{"id":"f5f59bc1-6c24-4427-abc8-a52583bb610f","arxiv_id":"2501.05314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Applying the SDGs-GENEPY complexity framework to NITI Aayog data yields new rankings of Indian states and goal weights, with SDG 9 emerging as the most important goal.","lead":"This paper applies a network-based complexity method to rank Indian states and union territories on their progress toward the UN Sustainable Development Goals, using official NITI Aayog scores. It identifies which goals (such as SDG 9, Industry and Innovation) carry the most weight in separating high- and low-performing states, offering an alternative to simple average rankings.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper replaces its nonlinear GENEPY iteration with an unverified linearized eigenproblem; if the approximation diverges from the fixed point, the top ranking of SDG 9 could be an artifact.","rationale":"The reader's weakest assumption concerns the interval-scale nature of NITI Aayog scores. That is a genuine limitation, but it applies broadly to any score-based analysis and is not decisively checkable from the paper alone. A more specific and internally checkable risk is that the eigenvector formula used for the results is not the same object as the nonlinear iteration introduced in the Methodology. The paper does not validate that the linearization is accurate for this data matrix, so the central claim that SDG 9 carries the top weight is not yet established. The proposed test would settle this by comparing the two algorithms on the same data. Since the authors can run this check and report it in a revision, the reader's CONDITIONAL verdict stands; I see no grounds to reject outright or to accept without the added verification.","tokens_in":8740,"tokens_out":9401,"duration_ms":89391,"concrete_test":"Recompute the 2023-24 analysis two ways: (i) iterate the nonlinear system in the Methodology until convergence (e.g., 10^5 steps with the stated nu_s and nu_g normalizations), and (ii) compute the principal eigenvectors of U and V as in the paper, then compare the resulting goal weight vectors W_g. If the goal with maximum W_g is not SDG 9 in the iterated solution, or if the Spearman rank correlation between the two W_g vectors is below 0.9, the linearization matters and the headline ranking is not robust. Repeat for the 2019 matrix to check whether any mismatch persists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Methodology first defines GENEPY scores through a nonlinear system: fD_s(n+1)=sum_g I_sg C_g(n), fC_g(n+1)=1/sum_s I_sg/D_s(n), with normalizations by nu_s and nu_g. It then states that Sciarra et al. showed this can be approximated by an eigenproblem, and uses the principal eigenvectors of U=N N' and V=N' N, where N_sg=I_sg/(k_s k'_g). This linearization is dataset-dependent: the eigenvector of the similarity matrix matches the fixed point of the nonlinear iteration only approximately, and the paper provides no convergence check or code. Moreover, the normalizations by nu_s and nu_g in the nonlinear system disappear in the eigenproblem, and the definition of k_s is misprinted as summing over s rather than over g, so the exact object being computed is ambiguous. If the linearization error is large, the normalized goal weights W_g=C_g/k'_g, and hence the claim that SDG 9 has the largest weight, may differ from the iterative solution. The central claim therefore rests on an unverified equivalence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies the SDG-GENEPY framework, previously introduced by Sciarra et al., to NITI Aayog SDG India Index scores for 36 Indian states and union territories across four editions (2018, 2019, 2020-21, and 2023-24). The states and SDGs are treated as a weighted bipartite network, and the authors compute state complexity scores as principal eigenvectors of a similarity matrix, derive goal weights, and rank states accordingly. The headline results are that SDG 9 (Industry, Innovation, and Infrastructure) receives the largest weight, Kerala ranks first among states, and the proposed rankings correlate only weakly with the official NITI Aayog rankings. The paper also presents evolutions of state ranks and goal weights over the four years.","tokens_in":8875,"tokens_out":6450,"duration_ms":56937,"significance":"If the computations are correct, the paper offers a useful, policy-relevant application of a known network-complexity method to subnational SDG data. The data are public and the method is standard, so the results are in principle reproducible; the paper does not, however, provide code or machine-checked proofs. The finding that SDG 9 dominates, and the identification of top-performing states, could inform Indian SDG discussions. The contribution is primarily empirical rather than methodological, and the current manuscript leaves several load-bearing technical points unverified, so the significance is conditional on those points being resolved.","major_comments":[{"comment":"The paper asserts, without derivation or numerical evidence, that the nonlinear iteration for D_s and C_g can be replaced by the principal eigenvectors of U = NN' and V = N'N. The displayed iterative equations include normalization factors ν_s and ν_g that do not appear