{"id":"1e93676a-6d0a-4405-8334-d39ea19add7d","arxiv_id":"2507.22747","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Joint exogeneity, a foundational assumption of potential-outcome causal models, is shown to be inconsistent with quantum correlations in a Bell-type experiment.","lead":"This paper argues that a standard causal-inference assumption, joint exogeneity, fails in quantum experiments. It builds a Bell-type example where the assumption would force a causal effect size, but quantum predictions contradict that bound, suggesting the assumption is false in quantum regimes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reductio rests on an unverified LP whose printed target does not match the ACE defined in Eq. (6); the reported bound 0.1339 must be independently reproduced.","rationale":"The manuscript's central claim is a reductio: potential outcomes plus joint exogeneity imply Balke–Pearl bounds; the quantum distribution violates those bounds while the true causal effect is 0. For this to work, Proposition 1 must be correct. The paper gives only a sketch of the LP argument, and the target function printed in Eq. (11) does not match the ACE defined in Eq. (6), so the numerical bound 0.1339 is currently unsupported by the text. This is a concrete, fixable issue, but it blocks verification of the main logical bridge. I partly agree with the reader that the reset-protocol definition of potential outcomes in Eq. (2) is the deeper modeling assumption; however, I would not call it an error, because the protocol is explicit and physically implementable. The more immediately load-bearing problem is that the quantitative consequence of that definition—the LP bound—is asserted rather than demonstrated and appears to contain a notational slip. The true ACE=0 claim also relies on Chaves et al., but it can be re-derived from Eq. (2) in the same check. No fatal internal inconsistency is apparent, and the thought-experiment framing is novel. The correct response is to require a completed derivation and a reproducible LP, which is exactly a conditional verdict. Hence I recommend keeping the reader's CONDITIONAL verdict.","tokens_in":8113,"tokens_out":20198,"duration_ms":248291,"concrete_test":"Implement the LP in Eq. (11) with the objective corrected to the ACE in Eq. (6): minimize over q_{x,y00,y01,y10,y11|z} the quantity Σ_{x,y00,y01,y11} q_{x,y00,y01,1,y11|0} − Σ_{x,y01,y10,y11} q_{x,1,y01,y10,y11|0}, subject to normalization, joint exogeneity, stratified exclusion, and the observation constraints using the probabilities in Eq. (13). Verify that the optimum equals ≈0.1339. In parallel, compute P(Y(1,0)=1) and P(Y(0,0)=1) directly from Eq. (2) to confirm the true ACE is 0. If the corrected LP optimum differs, or the objective as printed is infeasible, Proposition 1 and the central contradiction fail.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 1 is the bridge from joint exogeneity plus potential outcomes to the contradiction, but its proof is only sketched: the LP in Eq. (11) is said to be handled analytically and to yield the Balke–Pearl bound, with no derivation or code. As printed, the LP target is inconsistent with Eq. (6): the first sum runs over q_{x,y00,1,y10,y11|0}, i.e. it fixes y01=1 rather than y10=1, so the objective is P(Y(0,1)=1)−P(Y(0,0)=1), which is identically zero under the stratified exclusion restriction (3), not ACE_X→Y = P(Y(1,0)=1)−P(Y(0,0)=1). The reported lower bound ≈0.1339 therefore cannot be checked against the claimed target. A second load-bearing premise is the reset-protocol definition of potential outcomes via Eq. (2); although internally coherent, it is a modeling choice, and the statement that joint exogeneity is falsified is no stronger than that choice. The true ACE=0 also depends on an external result, though it is consistent with Eq. (2) and the Bell-state reduced density matrix.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript claims to exhibit the first randomized experiment violating joint exogeneity. It defines potential outcomes Y(x,z) through a reset protocol on a bipartite quantum system, derives the stratified exclusion restriction, and argues that assuming joint exogeneity together with the existence of these potential outcomes yields the Balke-Pearl bounds on the average causal effect. It then specifies a Bell state and measurement settings whose observed conditional distribution violates those bounds while the true ACE is claimed to be 0, and concludes that joint exogeneity is falsified in quantum regimes. The paper also discusses implications for the fatalism view of potential outcomes and for the realism-locality debate.","tokens_in":8347,"tokens_out":14134,"duration_ms":147457,"significance":"The central idea is original and thought-provoking: it connects a foundational assumption in causal inference to quantum violations of instrumental inequalities and does so with a concrete, implementable quantum configuration and no fitted free parameters. If the derivation is made fully valid, the paper would provide the first concrete counterexample to joint exogeneity, lending empirical support to Dawid's critique and sharpening the realism-locality discussion in potential-outcome modelling. The significance is currently tempered by the fact that the load-bearing linear program is not correctly or completely transcribed, the proof of the key proposition is only asserted, and the scope of the conclusion depends on a nonstandard reset-protocol definition of potential