{"id":"aa11d357-d950-49a8-9469-c32d41f14e86","arxiv_id":"1909.00847","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Helmholtz-Hodge decomposition of a network built from 1,700 smart sanctions lists places OECD and ICTR at the top and the UN Security Council near the bottom for Iran-North Korea sanctions.","lead":"This paper applies a network-decomposition method to 1,700 smart sanctions lists and claims to expose a hidden hierarchy of influence among the 85 countries and organizations that issue them. A generalist might read it as an example of quantifying soft power through public regulatory lists.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The temporal-precedence-to-influence edge construction is unvalidated, and the paper's own text states that data for open use are still being recollected, so the core inference cannot currently be checked.","rationale":"The reader's weakest assumption and my load-bearing concern coincide: the mapping from temporal order of first appearance on sanctions lists to directed influence is unvalidated, and the paper's own introduction acknowledges plausible non-influence explanations for the same pattern. My independent reading adds specificity: the same-date exclusion rule can swing specific findings such as the Iran-North Korea hierarchy; the reported gradient/loop ratios themselves imply substantial circular flow, which is evidence that simple hierarchical copying does not fully describe the system; and the proprietary data, explicitly noted as still being recollected for open use, prevent any independent check of the edge construction. I find no internal mathematical error in the Helmholtz-Hodge decomposition step, and the methodology is a standard, legitimate application. The core issue is external validity of the network construction, which is exactly the kind of correctable concern that warrants a CONDITIONAL rather than REJECT or ACCEPT verdict. I agree with the reader on the weakest assumption, and I endorse the recommended verdict without change.","tokens_in":8419,"tokens_out":1328,"duration_ms":12080,"concrete_test":"Obtain the raw list-entry data (entity, institution, first-listing date) and run a placebo test: shuffle first-listing dates for entities within each category while preserving the number of edges per institution, recompute the rank correlation of HHD potentials with the original, and compare the observed potential ranking to the distribution under the null. Also compute the potential ranking after reversing the direction of a random sample of edges and after dropping same-date events; if the headline findings such as UNSC-at-bottom for Iran-North Korea do not survive, the temporal-precedence assumption is not load-bearing for the conclusions.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central construction defines an influence edge from institution A to B whenever B first lists an entity after A first listed it (Section 2). This equates temporal precedence with causal influence. The paper itself motivates alternative mechanisms: independent intelligence, shared third-party sources, or copying to project international cooperation (Introduction). Under any of these, the edge set is not a faithful influence relation, and the Helmholtz-Hodge potentials inherit the misspecification. The paper provides no external validation, no robustness check against a null model, and no sensitivity analysis for the same-date exclusion rule. Additionally, the data are proprietary Dow Jones lists, and footnote 4 states only that the authors are 'recollecting the data for open use,' so the edge construction cannot be independently reproduced or tested as published. The 'UN Security Council at the bottom' finding for Iran-North Korea is particularly dependent on this assumption and on the unresolved same-date rule, and Table 2 shows material loop flow (non-negligible circular structure), which is consistent with non-hierarchical mechanisms such as shared external drivers, undermining the claim that the gradient component reflects true influence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes 1,700 smart-sanctions lists from 85 institutions (73 countries and 12 international organizations) and constructs two directed networks: one at the level of individual sanctions lists and one at the level of institutions. An edge is placed from A to B when institution/list B includes an entity that A listed earlier; cases with identical dates are ignored. The paper then applies Helmholtz–Hodge decomposition (HHD) to these networks, extracting a gradient component with node potentials and a loop component. The authors report potentials for institutions overall and for seven sanction categories, and interpret the potential ordering as a hierarchy of influence: for example, OECD and ICTR are at the top in the overall analysis, Switzerland appears as a consistently influential player, and the UN Security Council is unusually low in the Iran–North Korea category. The main claim is that HHD of this network yields meaningful insights into the hierarchical influence structure behind smart sanctions.","tokens_in":8718,"tokens_out":4404,"duration_ms":47136,"significance":"If the temporal-precedence edges were a valid proxy for influence, this would be a novel and interesting application of HHD to international political data, with the potential to inform political science and policy analysis. The mathematical framework is standard, the decomposition is correctly described, and the paper is honest about data limitations (footnote 4). However, the central interpretation rests on an unvalidated assumption that listing an entity earlier causes later listing of the same entity by another institution. The paper does not provide null models, significance tests, external validation, or uncertainty quantification for the potentials, and the data are not currently available for independent replication. These gaps currently limit