{"id":"3b333ead-adfa-4292-9d50-27e085780586","arxiv_id":"2506.19166","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Over the past decade, understanding of supermassive black hole growth has advanced through new facilities, larger datasets, new techniques, and conceptual shifts, with AGN now viewed as transient events in galaxy lifecycles.","lead":"This review summarizes a decade of progress in understanding how supermassive black holes grow, updating a 2012 review by the same lead authors. It adds a citation-network map of black-hole research and a qualitative analysis of the drivers of progress, aimed at researchers and science-policy audiences.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The five progress-driver categories in Table 1 are asserted post hoc under an explicitly 'know it when you see it' method; the citation-network analysis does not test the driver attribution, so the paper's central claim about what drives progress is unsupported.","rationale":"The reader's weakest assumption identifies exactly the same vulnerability: the driver analysis depends on subjective recognition of progress and attribution to the five categories, and the authors themselves acknowledge the 'know it when you see it' approach. My concern sharpens this by noting that the citation network, the paper's quantitative element, does not actually test the driver attribution; it only situates the review within the literature. This means the central causal claim—that specific drivers enabled progress—is currently an assertion rather than a measured result. However, the scientific review content is broadly accurate and valuable, and the driver analysis is a secondary, self-avowedly qualitative component. The appropriate outcome is therefore the same conditional acceptance the reader recommended: the paper should be accepted with the driver claims reframed as hypotheses, or with the coding framework and citation artifacts released for independent verification. I see no reason to move the verdict to reject or to upgrade to unconditional acceptance, since the concern is about a specific methodological gap rather than a demonstrated error in the scientific synthesis.","tokens_in":52079,"tokens_out":3182,"duration_ms":39805,"concrete_test":"Extract the enumerated progress claims from the section summaries (e.g., §3.8, §4.7, and analogous passages in §5–§6), remove the authors' driver labels, and have five independent AGN researchers, blind to the paper's attributions, classify each claim into the five Table 1 categories (allowing multiple labels) and rate their confidence. Compute Fleiss' kappa for inter-rater agreement. If κ < 0.6, the driver categories are not sufficiently objective to support the central claim; if κ ≥ 0.6, the attribution is at least reproducible across specialists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's abstract claims that progress was enabled by new facilities, larger datasets, new techniques, community engagement, and conceptual shifts. This driver attribution is load-bearing because it is the paper's distinctive addition beyond a conventional review. However, Section 2.3 explicitly states that, 'in the absence of any widely accepted approach to quantify or otherwise measure scientific progress, in this review we adopt a “know it when you see it” approach,' and Table 1 describes the five categories as 'postulated.' The subsequent sections label specific advances and attach them to these categories, but the labels are assigned by the same authors who selected the baseline (AH12) and who chose which advances count as progress. The citation-network analysis in Sections 2.1 and 7 identifies subfield clusters and locates AH12 within the literature, but it never maps clusters to driver categories, nor does it test whether community engagement or conceptual shifts causally preceded progress. Thus the central causal claim rests on an untested, non-reproducible classification. If the classification is unreliable, the paper's headline conclusion about the drivers of black-hole growth research is anecdotal rather than demonstrated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper is a review of the past decade of research on the growth of supermassive black holes, framed as an update to Alexander & Hickox (2012, AH12). It surveys the main scientific topics (gas accretion from kpc to sub-pc scales, the AGN-host-galaxy connection, high-redshift quasars, and AGN feedback), and it adds a bibliometric component: a citation-network analysis of the broader black-hole literature intended to place the review in context and to identify what has driven scientific progress over the last decade. The authors explicitly state that, in the absence of a widely accepted quantitative measure, they adopt a \"know it when you see it\" approach to recognizing progress, and they postulate five categories of drivers (facilities, data, techniques, community, and conceptual shifts; Table 1). The scientific synthesis is the main body of the paper, while the driver analysis is presented as a secondary objective and is summarized in the abstract as a causal statement about what has \"enabled\" progress.","tokens_in":52327,"tokens_out":4222,"duration_ms":53172,"significance":"If read as an expert synthesis, the review is valuable and broadly consonant with current consensus: AGN as recurrent events in galaxy lifecycles, a clumpy disk plus polar-wind obscuring medium, and major advances