in the eigenproblem, and the definition of k_s is printed as ∑_s I_sg(τ) rather than ∑_g I_sg(τ), so the exact object being computed is ambiguous. Since the headline finding that SDG 9 has the largest weight W_g (Results and Discussion, second paragraph) is an output of this eigenvector computation, the manuscript should either supply the equivalence proof/citation for this dataset or compare the eigenvector solution to the fixed point of the nonlinear map (e.g., by iterating the equations to convergence). Without this check, the central ranking claim is not established.","section":"Methodology and The SDGs-GENEPY framework"},{"comment":"The intertemporal comparison is compromised by non-comparable inputs. The data section states that the 2018 edition used 62 indicators and excluded goals 11, 12, and 13 (while goal 14 is always excluded), whereas later editions used 100-115 indicators and a different goal set. The number of states and union territories also changes across years (e.g., Telangana and Ladakh were created). The paper nonetheless presents 'evolution of ranks' and 'evolution of weights' across the four years as if they were comparable. The authors should restrict the analysis to a common set of goals (and, where possible, states) or explicitly quantify the effect of the changing goal set on the computed weights; otherwise the temporal claims in Figures 4 and 5 are not supported.","section":"Datasets and Results and Discussion (Figures 4-5)"},{"comment":"The paper claims 'there is a positive correlation between this study’s index and the NITI Aayog rankings, the relationship remains weak,' but no correlation coefficient, statistical test, or supporting figure is provided anywhere in the manuscript. This is a specific, testable claim about the validity of the proposed index; the authors should report, for each year, the rank correlation (e.g., Spearman's ρ) between D_s and k_s, with a confidence interval or p-value.","section":"Results and Discussion, paragraph 2"},{"comment":"No uncertainty or sensitivity analysis is reported. The complexity scores and goal weights are deterministic functions of the NITI Aayog score matrix, but those scores are themselves averages of indicator scores with NITI-defined target normalizations, and small perturbations in the scores could alter the eigenvector rankings. Because the paper's contribution is precisely the ranking of states and goals, the authors should assess robustness, for instance by bootstrapping over indicators within goals, jackknifing states, or perturbing scores, and report the stability of the top-ranked states and of the SDG 9 weight.","section":"All results (Figures 2-5)"}],"minor_comments":[{"comment":"The first sentence contains 'NITI The Aayog' and the text later refers to 'NIF' instead of 'NITI'; please correct these typographical errors.","section":"Datasets"},{"comment":"Reference [26] is cited as 'Sciarra et al.25' in the introduction; the reference numbering should be checked and corrected.","section":"Introduction"},{"comment":"The paper says the data contain scores in '16 sustainable development goals' but also excludes goal 14 always and goals 11, 12, and 13 in 2018; these statements are inconsistent and should be reconciled (the 2018 edition would then cover 13 goals, not 16).","section":"Datasets"},{"comment":"The iterative equations for fD and fC include a division by ν_s and ν_g whose purpose is not explained; please define these normalization factors and clarify how they are absorbed in the eigenproblem formulation.","section":"Methodology"},{"comment":"There are several figure-cross-reference errors: the weights are said to be shown in 'Figure 3c' instead of Figure 2C, and 'Kc ranks' should likely read 'k_s ranks'; please standardize figure labels and notation throughout.","section":"Results and Discussion"},{"comment":"The abstract states the study 'enables data-driven policy-making,' but the paper does not translate its results into concrete policy recommendations; consider adding a short policy implications subsection or softening the claim.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper is an application of an existing method (GENEPY, Sciarra et al. 2020/2021) to Indian subnational data. The novelty is modest, but the topic is appropriate for an applied policy-oriented journal. The main concern is that the presentation is quite rough: several misprints, inconsistent figure labels, and an unsupported correlation claim. The lack of code or convergence checks is a barrier to verifying the central claim. I would ask the authors to address the major comments before considering publication and to carefully proofread the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this is an application paper, not a methods paper. It takes Sciarra et al.'s GENEPY network framework, runs it on four vintages of the NITI Aayog SDG India Index, and reports new subnational rankings and goal weights, with SDG 9 coming out on top. That is a genuinely new empirical output, and the computation is probably close to right. But the paper does not give you enough to check it: no code, no data, no sensitivity analysis, and the one place where the method description matters most—the linearization of the nonlinear iteration into the eigenproblem—is asserted rather than verified.