven mDAGs that realize every binary support but may not be saturated form a natural test bed for whether support-level equality implies distribution-level equality; a single counterexample would separate possibilistic from probabilistic causal compatibility.
  • The facet-merging rules point toward a combinatorial closure operation on facets that might decide observational equivalence for arbitrary node counts; the paper does not claim such an operation exists.
  • For quantum causation, the ubiquity of nonalgebraic classes implies that new quantum-classical gaps should be sought in generic four-variable networks rather than specially designed ones; this is an implication the paper only gestures at.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper studies the observational dominance partial order on mDAGs (marginalized DAGs) with latent variables, under a fixed ordering of the visible variables. It assembles known dominance- and nondominance-proving rules, adds two new facet-merging rules (Moderate and Strong Facet-Merging), and augments the graphical nondominance rules with a support-based criterion derived from Fraser's algorithm. The central claims are: for three visible nodes, the 72 mDAGs consistent with a fixed nodal ordering form exactly 15 observational equivalence classes whose dominance relations are completely determined; for four visible nodes, the number of classes lies between 1253 and 1444, with 1156 classes fully identified. From these results the paper derives lower bounds on the prevalence of nonalgebraic classes (at least 85.2% for four nodes) and argues that conditional-independence constraints alone identify fewer than 10% of four-node classes, so that richer constraints are needed for causal discovery.

Significance. If the results are correct, the complete 3-node observational partial order and the partially characterized 4-node order are valuable reference points for causal discovery with latent variables, and the evidence that nonalgebraic classes are generic strengthens the case for going beyond conditional-independence constraints. The paper's care in distinguishing proven-equivalence from proven-inequivalence partitions, its systematic application of existing rules, and its public code repository are explicit strengths. The two new facet-merging rules are proved in the appendices and are likely to be useful beyond the specific classification. However, the classification leans heavily on an external computational algorithm, and one of the paper's own written proofs of a load-bearing support separation is incorrect; these issues must be fixed before the central claims can be accepted as established.

major comments (3)
  1. [Sec. 6.2] The direct proof that Instrumental CAB realizes the support S' is incorrect. With Xa = Xlambda*Xgamma, Xb = Xlambda*(Xa XOR 1), Xc = Xgamma, the four binary assignments of (Xlambda, Xgamma) give the support {(0,0,0),(0,0,1),(0,1,0),(1,0,1)}, not S' = {(1,0,0),(0,0,1),(0,1,1),(0,0,0)}. Since the inequivalence between Instrumental BAC and Instrumental CAB is required to close the 3-node classification, the written proof fails at a load-bearing point. The separation may still be true (indeed, S' is realizable by Instrumental CAB under a different parametrization), but the text must either supply a correct explicit construction or state plainly that this inequivalence rests on Fraser's algorithm and the associated computation.
  2. [Sec. 5.2 and Appendix B] The 3-node closure, the 4-node lower bound of 1253, the 1156 identified classes, and the derived 85.2% nonalgebraic fraction all depend on Fraser's algorithm for deciding which supports are realizable, including the bound k >= s on latent cardinalities. This algorithm and bound are cited from Ref. [39] and are not proved in the manuscript. Because a misclassification of a single support would over-split the proven-inequivalence partition and invalidate the interval [1253,1444], the paper should either include a self-contained proof of the algorithm's correctness and of the latent-cardinality bound, or state them as explicit theorems with precise references and make the computational verification (e.g., the repository scripts used to generate Tables 4 and 5) fully reproducible.
  3. [Sec. 6.2 and Table 2] For three binary visible variables, a support can contain up to 8 events, yet Table 2 and the text report comparisons only up to 4-event supports. The paper does not explain why checking supports of size at most 4 is sufficient to certify the absence of any further inequivalences. If supports of size 5-8 were not checked, the claimed completeness of the 3-node classification is not established. Please either report that all 2^8 possible supports were checked, or provide a reduction argument showing that any separating support of size greater than 4 would imply a separating support of size at most 4.
minor comments (5)
  1. [Sec. 8] The conclusion inverts the conditional-independence statistics: the text says 'for 4-node mDAGs ... 2/3 ... for 3-node mDAGs ... less than 10%', but Section 7.2.2 correctly reports 10/15 = 2/3 for three nodes and fewer than 10% for four nodes. This should be corrected.
  2. [Sec. 7] The text states that the best lower bound on the number of 4-node classes is 1256, while Table 4 and the subsequent calculation use 1253. These numbers must be reconciled.
  3. [Sec. 5.2.1] There is a typo in the paragraph after Lemma 7: 'it is possible to do find such a support' should read 'it is possible to find such a support'.
  4. [Sec. 6.2] The sentence 'We infern that Evans is an observational equivalence class' contains a typo ('inf ernn' should be 'infer').
  5. [Fig. 4.1 caption] The caption says 'Classification of the different nondominance-proving rules' for what appears to be the dominance-proving rules diagram; the two captions in Fig. 4.1 and Fig. 5.1 appear to be interchanged or duplicated.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the derivations are self-contained; self-citations are not load-bearing. One direct-support construction in Sec. 6.2 appears erroneous, but that is a correctness risk, not a circular reduction.