the strength of the causal claims and the generalizability of the results.","major_comments":[{"comment":"The load-bearing assumption is that a directed influence edge from A to B exists whenever B first lists an entity after A first listed it. This equates temporal precedence with causal influence, but the Introduction itself acknowledges alternative mechanisms: institutions might act independently, follow a shared third-party source, or copy lists to project international cooperation. Under any of these alternatives, the edge set is not a faithful influence relation and the HHD potentials inherit the misspecification. To support the central claim, the authors should validate this edge definition against external evidence (e.g., case studies or official documents), test robustness against null models such as random date permutations or configuration models that preserve list overlap, and analyze sensitivity to the same-date exclusion rule. Without such checks, the interpretation of the potentials as influence is unsupported.","section":"Section 2 (network construction)"},{"comment":"The potentials and gradient/loop ratios are reported as point estimates with no uncertainty quantification or statistical significance tests. For example, the claim that the UN Security Council is at the bottom of the Iran–North Korea hierarchy (Fig. 8b and Section 5) could be within noise, especially given the small number of entities in that category. A bootstrap over listed entities or a permutation test that randomly shuffles list dates would provide confidence intervals for the potentials and p-values for the observed ordering. Similarly, the gradient/loop ratios in Table 2 have no error bars; the non-negligible loop ratios (0.11–0.21) are consistent with non-hierarchical mechanisms such as shared external drivers, which undermines the interpretation of the gradient component alone as representing true influence.","section":"Section 4.1, Table 3, and Table 2"},{"comment":"The analysis excludes two of the nine detected categories (A: Japanese bureaucracy, and E: embargoes) because 'only a few countries were involved in the country-level network.' This post hoc selection is not accompanied by a principled criterion or threshold, and it weakens the generality of the claim that the analysis covers smart sanctions broadly. The authors should either include all categories with appropriate caveats, or provide a prespecified exclusion rule and show that the main findings are robust to alternative inclusion criteria. The current presentation leaves the impression that categories were selected to produce more interpretable results.","section":"Section 4.2"},{"comment":"The data are proprietary Dow Jones datasets, and footnote 4 states that the authors are 'recollecting the data for open use.' As published, the core network construction cannot be independently reproduced or tested, which is especially problematic because the edge definition and the resulting HHD potentials are the entire basis of the paper's claims. The authors should make the data available in some accessible form (e.g., a curated list of institution-entity-date triples for the analyzed lists) or provide a detailed replication script and summary statistics so that reviewers and readers can check the construction and run the suggested null-model analyses.","section":"Footnote 4"}],"minor_comments":[{"comment":"In the text, 'Table 1b summarizes the characteristics of each community' appears to refer to a panel in Fig. 1; the reference should be clarified.","section":"Section 2, Fig. 1"},{"comment":"The label 'Word Bank' is a typo and should read 'World Bank.'","section":"Fig. 2"},{"comment":"The sentence 'The arrow shows the location of where the y-axis being 0' is ungrammatical and should be revised, e.g., to 'The arrow indicates where the y-axis value is zero.'","section":"Section 4.1"},{"comment":"The phrase 'we could shed some light' should be 'we can shed some light' for consistency with the rest of the paper.","section":"Section 5"},{"comment":"The scatterplot comparing PageRank and HHD potential would benefit from reporting the Pearson or Spearman correlation coefficient and a regression line to make the claim of independence quantitative.","section":"Fig. 4a"},{"comment":"The HHD presentation would be clearer if the authors stated the exact least-squares objective and the normalization convention for the potentials (e.g., mean zero), and explained how disconnected nodes are treated in the decomposition.","section":"Section 3"}],"recommendation":"major_revision","confidential_remarks":"This is an interesting application paper with a standard method, but the central inference from temporal precedence to influence is not validated, and the data are not yet publicly available. I believe the paper can be made publishable if the authors add rigorous null-model and sensitivity analyses, provide uncertainty quantification, and address the post hoc category exclusion. If the data cannot be shared or the validation cannot be performed, the claims should be substantially toned down to describing the hierarchical structure of temporal dependencies rather than influence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this paper applies Helmholtz-Hodge decomposition to a newly constructed influence network of smart-sanctions issuers. The math is standard and correctly applied; the result is a new empirical ranking of institutions (OECD and ICTR at the top, UN Security Council near the bottom for Iran-North Korea). If you work on computational political science or sanctions, it's worth a look.\n\nWhat's genuinely new is the dataset construction: 1,700 lists from Dow Jones, aggregated into directed edges by temporal precedence of listing the same entity. That is a legitimate new object of study. The HHD itself is not new (Jiang et al. 2011), but applying it here is reasonable, and the authors are careful to separate gradient from loop flow.