\n\nWhat it does well: the writing is clear, the dataset description is honest about changing indicator counts and coverage gaps, and the figures are informative. The authors also correctly note that NITI Aayog ranks states and UTs separately while this paper ranks them together, which explains some rank mismatches. That transparency is worth crediting. The citation pattern is fair: the GENEPY and economic-complexity literature is credited properly.\n\nThe soft spots, in order. First, the eigenproblem step. The nonlinear system in Methodology has normalizations by ν_s and ν_g; the linearized version via N_sg = I_sg/(k_s k'_g) drops those, and there is a typo in the definition of k_s (it sums over s instead of g). The paper says the eigenproblem is an approximation but never checks, for this dataset, whether the principal eigenvector is close to the fixed point. If the approximation error is large, the top ranking of SDG 9 could shift. This is the load-bearing issue, and it is easy to fix with a convergence check. Second, inter-temporal comparison is shaky because the indicator set changes across years and some goals are absent in 2018; the paper acknowledges this but proceeds anyway. Third, the \"weak correlation\" with NITI Aayog rankings is stated without a number. Since k_s is essentially the NITI sum, the correlation should be high; \"weak\" needs support. Fourth, there are no reproducibility artifacts, which is a real limitation for a paper whose main output is a ranking.\n\nThe circularity concern is real but modest: the goal weights come from the same scores and are then used to reweight those scores, so calling them \"importance\" is interpretive. That is inherent to the method, not a fatal flaw.\n\nBottom line: this deserves a referee, but it needs heavy revision. Add the convergence check, provide the data and code, quantify the correlation, and soften the claims about goal importance. Then it becomes a serviceable empirical contribution for people working on subnational SDG tracking.","headline":"Straightforward application of a known network method to NITI Aayog's SDG India data; the new subnational rankings are plausible, but the paper skips verification of its linearized eigenproblem, provides no code or data, and overreaches on a few interpretive claims.","tokens_in":9475,"tokens_out":3409,"would_cite":false,"duration_ms":30422,"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":"Applying the generalized economic-complexity (GENEPY) algorithm to India's official state-level SDG scores, this paper claims that SDG 9 (Industry, Innovation, and Infrastructure) is the highest-weight goal and Kerala the leading state.","keywords":["sustainable development goals","India","bipartite network","economic complexity","GENEPY","eigenvector centrality","SDG 9","state-level ranking"],"falsifier":"A reader could rerun the algorithm on the same score matrix after replacing every score by its rank, which preserves all orderings but destroys the ratio structure. If the goal weights or state complexity ranks change materially under that transformation, the interval-scale premise is false and the rankings are not supported by the data.","tokens_in":8489,"feed_emoji":"🏭","tokens_out":10487,"duration_ms":94261,"temperature":0.7,"pith_summary":"The paper tries to extract a signal from India's official state-level SDG scores that a simple average hides: which states have the capability to achieve difficult goals, and which goals are hard enough that only capable states reach them. It applies the SDGs-GENEPY algorithm, which treats the 36 states and union territories and the 16 scored goals as the two sides of a weighted bipartite network and computes a complexity score for each side. The headline empirical result is that SDG 9 (Industry, Innovation, and Infrastructure) carries the largest goal weight, with SDG 2, SDG 5, and SDG 13 also prominent, while Kerala ranks first among states. The new state rankings correlate positively but only weakly with the official composite index, so the method changes the picture, not just the labels. If the result stands, it would give policymakers a data-driven way to see which SDGs are structurally harder and which states deserve targeted support on those goals.","feed_headline":"Industry and infrastructure is India's hardest SDG to achieve","feed_subtitle":"A network complexity score of India's official state SDG data ranks Kerala first and gives SDG 9 the biggest weight.","key_machinery":"The central object is a weighted bipartite network whose adjacency matrix is $I_{sg}(\\tau)$, the score of state $s$ on goal $g$ in year $\\tau$. The GENEPY iteration updates state complexity $D_s$ and goal complexity $C_g$ through coupled equations: $D_s$ sums a state's scores weighted by goal complexity, while $C_g$ is a harmonic mean of state scores weighted by inverse state complexity. The paper follows the earlier result that this iteration can be linearized: $D_s$ becomes the principal eigenvector of the state similarity matrix $U=NN^\\top$ and $C_g$ the principal eigenvector of $V=N^\\top N$, where $N_{sg}=I_{sg}/(k_s k'_g)$ is the score normalized by the state's total score and the goal's adjusted score. The reported goal weights are $W_g=C_g/k'_g$. This machinery turns the raw scores into centrality measures and gives a quantitative ranking for both nodes in the bipartite network.","core_discovery":"On its own terms, the discovery is that the GENEPY algorithm, run on four editions of the official state-goal score matrix (2018, 2019, 2020-21, 2023-24), yields a ranked list of states and a ranked list of goals that are not the same as the official composite ranking. For the most recent year, the largest goal weight is assigned to SDG 9; the other high-weight goals include SDG 2, SDG 5, and SDG 13. Among states, Kerala sits at the top of the complexity ranking. The authors interpret these scores as centralities in a bipartite network, showing that more complex states achieve more complex goals in the Indian federal context, and they argue the weighted performance profiles reveal where each state is strong or weak after accounting for goal difficulty.","pith_inferences":["The paper does not test this, but because GENEPY rewards goals achieved selectively by high-complexity states, SDG 9's top weight may reflect the wide spread of industrial and infrastructural performance across states rather than any inherent national priority; a goal on which all states scored alike would receive a low weight no matter how important it is.","A natural check, absent from the paper, would be to rerun the algorithm on rank-transformed or winsorized scores; if the goal weights and state rankings lurch under such a transformation, the interval-scale assumption is doing all the work.","The weak correlation with the official composite suggests that adopting complexity weights would reallocate policy attention, but whether that reallocation improves SDG outcomes is an empirical question this paper does not answer."],"forward_implications":["States that rank high on the official average but low on the complexity score are exposed as performers on easy goals whose capability is narrow, while the reverse holds for states that do comparatively well on high-weight goals.","Multiplying raw scores by goal weights, as in the paper's weighted performance figures, provides a direct visual and numerical diagnostic of which goals drag each state down.","Recomputing ranks each year, as done for 2018-2024, tracks how administrative splits such as the creation of new states and union territories change the relative complexity landscape.","Future editions of the official index can be plugged into the same algorithm, allowing continuous monitoring of whether goals become easier or harder over time."],"supporting_citations":[{"why":"Supplies the original method of reflection that defines complexity from how selectively outcomes are achieved.","marker":"[23]"},{"why":"Provides the non-linear fitness-complexity iteration that the GENEPY equations adapt.","marker":"[24]"},{"why":"Shows how the iterative scores can be recast as centralities and eigenvectors, the linearization used here.","marker":"[25]"},{"why":"Proposes the SDGs-GENEPY framework specifically for ranking countries and goals, the method applied to Indian states.","marker":"[26]"}],"fun_headline_variants":["Kerala tops India's SDG complexity ranking","SDG 9 is India's most complex goal to achieve","Network measure flips India's SDG state rankings","India's SDG leaders: Kerala first, industry hardest","Complex network ranks India's SDG states and goals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire ranking rests on treating the official 0-100 goal scores for each state as real, comparable numbers, so that taking averages, products, and ratios of them inside the complexity algorithm is meaningful; if the scores are only rough orderings or use different rubrics for different goals, the complexity ranks and goal weights are artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Kerala tops India's SDG complexity ranking","SDG 9 is India's most complex goal to achieve","Network measure flips India's SDG state rankings","India's SDG leaders: Kerala first, industry hardest","Complex network ranks India's SDG states and goals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000427,"raw_usage":{"total_tokens":2215,"prompt_tokens":1001,"completion_tokens":1214,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":617,"completion_tokens_details":{"reasoning_tokens":1136}},"tokens_in":617,"tokens_out":1214,"duration_ms":11428,"temperature":1.0,"reasoning_tokens":1136,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:12:28.682851+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could rerun the algorithm on the same score matrix after replacing every score by its rank, which preserves all orderings but destroys the ratio structure. If the goal weights or state complexity ranks change materially under that transformation, the interval-scale premise is false and the rankings are not supported by the data.","supporting_citations":[{"cited_title":"& Laio, F","cited_arxiv_id":null,"evidence_quote":"Shows how the iterative scores can be recast as centralities and eigenvectors, the linearization used here."},{"cited_title":"& Laio, F","cited_arxiv_id":null,"evidence_quote":"Proposes the SDGs-GENEPY framework specifically for ranking countries and goals, the method applied to Indian states."}],"review_version":1}