full rationale

The central claims do not reduce to their inputs. The 3-node proven-equivalence partition (upper bound 15) is obtained from dominance rules whose proofs (Appendix A, Bayesian updating and exogenization) do not assume the target classes; the proven-inequivalence partition (lower bound 15) is obtained from independent nondominance rules, including d-separation and Fraser's support-realizability algorithm from Ref. [39]. The 4-node bounds [1253, 1444] and the 1156 identified classes likewise combine provable dominance rules with support computations; no parameter is fitted to a subset of data and then renamed a prediction. The algebraicness criterion is imported from Evans's external theorem [33], and the algebraic-class upper bound 185 comes from d-separation counting of confounder-free mDAGs, not from the paper's fitted values. The paper cites prior work by the same authors ([25], [8], [35]), including the fixed-nodal-ordering motivation and the nonalgebraicness of particular structures, but these citations are not load-bearing for the final counts; the final lower and upper bounds rest on the independently stated algorithms and lemmas. One flagged passage is Section 6.2: the 'direct proof' for support S' defines Xa = Xλ·Xγ, Xb = Xλ·(Xa⊕1), Xc = Xγ, which generates the support {(0,0,0),(0,0,1),(0,1,0),(1,0,1)} rather than the claimed S' = {(1,0,0),(0,0,1),(0,1,1),(0,0,0)}. This apparent error leaves that particular written separation dependent on Fraser's algorithm and is a verification weakness, but it is not an instance of a result being defined or fitted into existence. There is therefore no circular step of the enumerated kinds.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

No parameters are fitted and no new entities are introduced. The paper's contribution is combinatorial classification; it relies on standard causal semantics, external theorems, and an external algorithm for support realizability.

assumptions (6)
  • domain assumption Classical unrestricted semantics with arbitrary cardinalities for latent variables (Definition 5).
    All realizability claims are relative to this semantics; for example, latent variables can absorb redundant effects, which underlies the mDAG reductions.
  • domain assumption Restriction to a fixed nodal ordering of visible variables (Definition 4).
    The partial orders and counts are stated only for mDAGs consistent with one ordering; results do not compare structures with different orderings.
  • standard math d-separation implies conditional independence and vice versa for classical models (Theorem 1).
    Used to prove nondominance rules by comparing d-separation relations.
  • domain assumption An mDAG is algebraic if and only if it is observationally equivalent to a confounder-free mDAG (Ref. [33]).
    Foundation for identifying algebraic classes and for the upper bound on the number of algebraic classes.
  • domain assumption Fraser's algorithm correctly characterizes realizable supports for finite visible cardinalities (Ref. [39]).
    Supports found by this algorithm are used to prove observational nondominance, including the completeness of the 3-node order.
  • domain assumption The set of d-separation equivalence classes of latent-free 4-node DAGs has 185 elements.
    Used to bound the number of algebraic 4-node classes; asserted without derivation in Section 7.2.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Observational Partial Order of Causal Structures with Latent Variables." pith.science (2026). https://pith.science/paper/EKZ6YFTS

@misc{pith2026250207891,
  author       = {Pith},
  title        = {Pith review of: The Observational Partial Order of Causal Structures with Latent Variables},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EKZ6YFTS}},
  note         = {Machine review of arXiv:2502.07891}
}
read the original abstract