\n\nThe soft spot is load-bearing: the assertion that if B lists an entity after A, then A influences B. The paper's own introduction lists alternative explanations—independent intelligence, shared third-party sources, or copying to project cooperation. None of those are ruled out. There are no null models, no significance tests, no external validation. The loop ratios in Table 2 are non-negligible (0.11 to 0.21), which is consistent with non-hierarchical mechanisms. And the data are proprietary; footnote 4 says the authors are 'recollecting the data for open use,' so as published the network cannot be independently reproduced.\n\nThe post-hoc exclusion of categories A and E in Section 4.2 is minor but should be better justified. The same-date exclusion rule deserves a sensitivity analysis.\n\nNone of this kills the paper. The derivation is self-contained, not circular, and the results are interesting enough to prompt follow-up. But the interpretation of the potentials as 'influence' needs external grounding or a serious robustness section.\n\nWho is this for? Political scientists studying sanctions, and network analysts looking for an applied example. I would give it a serious referee, with the referee pushing for data release and validation.\n\nBottom line: worth engaging, but as an invitation to further work, not a settled result.","headline":"First HHD application to a new sanctions-list influence network; interesting but the temporal-precedence-to-influence assumption is unvalidated and the data are not yet available.","tokens_in":9114,"tokens_out":1394,"would_cite":true,"duration_ms":14939,"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":"This paper claims that applying Helmholtz–Hodge decomposition to a network built from the order in which institutions add the same targets to smart sanctions lists exposes the hierarchical influence structure behind those lists, placing…","keywords":["smart sanctions","Helmholtz-Hodge decomposition","influence network","network analysis","sanctions lists","hierarchical structure","temporal precedence","international politics"],"falsifier":"Shuffle the listing dates within each sanctions category while preserving the content of each institution's lists, recompute the Helmholtz–Hodge potentials on many shuffles, and compare the resulting rankings with the reported ones; if the upstream/downstream ordering survives random date reassignment, then list timing is not what the hierarchy measures and the paper's interpretation would fail that test.","tokens_in":8213,"feed_emoji":"🌐","tokens_out":9480,"duration_ms":88267,"temperature":0.7,"pith_summary":"Smart sanctions lists—who is barred from doing business with whom—are issued by many countries and organizations, and the paper's idea is that the order in which institutions add the same named entity to their lists carries information about who is influencing whom. From 1,700 lists issued by 85 institutions, the authors construct a directed network in which an edge points from the institution that listed an entity first to the one that listed it later, then apply Helmholtz–Hodge decomposition to extract a scalar potential for each institution. The paper argues that this potential is a meaningful ranking of upstream initiators versus downstream followers, and that looking within issue-specific categories produces interpretable hierarchies—for example, the UN Security Council sits near the bottom for Iran–North Korea sanctions while some national and intergovernmental bodies sit near the top. The payoff, if the claim holds, is a data-driven window into the hidden political power structure behind international sanctions, independent of confidential diplomacy.","feed_headline":"Who lists first exposes who really drives smart sanctions","feed_subtitle":"Helmholtz-Hodge scoring of 1,700 sanctions lists puts the UN Security Council downstream on Iran-North Korea.","key_machinery":"The load-bearing object is the Helmholtz–Hodge decomposition of a weighted directed network: it writes the observed flow $F_{ij}$ along each edge as a gradient flow $w_{ij}(\\phi_i-\\phi_j)$ plus a circular loop flow, choosing the node potentials $\\phi_i$ to minimize the squared difference, so $\\phi_i$ becomes the hierarchy score of institution $i$. The input network is the paper's own construction: for each pair of institutions, whenever one institution lists an entity that another institution listed earlier, one unit of weight is added to the directed edge from the earlier to the later lister, and same-day listings are discarded because the direction is unclear. Modularity-based community detection then defines the issue categories (financial crimes, Libya, Africa, Burma, terrorism in general, Al-Qaeda, Iran–North Korea) used to build category-specific networks, and the gradient/loop ratio summarizes how much of the flow is hierarchical rather than cyclic.","core_discovery":"The central claim is that a Helmholtz–Hodge potential computed from the who-lists-first network recovers a real hierarchy of influence among sanctions issuers, not just list similarity or conventional centrality. In the network built from all 1,700 lists, the OECD and the International Criminal Tribunal for Rwanda receive the highest potentials, which the authors read as meaning these institutions are least influenced by others, while the G7 countries sit in the middle of the ranking. The category-specific networks shift the ranking in ways the paper argues are interpretable: Switzerland has positive potential in nearly every category, Australia is mostly a follower, the United States ranks high against Al-Qaeda, Libya, Africa, and terrorism in general but low for financial crimes, and the UN Security Council sits at the bottom for Iran–North Korea sanctions, a position the authors tie to that body's complex approval mechanism. The paper presents the gradient-to-loop ratio as a companion measure of how hierarchical