{"id":"104570a4-53a9-49ba-a7f5-b00d783e59b5","arxiv_id":"2507.03760","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"Causal analysis of 229 Trans-Neptunian Objects finds that orbital inclination does not cause color, which the authors interpret as evidence that TNO colors are primordial.","lead":"Using a causal-discovery computer algorithm on 229 Kuiper belt objects, this paper concludes that the objects' colors are not caused by their orbital tilt. The authors read this as evidence that TNO colors were fixed when the objects formed, which would settle a long-standing debate in planetary science.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The PAG only excludes inclination as a cause of color; the conclusion that colors are primordial requires assuming the latent confounder is formation location, which is the claim under test.","rationale":"The reader's REJECT verdict is well placed. The paper does useful work: the FCI pipeline is standard, the code is released, and the jackknife robustness check is a real strength. However, the central astrophysical claim is not a theorem of the computed PAG. The graphical result rules out only the specific causal direction 'inclination → color.' Because FCI is designed to remain agnostic about latent common causes, the PAG cannot, by itself, distinguish 'color → inclination' from 'unobserved confounder → both color and inclination.' The only justification offered for choosing the confounder to be primordial formation location is Section 2.1's prior assumption that colors are primordial; that is precisely the proposition the paper claims to prove. A non-primordial confounder would preserve the PAG and the 98.7% per-test statement while overturning the conclusion. The 98.7% figure is also not a confidence in the causal claim; it is the per-test α, with no correction for multiple conditional-independence tests, so the headline certainty is overstated. My recommendation is to keep the REJECT verdict (UNCHANGED): the paper is a reasonable methods demonstration, but it does not provide causal evidence for primordial TNO colors.","tokens_in":7603,"tokens_out":5015,"duration_ms":58164,"concrete_test":"Simulate a non-primordial generative model with latent U (collisional or irradiation processing) where U→color and U→inclination, with inclination having no causal path to color; generate 229 points matching the observed marginal distributions, and run the identical FCI pipeline (Yeo-Johnson transform, Fisher-Z test, α=0.013). If the recovered PAG reproduces the three edges of Fig. 2, then the observed PAG is equally consistent with a non-primordial confounder and the primordiality claim is not identified. If the PAG changes materially in a way that cannot reproduce edge (iii), the concern is weakened accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the interpretation of edge (iii) in Fig. 2. The PAG edge is color ◦→ i, which by the paper's own legend (Section 2.3) means 'i is not an ancestor of color'; it does not assert color → i. The same edge is compatible with an unobserved confounder causing both color and inclination. Section 3.2 concedes this: 'this link ... leaves open the possibility of an unobserved confounder causing both color and the inclination. This confounder can be the formation location itself, if we were to assume the color and initial location to be two distinct variables.' But Section 2.1 has already made the 'fundamental assumption' that colors are primordial and are a proxy for initial semimajor axis. Selecting 'formation location' as the confounder therefore imports the conclusion into the analysis. A non-primordial latent variable—for instance, collisional resurfacing or irradiation that influences both surface color and, through size or dynamical evolution, orbital inclination—would produce exactly the same PAG while making the central conclusion false. The '98.7% certainty' is α per conditional-independence test, not a posterior probability for the causal or PAG statement, so it cannot arbitrate this ambiguity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies the Fast Causal Inference (FCI) algorithm to a sample of 229 trans-Neptunian objects (TNOs), using semimajor axis, eccentricity, inclination, and spectral slope (color). The output is a Partial Ancestral Graph (PAG) with three edges: eccentricity is not an ancestor of semimajor axis, eccentricity and inclination share a latent common cause, and inclination is not an ancestor of color. The authors interpret these edges as causal evidence for the primordial origin of TNO colors: they claim that color is causally antecedent to inclination, that models involving irradiation or collisional color modification are excluded, and that the analysis 'rediscovers' Neptune. The paper argues that these conclusions follow with high certainty from a purely data-driven, model-agnostic procedure.","tokens_in":7834,"tokens_out":7027,"duration_ms":78150,"significance":"The paper is a clear and well-written demonstration of a modern causal discovery method applied to an astrophysical dataset. The FCI implementation is standard, the robustness checks (jackknife by subpopulation, multiple conditional independence tests, different significance thresholds) are thoughtful, and the authors make their code and data publicly available. If the causal interpretation were valid, the conclusion that TNO colors are predominantly