For two causal structures with the same set of visible variables, one is said to observationally dominate the other if the set of distributions over the visible variables realizable by the first contains the set of distributions over the visible variables realizable by the second. Knowing such dominance relations is useful for adjudicating between these structures given observational data. We here consider the problem of determining the partial order of equivalence classes of causal structures with latent variables relative to observational dominance. We provide a complete characterization of the dominance order in the case of three visible variables, and a partial characterization in the case of four visible variables. Our techniques also help to identify which observational equivalence classes have a set of realizable distributions that is characterized by nontrivial inequality constraints, analogous to Bell inequalities and instrumental inequalities. We find evidence that as one increases the number of visible variables, the equivalence classes satisfying nontrivial inequality constraints become ubiquitous. (Because such classes are the ones for which there can be a difference in the distributions that are quantumly and classically realizable, this implies that the potential for quantum-classical gaps is also ubiquitous.) Furthermore, we find evidence that constraint-based causal discovery algorithms that rely solely on conditional independence constraints have a significantly weaker distinguishing power among observational equivalence classes than algorithms that go beyond these (i.e., algorithms that also leverage nested Markov constraints and inequality constraints).

Figures

Figures reproduced from arXiv: 2502.07891 by the authors.

Figure 1.1
Figure 1.1. Two observationally equivalent causal struc [PITH_FULL_IMAGE:figures/full_fig_p004_1_1.png] view at source ↗
Figure 7
Figure 7. (a)-(d) are known to present inequality con [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 2.1
Figure 2.1. An example of nonalgebraic pDAGs that are [PITH_FULL_IMAGE:figures/full_fig_p010_2_1.png] view at source ↗
Figures from the paper (35 more)
Figure 2
Figure 2. Figure 2: ; the observational equivalence of the three [PITH_FULL_IMAGE:figures/full_fig_p010_2.png]
Figure 2.2
Figure 2.2. Figure 2.2: In Fig. 2.2(a), we have a pDAG where the [PITH_FULL_IMAGE:figures/full_fig_p011_2_2.png]
Figure 2
Figure 2. Figure 2: (c). The operation of transitioning from the [PITH_FULL_IMAGE:figures/full_fig_p011_2.png]
Figure 2.2
Figure 2.2. Figure 2.2: (a) A pDAG. (b) The pDAG obtained from (a) by exogenizing the latent node [PITH_FULL_IMAGE:figures/full_fig_p012_2_2.png]
Figure 2.4
Figure 2.4. Figure 2.4: (a) mDAG with trivial directed structure and with simplicial complex B = {{a}, {b}, {c}, {a, b}, {a, c}, {b, c}}. (b) mDAG with trivial directed structure and with simplicial complex B ′ = {{a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}}.The facets (inclusion-maxim…
Figure 2.5
Figure 2.5. Figure 2.5: Observational partial order for the set of two [PITH_FULL_IMAGE:figures/full_fig_p014_2_5.png]
Figure 3.1
Figure 3.1. Figure 3.1: The 72 mDAGs with three visible nodes with nodal ordering [PITH_FULL_IMAGE:figures/full_fig_p017_3_1.png]
Figure 3
Figure 3. Figure 3: illustrates how we express partial infor [PITH_FULL_IMAGE:figures/full_fig_p021_3.png]
Figure 3.2
Figure 3.2. Figure 3.2: The proven-equivalence partition (pink loops) is found by applying the equivalence rules of Section 4. [PITH_FULL_IMAGE:figures/full_fig_p022_3_2.png]
Figure 4.1
Figure 4.1. Figure 4.1: Classification of the different dominance-proving rules that we use. Classification of the different [PITH_FULL_IMAGE:figures/full_fig_p024_4_1.png]
Figure 4.2
Figure 4.2. Figure 4.2: Two mDAGs that can be proven to be obser [PITH_FULL_IMAGE:figures/full_fig_p024_4_2.png]
Figure 4.4
Figure 4.4. Figure 4.4: The first and third mDAG of this figure can be shown observationally equivalent by Evans’ rule, as well [PITH_FULL_IMAGE:figures/full_fig_p025_4_4.png]
Figure 4.5