outcomes.","major_comments":[{"comment":"The linear program as printed cannot be checked and does not match the stated target. In the objective, the first sum is written over x,y10,y01,y11 for q_{x,y00,y01,1,y11|0}, which leaves y00 unsummed and attempts to sum over y10 although that index is fixed to 1; the correct first term for the ACE in Eq. (6) is sum over x,y00,y01,y11 of q_{x,y00,y01,1,y11|0}. In addition, the two displayed stratified-exclusion constraints only enforce equality for the cases (x=0,y=0) and (x=1,y=1); the cases (x=0,y=1) and (x=1,y=0) are missing. Finally, the non-negativity constraint omits the index y11 and is written as q_{x,y00,y01,y10|z}. Without a corrected LP, the claimed equivalence to the Balke-Pearl bound and the numerical lower bound 0.1339 are unverifiable.","section":"Section 3, Eq. (11)"},{"comment":"The proof of Proposition 1 is not provided. The sentence 'This linear programming can be handled analytically, which yields the same bounds as the so-called Balke-Pearl bound' is an assertion rather than a derivation; no dual solution, vertex enumeration, or reproducible code is supplied. Since the contradiction in Section 2 depends entirely on the numerical lower bound 0.1339, the authors should provide either a complete analytical derivation of the Balke-Pearl equivalence or a reproducible computation and an optimality certificate for the LP.","section":"Section 3, Proposition 1"},{"comment":"The statement that 'the true value of ACE in this case is 0' is delegated to Chaves et al. without reproduction, and the definition of ACE used there may not coincide with the reset-protocol definition in this paper. The paper can compute this quantity directly from its own definitions: for rho = |Phi-><Phi-|, the reduced state on Y is the maximally mixed state I/2, so P(Y(x,z)=1) = 1/2 for all x,z and the ACE is 0. This computation should be included to make the contradiction self-contained and to fix the target quantity.","section":"Section 2, paragraph after Eq. (13)"},{"comment":"The conclusion that the experiment 'falsifies' joint exogeneity is only as strong as the modeling choice that the reset protocol in Eq. (2) correctly defines the potential outcomes Y(x,z) in a quantum system. A reader who does not accept this counterfactual interpretation could instead take the contradiction as showing that the potential outcome framework, or the stratified exclusion restriction derived from Eq. (2), is inapplicable to this quantum experiment. The authors should state the theorem explicitly as conditional on the reset-protocol definition and discuss this limitation, rather than claiming unconditional falsification of joint exogeneity.","section":"Abstract and Section 4"}],"minor_comments":[{"comment":"In the consistency equation, the third term is printed as X(1-Z)Y(0,1); it should be X(1-Z)Y(1,0).","section":"Section 3, Eq. (4)"},{"comment":"The phrase 'where the last inequality is directly from Assumption 1' should read 'last equality', since the displayed relation is an equality.","section":"Section 3, Eq. (6)"},{"comment":"There are several typographical errors, including 'Prelimiary' in the header, 'practial' in the abstract, 'attidude' in the introduction, and 'distribtuion' in Section 2; these should be corrected in a final version.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper has a genuinely interesting core idea, and the defects I found are fixable in principle: correcting the LP transcription, supplying the missing proof or code, and deriving the true ACE directly. I would not recommend rejection at this stage, but the current draft is not verifiable as written. If the LP cannot be repaired or the claimed bound 0.1339 cannot be reproduced, the central argument would collapse, so the revision should be judged on the completeness of those technical details."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Take this as a preliminary but interesting contribution. It is a thought experiment, not an experiment: no lab data, just a concrete Bell state and measurements. The novel step is to define potential outcomes via a reset protocol that yields a 'stratified exclusion restriction,' then show that adding joint exogeneity forces Balke-Pearl bounds which the quantum correlations violate. This cleanly isolates joint exogeneity as the assumption that fails, and it sharpens the realism-locality debate in the way the authors claim. The philosophical discussion is honest and well-connected to Dawid, Gill, and Robins et al.\n\nThe main soft spots are presentation and verification. Proposition 1 is the load-bearing step, but the LP in Eq. (11) is asserted to be analyzable without showing the solution, dual, or vertex. The reported bound 0.1339 is just stated. The notation in Eq. (11) is also garbled—summation indices don't match the subscripts—so a reader cannot verify the target. I don't think the stress-test's specific claim that the objective fixes y01=1 is right; the intended target is clearly P(Y(1,0)=1)-P(Y(0,0)=1). But even so, the whole derivation should be made checkable, e.g., with code or a formal LP solution. The 'true ACE=0' is also outsourced to Chaves et al.; a direct calculation from Eq. (2) would make the paper self-contained.\n\nThe bigger conceptual issue is the reset protocol in Eq. (2). It is a modeling choice, not a forced consequence of quantum mechanics. The paper acknowledges this implicitly but the abstract and title overclaim: calling it a 'quantum experiment' and saying joint exogeneity is falsified without qualifying that this is within their operational definition. With that qualification, the reductio is valid and the result is genuinely new in framing—Chaves et al. used a stronger individual exclusion restriction, and here the stratified version suffices.