from new facilities (ALMA, JWST, eROSITA, LOFAR, EHT) and large multiwavelength datasets. The paper is transparent about its qualitative method, and it makes a useful attempt to structure the field through the AH12 baseline and citation-network analysis. However, the paper's distinctive contribution, the identification of the drivers of progress, rests on the authors' own subjective classification and is not tested by the quantitative analysis. The claims about drivers should therefore be regarded as expert hypotheses rather than measured conclusions; in its current form the causal language in the abstract and Section 7 is stronger than the evidence presented. The strengths of the paper are its comprehensiveness, its explicit philosophical framing, and the reproducible citation-network construction; these make it a useful community resource even if the driver attribution is provisional.","major_comments":[{"comment":"The abstract states that the progress described in the paper \"has been enabled by\" new facilities, larger datasets, new techniques, and community engagement, but this causal attribution is not quantitatively supported. Section 2.3 explicitly adopts a \"know it when you see it\" approach and Table 1 labels the five categories as \"postulated\"; the citation-network analysis in Sections 2.1 and 7 identifies clusters of papers but does not map advances onto the five driver categories, nor does it test whether community engagement or conceptual shifts preceded the cited progress. The paper should either provide an explicit operational mapping between the citation-network statistics and the driver categories, or reframe the abstract and Section 7 conclusions as expert hypotheses rather than established causal claims.","section":"§2.3, Table 1, Abstract"},{"comment":"Community engagement is proposed as a driver of progress, with the citation network offered as a proxy, but citation networks measure bibliographic proximity and co-citation patterns, not community interactions such as workshops, discussions, or funding structures. The example of the 2017 Lorentz Center workshop leading to influential articles is asserted without evidence of a causal or even temporal link beyond co-occurrence. To make this category credible, the paper should state what specific observable would confirm or refute the claim that community engagement drove a particular advance, or explicitly acknowledge that this category is currently supported only by anecdote.","section":"§7.2 (and §2.3, category 4)"},{"comment":"The use of the authors' own AH12 review as the baseline snapshot, combined with the 2022 workshop sessions being arranged to mirror the 2010 meeting, introduces a selection channel: the same group that wrote the baseline review, organized the workshop, and wrote the current review also decides which advances count as \"progress\" and which driver category applies. This is a reasonable design for an expert review, but the paper should discuss this self-referential structure explicitly and identify what external evidence could disconfirm the driver classification. Without such a discussion, the driver analysis risks being circular, as the reader cannot independently assess whether the baseline choice biases the selection of progress examples.","section":"§1, §2, and §2.3"}],"minor_comments":[{"comment":"Figure 20 contains the visible placeholder text \"MAKE THIS POINT IN TEXT\", which is clearly an editing artifact and must be removed or implemented before publication.","section":"Figure 20"},{"comment":"There are several typographical errors, including \"occuring\" in the first sentence and \"Laser Inteferometer Space Antenna\" in the introductory section; these should be corrected.","section":"Introduction"},{"comment":"Equation (1) uses cgs units for the stellar density but does not state the units explicitly; the text should note that ρ* is in g cm^-3.","section":"Equation (1)"},{"comment":"The \"philosophical aside\" in footnote 13 interrupts the scientific narrative; consider moving this reflection to the main text or to a more appropriate location, or deleting it if space is limited.","section":"Footnote 13"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a review article, so the bar for quantitative novelty should be calibrated accordingly. However, the driver analysis is the paper's main new contribution beyond a conventional literature review, and the current gap between the abstract's causal language and the explicitly subjective method is a load-bearing issue. The paper can be made suitable by softening the causal claims and clearly labeling the driver identification as an expert hypothesis, which is why I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is a serious, comprehensive update of the Alexander & Hickox 2012 review, and it will probably become the standard citation for the state of AGN/SMBH growth in the 2020s. The core scientific sections are well organized, deeply referenced, and give an accurate picture of where the field stands: the AGN-as-event picture, the clumpy disk plus polar wind view of obscuration, the progress in accretion physics, simulations, and the census of AGN at high redshift. If you want one readable document that gets a non-specialist up to speed on a decade of progress, this is it. On that measure, the paper succeeds.