Figure 4.5. Figure 4.5: Two mDAGs that can be proven observa￾tionally equivalent by Moderate Facet-Merging, but not by Weak Facet-Merging. Here, the sets B, C and D from the statement of Proposition 5 are respectively B = {a, b, d, e}, C = {a, b} and D = {d, e}. 1. paD(C) ∪ C ⊆ paD(d) for e…
Figure 6.1
Figure 6.1. Figure 6.1: This result shows conditions under which multiple facets can be merged simultaneously. In the example of [PITH_FULL_IMAGE:figures/full_fig_p026_6_1.png]
Figure 4.6
Figure 4.6. Figure 4.6: Two mDAGs that can be shown observation [PITH_FULL_IMAGE:figures/full_fig_p026_4_6.png]
Figure 5.1
Figure 5.1. Figure 5.1: Classification of the different nondominance-proving rules. An arrow from one rule to another indicates [PITH_FULL_IMAGE:figures/full_fig_p028_5_1.png]
Figure 5.2
Figure 5.2. Figure 5.2: (a) An mDAG; and (b) Its skeleton. G ′ is also realizable by G, and thus has to satisfy Xx ⊥⊥ Xy∣Xz. By the second point of Theorem 1, this would imply that x ⊥⊥d y∣z in G ′ , which is a contradiction. Therefore, we can use the comparison of d￾separation relations to…
Figure 5.4
Figure 5.4. Figure 5.4: Example of mDAGs that have the same set of d-separation relations (i.e., no d-separation relations), but different skeletons. In [PITH_FULL_IMAGE:figures/full_fig_p029_5_4.png]
Figure 5
Figure 5. Figure 5: (b). This in turn gives us [PITH_FULL_IMAGE:figures/full_fig_p030_5.png]
Figure 5.5
Figure 5.5. Figure 5.5: The districts of this mDAG are {a, b, c, h}, {d} and {e, f, g}. The second preliminary definition we need is the closure of a set of nodes, first introduced in Ref. [37]. Definition 16 (Closure of a set of nodes). Let G = {D, B} be an mDAG, and let A ⊆ nodes(G). Set …
Figure 5.6
Figure 5.6. Figure 5.6: Steps used to find the closure of the set [PITH_FULL_IMAGE:figures/full_fig_p031_5_6.png]
Figure 5.7
Figure 5.7. Figure 5.7: Example of a pair of mDAGs that have the [PITH_FULL_IMAGE:figures/full_fig_p031_5_7.png]
Figure 5.8
Figure 5.8. Figure 5.8: Example of a pair of mDAGs that have the [PITH_FULL_IMAGE:figures/full_fig_p032_5_8.png]
Figure 3
Figure 3. Figure 3: , the mDAG of Fig. 5.7(b) is not observa [PITH_FULL_IMAGE:figures/full_fig_p032_3.png]
Figure 5.9
Figure 5.9. Figure 5.9: The directed-edge-free mDAGs are represented by blue dots, while the other mDAGs are red dots. [PITH_FULL_IMAGE:figures/full_fig_p033_5_9.png]
Figure 6.1
Figure 6.1. Figure 6.1: Three observationally equivalent mDAGs, that lie in the Instrumental ABC class. mDAG of [PITH_FULL_IMAGE:figures/full_fig_p035_6_1.png]
Figure 6
Figure 6. Figure 6: , and the third pDAG of Fig. 6.2 corresponds [PITH_FULL_IMAGE:figures/full_fig_p035_6.png]
Figure 6.2
Figure 6.2. Figure 6.2: Intermediary steps to show observational [PITH_FULL_IMAGE:figures/full_fig_p036_6_2.png]
Figure 6.3
Figure 6.3. Figure 6.3: Representation of the information obtained by applying the comparison of skeletons to the proven [PITH_FULL_IMAGE:figures/full_fig_p037_6_3.png]
Figure 6.4
Figure 6.4. Figure 6.4: Representation of the information obtained by applying the comparison of skeletons and the comparison [PITH_FULL_IMAGE:figures/full_fig_p038_6_4.png]
Figure 7
Figure 7. Figure 7: (b) shows an example of a causal struc [PITH_FULL_IMAGE:figures/full_fig_p042_7.png]
Figure 7.1
Figure 7.1. Figure 7.1: (a) Bell mDAG. (b) Square mDAG. Both of these belong to equivalence classes that are com￾pletely identified by d-separation, meaning that there is no mDAG that is observationally inequivalent to them while still imposing the same conditional independence constraints …
Figure 3.1
Figure 3.1. Figure 3.1: they are algebraic, but the confounder-free [PITH_FULL_IMAGE:figures/full_fig_p044_3_1.png]
Figure 7.2
Figure 7.2. Figure 7.2: Seven mDAGs that do not appear in the same [PITH_FULL_IMAGE:figures/full_fig_p046_7_2.png]
Figure 7.3
Figure 7.3. Figure 7.3: Diagrammatic representation of the proven-equivalence and proven-inequivalence partitions of 4-node [PITH_FULL_IMAGE:figures/full_fig_p047_7_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 43 canonical work pages