\n\nThe paper is not ready as is, but the core idea deserves serious referee time. I would ask for a completed proof of Proposition 1 (or a verifiable computational certificate), a direct computation of ACE=0, a clean rewrite of Eq. (11), and a softer title/abstract. For the right reader—quantum causal inference, potential outcomes, Bell inequalities—this is worth engaging with.","headline":"A thought experiment that credibly isolates joint exogeneity as the falsified assumption in a Bell-type quantum setup, but the proof is not yet reproducible and the title overclaims an experiment.","tokens_in":8858,"tokens_out":4805,"would_cite":true,"duration_ms":53024,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A Bell-type quantum experiment falsifies joint exogeneity: assuming potential outcomes exist yields Balke-Pearl bounds, and the observed correlations violate them while the true causal effect is zero.","keywords":["potential outcomes","joint exogeneity","Balke-Pearl bounds","quantum causal inference","Bell correlations","instrumental variables","stratified exclusion restriction","fatalism"],"falsifier":"A reader could settle the matter by building the photonic experiment described by Eqs. (12) and (13): if observed correlations match Eq. (13) and the interventional protocol yields a true effect of 0 while the linear-program bound is about 0.1339, the violation is real; if an alternative, independently motivated interventional semantics yields an effect inside the bounds, the claim is not.","tokens_in":7860,"feed_emoji":"⚛️","tokens_out":9567,"duration_ms":94224,"temperature":0.7,"pith_summary":"The paper aims to show that joint exogeneity—the assumption that treatment assignment is statistically independent of the entire vector of potential outcomes—is not merely philosophically contestable but empirically false. Working in a Bell-type quantum network where a binary instrument Z controls which measurement is made on an entangled state, the authors argue that merely postulating potential outcomes forces a 'stratified exclusion restriction' on their joint distribution. Adding joint exogeneity then makes the average causal effect identifiable only up to the Balke-Pearl bounds. The measured quantum correlations fall outside those bounds: the true causal effect computed from the interventional protocol is 0, while the bounds demand a lower bound of about 0.134. The conclusion is that, in quantum regimes, any potential-outcome model accepting the paper's interventional semantics must give up joint exogeneity, undercutting the fatalist view that outcomes pre-exist the treatment assignment.","feed_headline":"Bell experiment falsifies a core causal assumption","feed_subtitle":"Quantum correlations violate Balke-Pearl bounds, so joint exogeneity fails even though the true causal effect is zero.","key_machinery":"The central object is the interventional distribution $Q_{xz}$ defined by a thought experiment: generate $(X,Z)$ as in the network, then reset them to fixed values $(x,z)$ before measuring $Y$. In quantum theory it satisfies $Q_{xz}(X=x',Y=y | Z=z') = tr[(M_x^{z'} \\otimes N_y^x)\\rho]$, which yields the stratified exclusion restriction. This restriction replaces the usual individual exclusion restriction, so the only added assumption beyond existence of potential outcomes is joint exogeneity. The argument's second piece is a linear program over the 64-dimensional conditional distribution of $(X, Y(0,0),...,Y(1,1))$ given $Z$, whose feasible region under joint exogeneity and the stratified exclusion restriction reproduces the Balke-Pearl bounds. Feeding the Bell-state probabilities from Eq. (13) into this program produces a lower bound of about 0.1339 for the average causal effect, whereas the interventional protocol gives 0.","core_discovery":"The discovery is that joint exogeneity can be violated in a realizable quantum experiment. In the network of Figure 2a, Z is a randomized instrument, X is a measurement whose setting depends on Z, and Y is a later measurement whose setting depends on X. Assuming potential outcomes $Y(x,z)$ exist, the reset protocol of Eq. (2) yields the stratified exclusion restriction $P(Y(x,1)=y | X=x', Z=z') = P(Y(x,0)=y | X=x', Z=z')$ for all $x,x',y,z'$. Under joint exogeneity, the average causal effect $E(Y(1,z)-Y(0,z))$ is bounded exactly by the Balke-Pearl bounds. Yet for the Bell state and the four Pauli-type measurements in Eq. (12), the observed distribution in Eq. (13) gives a Balke-Pearl lower bound around 0.1339 while the direct interventional computation gives a true effect of 0. The authors read this contradiction as falsifying joint exogeneity, with the stratified exclusion restriction derived rather than assumed.","pith_inferences":["A testable extension: a photonic realization of Eqs. (12)-(13) would let an experimenter directly verify the interventional effect of 0; observing the predicted violation in a real device would make the falsification of joint exogeneity an experimental result rather than a calculation.","If the violation is taken at face value, then future causal analyses involving quantum devices, such as entangled communication networks, should either test joint exogeneity or replace potential outcomes with quantum-compatible