\n\nThe genuinely new pieces are the citation-network map and the five-category driver taxonomy. The citation network is a useful visual and classification exercise, though I would echo the reproducibility concern: the construction is described but no data or code are provided, so the specific clustering is not independently checkable. Having the release be optional for a review, but it would strengthen the bibliometric claim.\n\nThe soft spot the stress-test flags is real but, in my view, less damaging than it is framed. The authors explicitly say they adopt a \"know it when you see it\" approach and that the five drivers are \"postulated.\" So the paper is transparent about the method, and the resulting attribution is a plausible, common-sense taxonomy rather than a hidden agenda. The bigger issue is the abstract: it says the paper identifies the drivers that enabled progress, which reads as a stronger claim than the evidence supports. The citation network does not test causal priority, and the labels in Table 1 are assigned post hoc. That's a legitimate caveat, but it applies to a secondary, meta-scientific component of the paper, not to the review's scientific content. The central scientific argument — that the last decade has produced major progress and this is what the progress looks like — holds up fine.\n\nOne more minor point: the driver attribution uses AH12 as the baseline and the Iceland workshop sessions mirror the 2010 meeting, so the selection of \"progress\" examples is structured by the authors' own framing. They note this, but it's worth keeping in mind when reading the progress summary.\n\nBottom line: this is a paper for the whole AGN community and for students entering the field. It deserves a serious referee and, with minor revision, publication. I would ask the authors to soften the abstract's causal claim about drivers, add a sentence acknowledging the post-hoc nature of the taxonomy, and consider releasing the citation-network data. I'm happy to engage with it and would cite it.","headline":"A thorough, authoritative update of AH12 that will serve as a standard reference; the driver-of-progress taxonomy is a transparently subjective add-on, not a demonstrated result, but the science review itself is solid.","tokens_in":52883,"tokens_out":1861,"would_cite":true,"duration_ms":23073,"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":"The central claim of this review is that the past decade established active galactic nuclei as transient, recurrent events within galaxy lifecycles, and that the field's progress was driven by five identifiable forces: new facilities…","keywords":["black holes","accretion","active galactic nuclei","quasars","feedback","galaxies","citation network analysis","scientific progress"],"falsifier":"Give an independent panel of AGN specialists the same 2012-vs-now comparison and ask them to attribute each landmark advance to one of the five driver categories; if inter-rater agreement is no better than chance, the driver taxonomy is not a stable empirical claim. Alternatively, apply the same citation-network method starting from a different decade baseline and ask whether the five categories still reproduce; systematic disagreement would falsify the taxonomy.","tokens_in":51954,"feed_emoji":"🌌","tokens_out":5449,"duration_ms":55562,"temperature":0.7,"pith_summary":"This review argues that the last decade produced a qualitative shift in how astronomers think about supermassive black hole growth: an active galactic nucleus is not a freak event but a recurrent, relatively short phase in the ordinary life of galaxies. It reports a new consensus that the obscuring material around accreting black holes is a dynamic, clumpy disk plus a polar wind, rather than a static doughnut. The review also attempts to explain the progress itself, using the authors' own 2012 review as a decade-old snapshot and a citation network of 60,924 papers to map the field. It concludes that progress in this area has been driven mainly by new facilities, larger datasets, new analysis techniques, community engagement, and conceptual shifts.","feed_headline":"AGN are short-lived events in galaxy lifecycles","feed_subtitle":"A decade of progress came from facilities, data, techniques, community, and conceptual shifts.","key_machinery":"The carrying objects are (1) the AGN-as-event framework, quantified by the probability distribution of specific black hole accretion rates $p(\\lambda_{\\rm sBHAR})$ (the AGN luminosity per unit host stellar mass), which explains why instantaneous AGN luminosity correlates weakly with star formation while average accretion tracks galaxy growth; (2) the dynamic obscurer model, in which a clumpy molecular disk and a polar dusty wind replace the static torus; and (3) a citation-network analysis built from 60,924 papers and 1,378,057 links, clustered with the Leiden algorithm, which lets the authors identify sub-fields and locate their own review relative to the field. Driving all three is the 'know it when you see it' method: specialists compare the present field against the 2012 snapshot and attribute progress to the five categories of Table 1.","core_discovery":"On the paper's own terms, the central claim is that the growth of supermassive black holes is best understood as a sequence of AGN 'events' inside galaxy lifecycles: galaxies repeatedly pass through short (<100 Myr) phases of strong accretion, separated by longer inactive periods, and the same galaxy can host many such events. A second claim is that the classic 'dusty torus' picture of AGN obscuration