  1. [24]

    Robin J. Evans. Graphs for Margins of Bayesian Networks. Scandinavian Journal of Statistics, 43(3):625–648, 2016. Note: Whenever we cited Propositions and Lem- mas from this reference, we mentioned the num- bering of the arXiv version, that is different from the numbering of the published version. This choice was made because the arXiv version may be more...

  2. [39]

    A Combinatorial Solution to Causal Compatibility

    Thomas Fraser. A Combinatorial Solution to Causal Compatibility. Journal of Causal Infer- ence, 8:22–53, 07 2020. A Proof of Strong Facet- Merging In this appendix we will prove the Strong Facet- Merging proposition, which is reproduced below. Proposition 8 (Strong Facet-Merging (Strong FM)). Let G = {D, B} be an mDAG whose simpli- cial complex contains a...

  3. [1]

    J. S. Bell. On the Einstein Podolsky Rosen para- dox. Physics Physique Fizika, 1:195–200, Nov 1964

  4. [2]

    Beyond Bell 's theorem: cor- relation scenarios

    Tobias Fritz. Beyond Bell 's theorem: cor- relation scenarios. New Journal of Physics , 14(10):103001, oct 2012

  5. [3]

    Quantum vi- olations in the Instrumental scenario and their relations to the Bell scenario.Quantum, 3:186, September 2019

    Thomas Van Himbeeck, Jonatan Bohr Brask, Stefano Pironio, Ravishankar Ramanathan, Ana Belén Sainz, and Elie Wolfe. Quantum vi- olations in the Instrumental scenario and their relations to the Bell scenario.Quantum, 3:186, September 2019

  6. [4]

    Quantum non-classicality in the simplest causal network, 2024

    Pedro Lauand, Davide Poderini, Rafael Rabelo, and Rafael Chaves. Quantum non-classicality in the simplest causal network, 2024

  7. [5]

    Escap- ing the shadow of bell’s theorem in network nonlocality, 2024

    Maria Ciudad-Alañón, Emanuel-Cristian Boghiu, Paolo Abiuso, and Elie Wolfe. Escap- ing the shadow of bell’s theorem in network nonlocality, 2024

  8. [6]

    Joe Henson, Raymond Lal, and Matthew F. Pusey. Theory-independent limits on cor- relations from generalized Bayesian networks. 16(11):113043, 2014. Publisher: IOP Publish- ing

Show all 44 references
  1. [7]

    Which causal structures might support a quantum–classical gap?New Journal of Physics, 19(4):043021, apr 2017

    Jacques Pienaar. Which causal structures might support a quantum–classical gap?New Journal of Physics, 19(4):043021, apr 2017

  2. [8]

    Pusey, and Elie Wolfe

    Shashaank Khanna, Marina Maciel Ansanelli, Matthew F. Pusey, and Elie Wolfe. Classify- ing causal structures: Ascertaining when classi- cal correlations are constrained by inequalities. Phys. Rev. Res., 6:023038, Apr 2024

  3. [9]

    Equivalence and synthesis of causal models

    Thomas Verma and Judea Pearl. Equivalence and synthesis of causal models. InProceedings of the Sixth Annual Conference on Uncertainty in Artificial Intelligence, UAI ’90, page 255–270, USA, 1990. Elsevier Science Inc

  4. [10]