versus reciprocally circular each subnetwork is.","pith_inferences":["The temporal-precedence assumption is testable: if a few documented cases show a 'later' lister acting on independent intelligence before the earlier list appeared, or both copying a third source, those edges are miscoded and the magnitude of the resulting ranking error could be estimated.","Since the lists carry dates, the same decomposition could be run in a sliding window to trace how influence hierarchies shift after major events such as new UN resolutions or terrorist attacks; the paper only reports static rankings.","The who-acted-first construction should transfer to other domains where organizations publish denial or warning lists over time—travel advisories, product-safety recalls, cybersecurity threat lists—where the resulting network could expose de facto standard-setters."],"forward_implications":["If the potentials are meaningful, the ordering of sanctions issuers from upstream initiators to downstream followers can be estimated from public list data alone, without access to internal decision records.","Category-specific decompositions separate issue areas: the Iran–North Korea network puts the UN Security Council at the bottom, which the paper attributes to that council's slow approval mechanism, while the Al-Qaeda network shows the UN and the EU acting as information-gathering hubs.","The gradient/loop ratio quantifies whether a sanctions domain is driven by hierarchical diffusion or by reciprocal influence; terrorism-in-general lists are almost purely hierarchical (loop ratio about 0.02), while Al-Qaeda lists carry the most loop structure (about 0.21).","Because the estimated potentials are not redundant with PageRank, the decomposition adds a dimension to standard network-centrality tools for political analysis."],"supporting_citations":[{"why":"Supplies the optimization formulation of Helmholtz–Hodge potentials that the paper computes for each sanctions network.","marker":"[6]"},{"why":"Previous application of Helmholtz–Hodge decomposition to economic transaction networks, the direct methodological template for this study.","marker":"[7]"},{"why":"PageRank comparison that the paper uses to argue the Helmholtz–Hodge potential captures information independent of standard centrality.","marker":"[3]"},{"why":"Fast unfolding community detection used to identify the issue categories that structure the category-specific analyses.","marker":"[2]"},{"why":"Multiscale modularity method also used to define communities in the influence networks.","marker":"[8]"},{"why":"Source cited for the UN Security Council's complex approval mechanism, which the paper invokes to explain the council's bottom ranking for Iran–North Korea sanctions.","marker":"[5]"}],"fun_headline_variants":["Sanctions lists expose hidden influence hierarchy","Helmholtz-Hodge reveals top influencers in smart sanctions","OECD and ICTR least influenced; UNSC trails on Iran-North Korea","Who lists first? Network analysis shows true power structure","Smart sanctions network reveals least-influenced issuers: OECD and ICTR"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire ranking rests on treating 'listed the same entity earlier' as 'influenced the later lister'; if institutions act independently, lag for bureaucratic reasons, or copy a shared third source, the edges—and the potentials built from them—misrepresent the influence structure.","fun_headline_variants_meta":{"raw":{"variants":["Sanctions lists expose hidden influence hierarchy","Helmholtz-Hodge reveals top influencers in smart sanctions","OECD and ICTR least influenced; UNSC trails on Iran-North Korea","Who lists first? Network analysis shows true power structure","Smart sanctions network reveals least-influenced issuers: OECD and ICTR"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001143,"raw_usage":{"total_tokens":4741,"prompt_tokens":941,"completion_tokens":3800,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":557,"completion_tokens_details":{"reasoning_tokens":3716}},"tokens_in":557,"tokens_out":3800,"duration_ms":31665,"temperature":1.0,"reasoning_tokens":3716,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:33:20.679493+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Shuffle the listing dates within each sanctions category while preserving the content of each institution's lists, recompute the Helmholtz–Hodge potentials on many shuffles, and compare the resulting rankings with the reported ones; if the upstream/downstream ordering survives random date reassignment, then list timing is not what the hierarchy measures and the paper's interpretation would fail that test.","supporting_citations":[{"cited_title":"Statistical ranking and combinatorial hodge theory","cited_arxiv_id":null,"evidence_quote":"Supplies the optimization formulation of Helmholtz–Hodge potentials that the paper computes for each sanctions network."},{"cited_title":"Iyetomi H","cited_arxiv_id":null,"evidence_quote":"Previous application of Helmholtz–Hodge decomposition to economic transaction networks, the direct methodological template for this study."},{"cited_title":"Brin and L","cited_arxiv_id":null,"evidence_quote":"PageRank comparison that the paper uses to argue the Helmholtz–Hodge potential captures information independent of standard centrality."},{"cited_title":"Fast unfolding of communities in large networks","cited_arxiv_id":null,"evidence_quote":"Fast unfolding community detection used to identify the issue categories that structure the category-specific analyses."},{"cited_title":"Kitacyosen Kaku no Shikingen Kokuren Sousa no Hiroku [Funding Source of North Korea: A Note on United Nation ’s Investigation] Funding source","cited_arxiv_id":null,"evidence_quote":"Source cited for the UN Security Council's complex approval mechanism, which the paper invokes to explain the council's bottom ranking for Iran–North Korea sanctions."}],"review_version":1}