has been replaced by a more physical picture in which a clumpy, geometrically thin molecular disk supplies the fuel and a polar dusty cone, shaped by an AGN-driven wind, contributes much of the line-of-sight obscuration. A third claim is that the field's progress over the last decade is attributable to five identifiable drivers, and the citation network shows where this review and its predecessor sit within black-hole research.","pith_inferences":["An extension the authors leave implicit: the AGN-as-event picture implies that the local SMBH mass function is built by repeated short bursts rather than continuous growth, so models that enforce a fixed Eddington ratio will misestimate the scatter in the $M_{\\rm BH}$–$M_{\\star}$ relation.","A testable extension is to compare the $p(\\lambda_{\\rm sBHAR})$ distribution measured from deep X-ray surveys against the stochastic accretion variability in modern magnetohydrodynamic simulations; the simulations' duty cycles are a direct prediction of the event picture.","The citation-network method could be turned from retrospective into prospective: track whether newly commissioned facilities and upcoming time-domain surveys produce the predicted acceleration in AGN discovery rates, and whether the community-engagement driver shows up as citation bursts at specific workshops."],"forward_implications":["If AGN are recurrent short events, then single-epoch surveys systematically miss most SMBH growth, and interpreting AGN demographics requires duty cycles, not just triggering rates.","If the obscurer is a clumpy disk plus polar wind, then the simple Type 1/Type 2 unification dichotomy must be replaced by a distribution of covering factors tied to accretion state and wind activity.","If citation clusters track sub-fields, then the 13-cluster map of black-hole research can be used to forecast which areas are positioned to grow, such as time-domain AGN and high-redshift JWST samples.","If the five drivers are real, then deliberately investing in community events and cross-disciplinary workshops, alongside facilities and data, should measurably accelerate progress."],"supporting_citations":[{"why":"The 2012 review that this article updates; it serves as the decade-old snapshot used to measure progress.","marker":"Alexander and Hickox (2012)"},{"why":"Introduced the probability distribution of specific black hole accretion rates, the key tool behind the AGN-as-event picture.","marker":"Aird et al. (2012)"},{"why":"Showed that Myr-timescale AGN variability reconciles the weak AGN-star-formation correlation with an underlying average accretion-SFR relation.","marker":"Hickox et al. (2014)"},{"why":"Provided the polar-dust/wind picture of the parsec-scale obscurer, central to the revised unification model.","marker":"Hönig (2019)"},{"why":"Review of torus/obscuration observations whose schematic style the paper adopts; supplies evidence for the clumpy dynamic obscurer.","marker":"Ramos Almeida and Ricci (2017)"},{"why":"Demonstrated that citation clusters correspond to research sub-fields, the premise of the paper's citation-network analysis.","marker":"Klavans and Boyack (2017)"},{"why":"Supplies the Leiden algorithm used to detect the 13 clusters in the black-hole citation network.","marker":"Traag et al. (2019)"},{"why":"Used ionisation echoes (Voorwerpje) to estimate AGN event timescales, observational support for AGN as transient events.","marker":"Schawinski et al. (2015)"},{"why":"Established the SMBH-galaxy scaling relations that anchor the link between black-hole growth and host galaxy assembly.","marker":"Kormendy and Ho (2013)"}],"fun_headline_variants":["Black hole growth: a tale of short feeding frenzies","AGN are fleeting events, not permanent states","New picture: black holes grow in bursts with long quiet gaps","Clumpy gas and winds reshape black hole feeding model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The driver analysis rests on the assumption that scientific progress can be reliably recognised by specialists and fairly attributed to the five categories of Table 1, using the authors' own 2012 review as a baseline; if that judgement is subjective, the claimed identification of progress drivers is anecdotal rather than measured.","fun_headline_variants_meta":{"raw":{"variants":["Black hole growth: a tale of short feeding frenzies","AGN are fleeting events, not permanent states","New picture: black holes grow in bursts with long quiet gaps","Clumpy gas and winds reshape black hole feeding model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1438,"prompt_tokens":973,"completion_tokens":465,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":589,"completion_tokens_details":{"reasoning_tokens":411}},"tokens_in":589,"tokens_out":465,"duration_ms":5843,"temperature":1.0,"reasoning_tokens":411,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T23:07:08.309465+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Give an independent panel of AGN specialists the same 2012-vs-now comparison and ask them to attribute each landmark advance to one of the five driver categories; if inter-rater agreement is no better than chance, the driver taxonomy is not a stable empirical claim. Alternatively, apply the same citation-network method starting from a different decade baseline and ask whether the five categories still reproduce; systematic disagreement would falsify the taxonomy.","supporting_citations":[],"review_version":1}