    Spirtes, C

    P. Spirtes, C. Glymour, and R. Scheines.Cau- sation, Prediction, and Search. MIT press, 2nd edition, 2000

  5. [11]

    Evans, Thomas S

    Ilya Shpitser, Robin J. Evans, Thomas S. Richardson, and James M. Robins. Introduction to Nested Markov Models. Behaviormetrika, 41:3–39, 01 2014

  6. [12]

    Robin J. Evans. Latent free equivalent mdags. Algebraic Statistics, 14(1):3–16, November 2023

  7. [13]

    Information–theoretic implications of quantum causal structures

    Rafael Chaves, Christian Majenz, and David Gross. Information–theoretic implications of quantum causal structures. Nature Communi- cations, 6(1):5766, January 2015

  8. [14]

    Quantum Inflation: A Gen- eral Approach to Quantum Causal Compatibil- ity

    Elie Wolfe, Alejandro Pozas-Kerstjens, Matan Grinberg, Denis Rosset, Antonio Acín, and Miguel Navascués. Quantum Inflation: A Gen- eral Approach to Quantum Causal Compatibil- ity. Phys. Rev. X, 11:021043, May 2021

  9. [15]

    Smith, Elie Wolfe, and Robert W

    Isaac D. Smith, Elie Wolfe, and Robert W. Spekkens. Fully quantum inflation: quantum marginal problem constraints in the service of causal inference, 2025

  10. [16]

    Uni- versal bound on the cardinality of local hidden variables in networks.Quantum Info

    DenisRosset, NicolasGisin, andElieWolfe. Uni- versal bound on the cardinality of local hidden variables in networks.Quantum Info. Comput., 18(11–12):910–926, September 2018

  11. [17]

    On the Logic of Causal Models.Proc

    Dan Geiger and Judea Pearl. On the Logic of Causal Models.Proc. 4th Conf. UAI, pages 136– –147, 1988

  12. [18]

    The lesson of causal discovery algorithms for quan- tum correlations: Causal explanations of Bell- inequality violations require fine-tuning

    Christopher Wood and Robert Spekkens. The lesson of causal discovery algorithms for quan- tum correlations: Causal explanations of Bell- inequality violations require fine-tuning. New Journal of Physics, 17, 08 2012

  13. [19]

    Instrumentality tests revisited

    Blai Bonet. Instrumentality tests revisited. In Proceedings of the Seventeenth Conference on 49 Uncertainty in Artificial Intelligence, UAI’01, page48–55, SanFrancisco, CA,USA,2001.Mor- gan Kaufmann Publishers Inc

  14. [20]

    On the Testability of Causal Mod- els with Latent and Instrumental Variables

    Judea Pearl. On the Testability of Causal Mod- els with Latent and Instrumental Variables. In Proceedings of the Eleventh Conference on Un- certainty in Artificial Intelligence, UAI’95, page 435–443, SanFrancisco, CA,USA,1995.Morgan Kaufmann Publishers Inc

  15. [21]

    Causal networks: Semantics and expressiveness

    Tom Verma and Judea Pearl. Causal networks: Semantics and expressiveness. Proceedings of the Fourth Workshop on Uncertainty in Artifi- cial Intelligence, 4, 03 2013

  16. [22]

    D. Geiger. Toward the formalization of informa- tional dependencies. 1988

  17. [23]

    Ayesha Ali, Thomas S

    R. Ayesha Ali, Thomas S. Richardson, Peter Spirtes, and Jiji Zhang. Towards characterizing markov equivalence classes for directed acyclic graphs with latent variables. InProceedings of the Twenty-First Conference on Uncertainty in Artificial Intelligence, UAI’05, page 10–17, ...

