{"id":"48f7913c-20b6-4601-90ee-9131b230a734","arxiv_id":"2606.00965","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper introduces Horvitz-Thompson estimators for edge-level causal effects under dyadic interference, a three-fold cross-fitting scheme to enable machine learning covariate adjustment, and a calibration step ensuring no asymptotic efficiency loss.","lead":"This paper develops estimators for causal effects defined on the directed connections between units in a network, where outcomes depend on treatment pairs and standard methods fail due to dependence. A smart generalist might read it to understand how to analyze interaction data in social, biological, or economic networks with valid uncertainty and machine learning adjustments.","discovery_kind":"unclear","skeptic_critique":{"model":"grok-4.3","headline":"Three-fold cross-fitting may leave residual dependence from shared units in directed dyadic networks","rationale":"The reader's weakest_assumption directly identifies the load-bearing technical step for the ML-adjusted estimator. All other claims (plain HT normality, variance bounds) can be verified independently of the splitting device; the efficiency gain with flexible ML cannot.","tokens_in":1774,"tokens_out":312,"duration_ms":16173,"concrete_test":"In the section deriving the three-fold cross-fit estimator, extract the precise partition rule and the statement that E[residual | fold] = 0; construct a small directed graph (n=9 nodes, all possible directed edges) and enumerate whether any edge can have both endpoints in the same fold under the rule; if any such edge exists, recompute the bias term for the ML-adjusted estimator on that graph.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central efficiency claim requires that the three-fold sample-splitting scheme restores the conditional independence needed for unbiased ML adjustment of the Horvitz–Thompson edge-level estimator. Standard two-fold splitting fails because an edge outcome depends on two units that can appear in both folds. The paper asserts that three folds suffice to break this dependence under the stated stability condition. If any pair of units connected by an edge can still share a fold after the three-way partition (possible in dense directed graphs), the cross-fit nuisance estimator remains correlated with the outcome, invalidating both unbiasedness and the subsequent asymptotic normality argument.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript develops design-based Horvitz-Thompson estimators for a general class of edge-level causal effects in directed networks under dyadic interference. It establishes asymptotic normality of these estimators under mild regularity conditions, constructs variance estimators that exploit identifiable components of the network dependence structure, and introduces a three-fold sample splitting and cross-fitting scheme to enable covariate adjustment (including via machine learning) while restoring the conditional independence needed for unbiased estimation. A calibration step is proposed to ensure the adjusted estimator incurs no asymptotic efficiency loss relative to the unadjusted version. The results are illustrated via simulations and a real-data application.","tokens_in":1921,"tokens_out":452,"duration_ms":25322,"significance":"If the central technical claims hold, particularly the validity of the three-fold splitting procedure and the stability condition, the work would advance causal inference methodology for network data by extending design-based estimation to edge-level outcomes with complex dependence and by providing a practical route to efficiency gains via flexible ML adjustment without asymptotic penalty. The explicit design-based foundation (avoiding outcome model assumptions) and the calibration device are notable strengths that could be useful in applied network settings.","major_comments":[{"comment":"The section describing the three-fold sample splitting and cross-fitting scheme (the paragraph addressing the failure of two-fold splitting due to shared units and the proposed solution): the claim that three folds suffice to restore the conditional independence between the nuisance estimator and the edge outcome for all pairs requires an explicit lemma or argument showing that, after the three-way partition, no two units connected by a directed edge can appear together in a fold used for nuisance estimation. Without this, the unbiasedness and subsequent asymptotic normality of the ML-adjusted estimator remain at risk in dense directed graphs, which is load-bearing for the efficiency claim.","section":"section on three-fold sample splitting"}],"minor_comments":[{"comment":"The abstract invokes a 'stability condition' for asymptotic normality of the covariate-adjusted estimator but does not state its precise form; moving a brief definition or reference to the relevant assumption into the abstract would improve readability.","section":"abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their careful reading and constructive feedback. We respond to the single major comment below.","responses":[{"response":"We appreciate the referee highlighting this point. The manuscript describes the three-fold scheme and states that it restores the conditional independence needed for unbiased estimation after noting the failure of two-fold splitting. We agree that an explicit lemma would make the argument more rigorous and directly address concerns in dense directed graphs. In the revision we will add a lemma proving that a random three-way partition of the units ensures that, for any directed edge (i,j), units i and j do not co-occur in any fold used for nuisance estimation. This will formally establish the required separation and support the unbiasedness and asymptotic normality of the ML-adjusted estimator. The addition will be placed in the relevant section and will not alter the existing results or proofs.","revision_made":"yes","referee_comment":"The section describing the three-fold sample splitting and cross-fitting scheme (the paragraph addressing the failure of two-fold splitting due to shared units and the proposed solution): the claim that three folds suffice to restore the conditional independence between the nuisance estimator and the edge outcome for all pairs requires an explicit lemma or argument showing that, after the three-way partition, no two units connected by a directed edge can appear together in a fold used for nuisance estimation. Without this, the unbiasedness and subsequent asymptotic normality of the ML-adjusted estimator remain at risk in dense directed graphs, which is load-bearing for the efficiency claim."}],"tokens_in":1402,"tokens_out":329,"duration_ms":22256,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper's main advance is the three-fold sample splitting and cross-fitting scheme that restores conditional independence for covariate-adjusted Horvitz-Thompson estimators of edge-level effects under dyadic interference. The calibration step that preserves asymptotic efficiency relative to the unadjusted estimator is a clean addition, and the variance estimators that use identifiable network dependence components give tighter bounds than classical ones.