primordial would be an important contribution to the solar-system community. However, the central claim as stated is not supported by the PAG output: the edge color ◦→ i does not establish that color causes inclination, and the assumption that colors are primordial is built into the interpretation of the latent confounder. These are load-bearing issues that undermine the paper's main conclusion.","major_comments":[{"comment":"The PAG edge color ◦→ i is interpreted as evidence that 'TNO color is causally antecedent to inclination' (abstract and §4), but the paper's own legend in §2.3 states that X ◦→ Y means 'Y is not an ancestor of X.' Therefore the edge only rules out i → color; it does not assert color → i. The causal direction from color to inclination is an interpretation, not a result of the algorithm. This overreach is central to the paper's claim of primordiality.","section":"§2.3, Fig. 2, §3.2, §4"},{"comment":"The '98.7% certainty' is the per-test conditional-independence significance level α=0.013, not a posterior probability for the PAG or for any causal edge. The paper states this correctly in §3.1 but then transfers the number to the conclusion ('with 98.7% certainty that TNO color is causally antecedent to inclination'), which is not justified. The number of tests performed and any multiple-testing correction are not reported, so the claimed confidence is not well defined.","section":"§3.1, Fig. 2 caption"},{"comment":"The analysis is circular with respect to the primordiality of colors. Section 2.1 states as a 'fundamental assumption' that colors are primordial and are a proxy for initial semimajor axis. Section 3.2 then interprets the unobserved confounder that the PAG leaves open between color and inclination as 'the formation location itself.' The conclusion that colors are predominantly primordial therefore rests on an assumption that is equivalent to the claim being tested. A non-primordial latent confounder—for example, a collisional or irradiation-related variable that affects both surface color and, through size or dynamics, orbital inclination—would produce the same PAG but make the central conclusion false. The paper acknowledges this possibility in §3.2 but does not address it in the interpretation.","section":"§2.1, §3.2"},{"comment":"The claim that the model 'excludes' irradiation and collisional evolution is too strong. The PAG only excludes i as an ancestor of color (edge (iii)); no edge involving a and color is listed, and even if an edge color ◦→ a were present it would only rule out a → color, not indirect or latent mechanisms. Collisional resurfacing or irradiation could enter through an unobserved confounder, as the paper itself concedes for the color–inclination link. The data do not distinguish these scenarios from the primordial-formation-location interpretation.","section":"§3.2"},{"comment":"The 'Neptune rediscovery' argument is an interesting illustrative exercise but does not validate the method for the color question. The e ↔ i edge in the PAG indicates only a latent common cause; many mechanisms could produce such dependence, and Neptune is one possible identification, not a consequence of the causal discovery algorithm. This interpretive step is not evidence that the other edges are correctly interpreted as causal.","section":"§3.2"}],"minor_comments":[{"comment":"The phrase 'with 98.7% certainty' should be reworded to 'with per-test significance α=0.013' or similar, to avoid conveying a posterior probability.","section":"Abstract"},{"comment":"The sentence 'removing any subsample of 48 Classicals, 102 Resonant, 36 Centaurs, 28 Scattered, or 15 Detached TNOs results in no change' is unclear; it should read 'removing each subpopulation in turn' or similar.","section":"§3.1"},{"comment":"The notation 'color ← a' is not defined in the paper; the standard edge notation from §2.3 should be used, e.g., 'a → color' or 'a is not an ancestor of color.'","section":"§3.2"},{"comment":"The parenthetical list of subpopulations is missing a verb; 'it consists of a total of 229 TNOs including hot classicals, centaurs, and resonant/scattered objects, in a dataset for which discovery biases were modeled' is grammatically awkward.","section":"§2.1"},{"comment":"The choice of α=0.09 for the KCI test is not justified; a brief explanation of how this threshold was selected would improve reproducibility.","section":"§3.1"},{"comment":"The GitHub URL is malformed ('/gtbhttps:'); it should be 'https://github.com/ZehaoJin/causalTNOs'.","section":"Acknowledgements"}],"recommendation":"reject","confidential_remarks":"The paper is likely to attract attention in the astro-ph community because it applies a modern causal discovery method to a classic problem. However, the central conclusion is not supported by the PAG output as presented, and the circular use of the primordiality assumption is a fundamental issue. A revision that substantially weakens the claims could be considered, but the current manuscript's stated conclusions overreach the evidence in a way that cannot be fixed by local edits."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about this one. First, the causal discovery analysis is competently done and the robustness checks are genuinely good – jackknifing by subpopulation, three different conditional independence tests, and the code is on GitHub. Second, the paper's headline claim – '98.7% certainty' that TNO colors are primordial – is not supported by the output. That number is just the per-test alpha (0.013), not a posterior probability for the PAG or the causal conclusion.