  18. [25]

    Spekkens

    Marina Maciel Ansanelli, Elie Wolfe, and Robert W. Spekkens. Everything that can be learnedaboutacausalstructurewithlatentvari- ables by observational and interventional prob- ing schemes, 2024

  19. [26]

    Causality: Models, Reasoning and Inference

    Judea Pearl. Causality: Models, Reasoning and Inference. Cambridge University Press, 2nd edi- tion, 2009

  20. [27]

    Andersson, David Madigan, and Michael D

    Steen A. Andersson, David Madigan, and Michael D. Perlman. A characterization of Markov equivalence classes for acyclic digraphs. The Annals of Statistics, 25(2):505 – 541, 1997

  21. [28]

    Optimal structure identification with greedy search

    David Maxwell Chickering. Optimal structure identification with greedy search. J. Mach. Learn. Res., 3(null):507–554, March 2003

  22. [29]

    Information- theoretic Inference of Common Ancestors.En- tropy, 17(4):2304–2327, 2015

    Bastian Steudel and Nihat Ay. Information- theoretic Inference of Common Ancestors.En- tropy, 17(4):2304–2327, 2015

  23. [30]

    Causation, prediction, and search

    Peter Spirtes, Clark Glymour, and Richard Scheines. Causation, prediction, and search. MIT press, 2001

  24. [31]

    Daley, Kevin J

    Patrick J. Daley, Kevin J. Resch, and Robert W. Spekkens. Experimentally adjudicating between different causal accounts of bell-inequality viola- tions via statistical model selection.Phys. Rev. A, 105:042220, Apr 2022

  25. [32]

    commenton’experimentally adjudicating between different causal accounts of bell-inequality violations via statistical model selection’

    Patrick Daley, Kevin J Resch, and Robert W Spekkens. Replyto"commenton’experimentally adjudicating between different causal accounts of bell-inequality violations via statistical model selection’". arXiv preprint arXiv:2412.02829, 2024

  26. [33]

    Robin J. Evans. Latent-free equivalent mDAGs. arXiv:2209.06534, 2022

  27. [34]

    Robin J. Evans. Graphical Methods for In- equality Constraints in Marginalized DAGs. 2012 IEEE International Workshop on Machine Learning for Signal Processing, pages 1–6, 2012

  28. [36]

    Robin J. Evans. Dependency in DAG models with Hidden Variables.arXiv:2106.07523, 2021

  29. [37]

    Ilya Shpitser, Robin Evans, and Thomas Richardson. Acyclic linear sems obey the nested markov property.Uncertainty in artificial intel- ligence : proceedings of the Conference on Un- certainty in Artificial Intelligence, 2018, 082018. 50

  30. [38]

    A general identifica- tion condition for causal effects

    Jin Tian and Judea Pearl. A general identifica- tion condition for causal effects. page 567–573, 2002

  31. [40]

    paD(Ci)∪ Ci ⊆ paD(d) for each d ∈ D, i = 1, ..., n

  32. [41]

    In this case, G observationally dominates G′, i.e., G⪰ G′

    If for a faceB′ ∈ B such that B′ /⊆ ∪i=1,...,nCi we have c ∈ B′ for some c ∈ ∪i=1,...,nCi, then D⊆ B′. In this case, G observationally dominates G′, i.e., G⪰ G′. Note that, since G′ structurally dominates G, by Lemma 3 we know thatG′ observationally dominates G, i.e., G′ ⪰ G. ...

  33. [42]

    This has to hold for both visible and latent parents

    paG1(C)∪ C ⊆ paG1(d) for eachd∈ D. This has to hold for both visible and latent parents. Then, G1 is observationally equivalent to the pDAG G3 defined by starting from G1, removing the nodes LCD and adding the new latent nodes{lC i }i=1,...,n and lD, whose children are respect...

  34. [43]

    We can see that G1 ⪰ G3 by structural domi- nance (Lemma 3)

    Fix the cardinalities of every latent variable to be k and the cardinality of each visible variable 53 x l1 CD y z t u v l2 CD d w G1 x l1 C y z t u v l2 C d w G2 lD x l1 C y z t u v l2 C d w G3 lD Figure A.2: The pDAGsG1 and G3 can be shown obser- vationally equivalent by Pro...

  35. [44]

    For each visible variable, enumerate all possible response functions it can have to the valuations of its parents. For example, if a visiblev has only one parent and this parent is latent, it has(cv)k possible response functions; we could have thatv reacts withtheeachoneofits ...

  36. [45]

    The question that is left is how to choosek, the car- dinality of latent variables

    For each possibility of response functions of all the visible variables ofG, we compute the sup- port: the set of visible events that occur under that response function, for some valuation of the latent variables. The question that is left is how to choosek, the car- dinality ...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.