\n\nIt handles the practical problem that standard two-fold splitting fails when an edge outcome depends on two units that can appear across folds. The design-based framing with known assignment probabilities keeps things grounded, and the abstract indicates the asymptotic normality holds under mild conditions plus a stability requirement. Simulations and a real-data example are said to show efficiency gains.\n\nThe soft spot is the claim that three folds suffice to eliminate residual dependence from shared units. In denser directed networks it is possible for connected pairs to still overlap after partitioning, which would leave the nuisance estimator correlated with the outcome and undermine both unbiasedness and the normality result. How mild the stability condition actually is, and whether it covers realistic dense cases, is not obvious from the abstract.\n\nThis is for causal inference researchers working with network or dyadic data. The technical adaptation is real and the thinking is engaged with the design-based literature. It deserves peer review so the full derivations and conditions can be checked.","headline":"Three-fold splitting lets ML adjust edge-level network causal estimators without efficiency loss, but whether it fully breaks shared-unit dependence in directed graphs is the key open question.","tokens_in":2434,"tokens_out":344,"would_cite":false,"duration_ms":20515,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Horvitz-Thompson estimators with three-fold splitting support valid causal inference for outcomes defined on network edges.","keywords":["causal inference","network data","edge-level outcomes","Horvitz-Thompson estimator","sample splitting","machine learning","dyadic interference","covariate adjustment"],"falsifier":"A Monte Carlo experiment in which the three-fold cross-fitting procedure is applied to data with known edge-level effects yet the covariate-adjusted estimator exhibits persistent finite-sample bias that the calibration step does not remove would falsify the independence-restoration claim.","tokens_in":2683,"feed_emoji":"📊","tokens_out":731,"duration_ms":25038,"temperature":0.7,"pith_summary":"This paper develops design-based methods for estimating causal effects when the outcomes sit on the directed edges of a network rather than on the nodes. Edge outcomes depend on the joint treatment assignments of pairs of units, creating dependence that invalidates standard node-level procedures. The authors construct Horvitz-Thompson estimators for a broad class of such effects, prove asymptotic normality under mild conditions, and supply variance estimators that use identifiable network dependence to produce tighter bounds. They further show how to adjust for covariates with linear or machine-learning methods by replacing the usual two-fold sample split with a three-fold scheme that restores the conditional independence needed for unbiased estimation, then add a calibration step that prevents any asymptotic efficiency loss relative to the unadjusted estimator.","feed_headline":"Three-fold splitting enables ML adjustment for edge causal effects","feed_subtitle":"Standard two-fold methods fail due to shared-unit dependence in networks, but the new scheme supports unbiased, asymptotically normal estima","key_machinery":"Horvitz-Thompson estimators for edge-level causal effects paired with three-fold sample splitting and cross-fitting for covariate adjustment","core_discovery":"We construct Horvitz-Thompson estimators for a general class of edge-level causal effects in directed networks under dyadic interference and establish their asymptotic normality under mild regularity conditions. We develop variance estimators that exploit identifiable components of network dependence. To improve efficiency we introduce a three-fold sample splitting and cross-fitting procedure that restores the conditional independence required for unbiased covariate-adjusted estimation; under a stability condition the resulting estimator is asymptotically normal for both linear and flexible machine-learning adjustments, and a calibration step guarantees no asymptotic efficiency loss relative","pith_inferences":["The three-fold splitting logic may generalize to other settings with overlapping units, such as spatial or temporal data with local dependence.","Analysts evaluating interventions that affect pairwise interactions in social or biological networks could obtain more precise estimates than node-level methods allow.","If the stability condition holds in practice, the procedure offers a template for incorporating flexible machine-learning adjustments into other design-based network estimators without efficiency penalties."],"forward_implications":["The estimators are asymptotically normal under mild regularity conditions.","Variance estimators that use network dependence components produce substantially less conservative bounds than classical approaches.","Covariate-adjusted estimators remain asymptotically normal when linear or machine-learning methods are used under the stated stability condition.","The calibration step ensures the adjusted estimator loses no asymptotic efficiency relative to the unadjusted version.","Simulation studies and a real-data application demonstrate the efficiency gains."],"fun_headline_variants":["Three-fold splitting for unbiased edge-level causal inference with ML","Covariate adjustment via three-fold splitting in directed network edges","Horvitz-Thompson edge estimators with three-fold cross-fitting","Asymptotically normal estimators for edge effects under dyadic interference"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The three-fold sample splitting scheme restores the conditional independence required for unbiased estimation of the covariate-adjusted estimator.","fun_headline_variants_meta":{"raw":{"variants":["Three-fold splitting for unbiased edge-level causal inference with ML","Covariate adjustment via three-fold splitting in directed network edges","Horvitz-Thompson edge estimators with three-fold cross-fitting","Asymptotically normal estimators for edge effects under dyadic interference"]},"model":"grok-4.3","cost_usd":0.007449,"raw_usage":{"total_tokens":3434,"prompt_tokens":694,"num_sources_used":0,"completion_tokens":68,"cost_in_usd_ticks":74487000,"prompt_tokens_details":{"text_tokens":694,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2672,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":694,"tokens_out":68,"duration_ms":17372,"temperature":1.0,"reasoning_tokens":2672,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T17:02:24.342054+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A Monte Carlo experiment in which the three-fold cross-fitting procedure is applied to data with known edge-level effects yet the covariate-adjusted estimator exhibits persistent finite-sample bias that the calibration step does not remove would falsify the independence-restoration claim.","supporting_citations":[],"review_version":1}