\n\nWhat's actually new is the application of FCI to TNO orbital elements and colors, and the specific finding that inclination is not an ancestor of color (the color ◦→ i edge). That directionality is a new piece of evidence, even though the broader conclusion of primordial colors is already in the literature (Nesvorny et al. 2020; Ali-Dib et al. 2021). The Neptune 'rediscovery' via the e↔i latent confounder is a nice sanity check, showing the method can pick up known dynamical structure.\n\nThe soft spot is the interpretive leap. The PAG's color ◦→ i edge only rules out i causing color. It is fully compatible with an unobserved confounder causing both. Section 3.2 says as much. But Section 2.1 already assumes colors are primordial and a proxy for initial location, so when the authors then suggest the confounder is 'the formation location itself,' they are importing the conclusion into the evidence. A non-primordial latent process – e.g., collisional resurfacing that affects color and, through size or dynamical state, inclination – would produce the same PAG. The exclusion of irradiation models is also too strong: the absence of an a–color edge does not eliminate latent irradiation variables that influence both color and inclination.\n\nThe paper is transparent and well-written, and the authors are honest about the fundamental assumption. But the abstract and conclusion present the result as a definitive resolution of the debate. That overreach matters.\n\nMy take: this is a solid methods demonstration worth a read by anyone interested in causal discovery in astronomy, and it deserves a serious peer review. But I would not treat it as strong evidence for primordial colors without a much more careful treatment of latent confounders.","headline":"Clever causal-discovery application undermined by an overclaimed certainty and a circular latent-confounder interpretation.","tokens_in":8399,"tokens_out":3396,"would_cite":false,"duration_ms":36632,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A causal analysis of 229 trans-Neptunian objects concludes, at 98.7% confidence, that surface color drives orbital inclination, making TNO colors predominantly primordial.","keywords":["trans-Neptunian objects","Kuiper belt","causal discovery","primordial colors","color–inclination correlation","orbital inclination","Fast Causal Inference","solar system formation"],"falsifier":"Rerun the same causal analysis on a TNO sample that includes a measured proxy for collisional resurfacing, such as crater density or albedo, and condition on that proxy. If the color–inclination association weakens or reverses, or a color←inclination edge appears, the primordial-color conclusion is refuted; a collisional family with demonstrably altered surfaces showing such an edge would be a direct contradiction.","tokens_in":7375,"feed_emoji":"🪐","tokens_out":14714,"duration_ms":149417,"temperature":0.7,"pith_summary":"The paper tries to settle a long-standing dispute in planetary science: whether the surface colors of trans-Neptunian objects were fixed by where they formed in the protoplanetary disk, or later altered by collisions and radiation. Using only orbital elements and spectral color for 229 objects, it runs a statistical causal discovery algorithm that makes no physics-based assumptions, and reports 98.7% confidence that color is causally antecedent to orbital inclination rather than the reverse. If that is right, the well-known color–inclination correlation is a fossil of formation location, and the main competing explanations—collisional resurfacing and irradiation-driven color change—are excluded as dominant processes. The same blind analysis also recovers the dynamical signature of an unseen perturbing body, which the authors interpret as Neptune, serving as a sanity check on the method.","feed_headline":"Kuiper belt object colors are set at birth, causal graph finds","feed_subtitle":"A data-driven causal analysis of 229 objects finds, at 98.7% confidence, color drives inclination, pointing to primordial colors.","key_machinery":"The central object is a Partial Ancestral Graph (PAG), a causal graph that encodes direction information while allowing for unmeasured variables. It is produced by Fast Causal Inference (FCI), a constraint-based causal discovery algorithm that tests conditional independences among measured variables; the PAG's edge symbols distinguish 'is an ancestor of,' 'is not an ancestor of,' and 'shares a latent common cause.' The paper's decisive result is the edge color $\\circ\\!