interventions; the paper leaves the latter framework unspecified.","The stratified exclusion restriction is derived from the reset protocol, so the same Bell correlations could be reinterpreted as a violation of that restriction if one adopts a different semantics for 'setting X and Z'; the argument thus locates the contradiction specifically at the choice of interventional meaning.","By showing realism fails without any appeal to locality, the result dissolves the realism-locality dilemma in potential-outcome modeling: keeping locality no longer protects realism."],"forward_implications":["In any experiment where quantum correlations are non-negligible, invoking joint exogeneity together with the existence of potential outcomes can produce bounds that contradict the true effect, so the assumption cannot be treated as a free background condition.","Prior causal conclusions that rely on joint exogeneity in classical settings are not overturned, because the violation appears precisely where a classical description fails.","The result pins down the average causal effect in the Bell configuration as zero, while the Balke-Pearl lower bound from the observed distribution is about 0.1339, so the observed correlations are the empirical witness of the violation.","The authors' reading shifts the philosophical default: potential outcomes should be thought of as generated at or after treatment assignment, with a joint distribution that may depend on the experimenter's choice, rather than as pre-existing attributes."],"supporting_citations":[{"why":"Supplies the linear-programming method and the bounds on average treatment effects that the quantum distribution is shown to violate.","marker":"[Balke and Pearl, 1997]"},{"why":"Provides the quantum instrumental-test violation and the calculation that the true causal effect in this Bell-state configuration is zero.","marker":"[Chaves et al., 2018]"},{"why":"Defines the potential outcomes framework and the standard exogeneity assumptions whose consistency is under test.","marker":"[Imbens and Rubin, 2015]"},{"why":"Gives the 'fatalism' critique that the paper's violation empirically supports.","marker":"[Dawid, 2000]"},{"why":"Argues for abandoning realism while preserving locality, a recommendation this paper's no-locality-needed violation reinforces.","marker":"[Gill, 2014]"},{"why":"Proves the realism-locality tension in Bell experiments and recommends dropping realism; the paper extends this to show realism fails even without locality.","marker":"[Robins et al., 2015]"}],"fun_headline_variants":["Quantum experiment falsifies joint exogeneity","Bell test shows causal assumption fails","Causal assumption violated in quantum setup","Joint exogeneity refuted by Bell correlations","Quantum correlations break causal assumption"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument stands or falls with the paper's choice to define the potential outcome distribution by the reset protocol of Eq. (2)—that setting X and Z in the quantum system means re-running the measurement with predetermined values—so if that semantics is not the right formalization, the contradiction with joint exogeneity does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Quantum experiment falsifies joint exogeneity","Bell test shows causal assumption fails","Causal assumption violated in quantum setup","Joint exogeneity refuted by Bell correlations","Quantum correlations break causal assumption"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0002,"raw_usage":{"total_tokens":1323,"prompt_tokens":840,"completion_tokens":483,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":456,"completion_tokens_details":{"reasoning_tokens":424}},"tokens_in":456,"tokens_out":483,"duration_ms":6210,"temperature":1.0,"reasoning_tokens":424,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T11:21:22.459336+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could settle the matter by building the photonic experiment described by Eqs. (12) and (13): if observed correlations match Eq. (13) and the interventional protocol yields a true effect of 0 while the linear-program bound is about 0.1339, the violation is real; if an alternative, independently motivated interventional semantics yields an effect inside the bounds, the claim is not.","supporting_citations":[{"cited_title":"Bounds on treatment effects from studies with imperfect compliance","cited_arxiv_id":null,"evidence_quote":"Supplies the linear-programming method and the bounds on average treatment effects that the quantum distribution is shown to violate."},{"cited_title":"Quantum violation of an instrumental test","cited_arxiv_id":null,"evidence_quote":"Provides the quantum instrumental-test violation and the calculation that the true causal effect in this Bell-state configuration is zero."},{"cited_title":"Causal inference without counterfactuals","cited_arxiv_id":null,"evidence_quote":"Gives the 'fatalism' critique that the paper's violation empirically supports."},{"cited_title":"Statistics, causality and bell’s theorem","cited_arxiv_id":null,"evidence_quote":"Argues for abandoning realism while preserving locality, a recommendation this paper's no-locality-needed violation reinforces."},{"cited_title":"A proof of bell's inequality in quantum mechanics using causal interactions","cited_arxiv_id":null,"evidence_quote":"Proves the realism-locality tension in Bell experiments and recommends dropping realism; the paper extends this to show realism fails even without locality."}],"review_version":1}