\\to i$, which says inclination is not an ancestor of color. The analysis uses a linear conditional independence test on power-transformed data at stringent thresholds, and the same graph is recovered with untransformed data, with a non-linear test, and when each TNO subpopulation is jackknifed out.","core_discovery":"The paper's central claim is that the colors of trans-Neptunian objects are predominantly primordial, set by the chemical composition of their formation location. The evidence is a Partial Ancestral Graph learned from 229 objects at a 98.7% significance threshold, whose decisive edge states that inclination is not an ancestor of color; the authors take this to mean that color is the root cause of the inclination distribution, with formation location acting through inclination-raising secular resonances. A second edge, stating that current semimajor axis is not an ancestor of color, disfavors irradiation-driven color modification. A third, a latent common cause between eccentricity and inclination, matches the dynamical signature of an external perturber, and the authors read it as a blind rediscovery of Neptune. Together these edges lead to the conclusion that TNO colors reflect conditions at formation rather than subsequent collisional or radiative evolution.","pith_inferences":["Beyond the paper, the same pipeline could be applied to asteroid families or exoplanet populations, where primordial-versus-evolutionary color questions also arise, to test whether such formation-location locking is a general planetary-system phenomenon.","An open extension is to treat the rediscovery of Neptune as a template for predicting other unseen perturbers; a distant massive planet would be expected to imprint a similar latent-common-cause signature on the eccentricity and inclination of outer TNOs.","The 98.7% figure constrains the direction color→inclination but does not by itself exclude a common cause; future data on physical properties such as albedo, density, or cratering state would be needed to confirm that the common cause is indeed formation location rather than a later process."],"forward_implications":["If colors are set at formation, the color–inclination correlation becomes a map of where each TNO formed in the protoplanetary disk.","Collisional resurfacing models that make inclination drive color are excluded as dominant, because that causal direction is disallowed.","Irradiation-driven color models are disfavored because the current semimajor axis is not an ancestor of color.","The same causal pipeline, without physical inputs, identifies a latent common cause of eccentricity and inclination that the authors interpret as the dynamical signature of Neptune.","The inferred graph is robust to removing any one TNO subpopulation, so the result is not driven by classicals, resonants, centaurs, scattered, or detached objects alone."],"supporting_citations":[{"why":"Supplies the photometric survey data and the observed color–inclination correlation the analysis builds on.","marker":"Marsset et al. (2019)"},{"why":"Adds the eccentricity–color trend and the prior primordial-origin interpretation.","marker":"Ali-Dib et al. (2021)"},{"why":"Defines the FCI/PAG causal discovery framework that allows latent variables.","marker":"Spirtes et al. (2001)"},{"why":"Provides the FCI implementation used to compute the reported graph.","marker":"Zheng et al. (2024)"},{"why":"Presents the primordial-origin hypothesis for Kuiper belt colors that the paper concludes is favored.","marker":"Nesvorný et al. (2020)"},{"why":"Represents the collisional color-evolution scenario the result rules out.","marker":"Luu & Jewitt (1996)"},{"why":"Proposes the collisional mechanism for Kuiper belt colors that the causal graph excludes.","marker":"Stern (2002)"},{"why":"Proposes the irradiation/volatile-loss color diversity model disfavored by the result.","marker":"Brown et al. (2011)"},{"why":"Extends the irradiation/volatile-loss model, also disfavored by the color–semimajor axis result.","marker":"Wong & Brown (2017)"}],"fun_headline_variants":["Causal analysis: TNO colors are primordial","Color drives TNO inclination, causal graph finds","TNO colors set at birth, causal model says","Root cause of TNO inclination is color, study says","TNO color is causal, not consequence, at 98.7%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim presumes that the unobserved factor linking color and inclination is the primordial formation location; if some later process such as collision-driven resurfacing is the real common cause, the conclusion that colors are primordial does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Causal analysis: TNO colors are primordial","Color drives TNO inclination, causal graph finds","TNO colors set at birth, causal model says","Root cause of TNO inclination is color, study says","TNO color is causal, not consequence, at 98.7%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000166,"raw_usage":{"total_tokens":1268,"prompt_tokens":972,"completion_tokens":296,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":216}},"tokens_in":588,"tokens_out":296,"duration_ms":4196,"temperature":1.0,"reasoning_tokens":216,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:03:34.875758+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rerun the same causal analysis on a TNO sample that includes a measured proxy for collisional resurfacing, such as crater density or albedo, and condition on that proxy. If the color–inclination association weakens or reverses, or a color←inclination edge appears, the primordial-color conclusion is refuted; a collisional family with demonstrably altered surfaces showing such an edge would be a direct contradiction.","supporting_citations":[{"cited_title":"Causal-learn: Causal discovery in python","cited_arxiv_id":null,"evidence_quote":"Provides the FCI implementation used to compute the reported graph."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proposes the collisional mechanism for Kuiper belt colors that the causal graph excludes."}],"review_version":1}