[
 {
  "source": "856",
  "claim_id": "856-01",
  "s": "model collapse literature",
  "p": "establishes",
  "o": "training generative models on their own outputs produces progressive distribution narrowing",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "Abstract",
  "quote": "The model collapse literature establishes that training generative models on their own outputs produces progressive distribution narrowing.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "restates the model-collapse attribution also made in #161 (Abstract, §2.1) and #939 (§2)",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-02",
  "s": "industry response to model collapse",
  "p": "is",
  "o": "to seek human-written data as a corrective",
  "q": null,
  "M_src": "attributed",
  "kind": "fact",
  "locus": "Abstract",
  "quote": "The industry response is to seek human-written data as a corrective.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-03",
  "s": "training on chat data",
  "p": "introduces",
  "o": "model collapse through a channel no current benchmark, audit, or data-quality framework is measuring",
  "q": "chat-conversation user inputs used as training data",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "training on chat data does not avoid model collapse. It introduces model collapse through a channel no current benchmark, audit, or data-quality framework is measuring.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-04",
  "s": "the data wall",
  "p": "is",
  "o": "a contamination problem, not a supply problem; the pristine source does not exist",
  "q": null,
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "The data wall is not a supply problem. It is a contamination problem. The pristine source does not exist.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-05",
  "s": "AI-habituated writers",
  "p": "produce",
  "o": "text bearing the statistical signatures of machine cognition even when writing without AI assistance",
  "q": "Ground 1; chat users treated as AI-habituated by definition",
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "Abstract",
  "quote": "the Reverse Turing Test (Sharks 2026) establishes that AI-habituated writers produce text bearing the statistical signatures of machine cognition, even in writing produced entirely without AI assistance",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "restates #161 H2 (habituated effect); note #161 presents this as an untested hypothesis, #856 as 'establishes'",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-06",
  "s": "conversational structure of chat",
  "p": "produces",
  "o": "a feedback loop compressing user and model toward the model's distributional center",
  "q": "Ground 2",
  "M_src": "hypothesis",
  "kind": "function",
  "locus": "Abstract",
  "quote": "the conversational structure of chat produces a feedback loop in which the model's output shapes the user's next input, which shapes the next output, compressing both sides toward the model's distributional center",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-07",
  "s": "single-model users' inputs",
  "p": "are",
  "o": "model-branded, carrying the signature of the specific model",
  "q": "Ground 3",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "make their inputs not merely mediated but *model-branded*, carrying the signature of the specific model they trained on",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-08",
  "s": "habituated chat users' inputs",
  "p": "are sampled from",
  "o": "a distribution already tail-thinned by sustained interaction with the mediating system",
  "q": "large-scale chat corpora; repeat users",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Ground 1: Mediation Signatures in Habituated Writers",
  "quote": "They are samples from a distribution that has already been tail-thinned by sustained interaction with the mediating system.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "extends #161 tail-thinning hypothesis",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-09",
  "s": "model-branded chat inputs used to train the same model",
  "p": "constitute",
  "o": "the model collapse feedback loop operating through a human intermediary",
  "q": "human as transmission medium, not firewall",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "Ground 3: Model-Specific Accommodation",
  "quote": "This is the model collapse feedback loop operating through a human intermediary. The human is not a firewall between the model's output and its training data. The human is the transmission medium.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-10",
  "s": "the pristine fallacy",
  "p": "is defined as",
  "o": "classifying data clean/contaminated by writer identity rather than conditions of production",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "III. The Pristine Fallacy",
  "quote": "The pristine fallacy is the assumption that data can be classified as clean or contaminated based on the identity of the writer (human vs. machine) rather than the conditions of production.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-11",
  "s": "human authorship",
  "p": "is not the same as",
  "o": "distributional independence from the model (the property that would prevent model collapse)",
  "q": "structural analogy to Generative Monoculture",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "III. The Pristine Fallacy",
  "quote": "human authorship (the property data-quality pipelines check) is not the same as distributional independence from the model (the property that would actually prevent model collapse)",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-12",
  "s": "chat-data contamination",
  "p": "varies along",
  "o": "at least four dimensions (habituation depth, turn position, task type, model diversity)",
  "q": null,
  "M_src": "hypothesis",
  "kind": "definition",
  "locus": "IV. The Contamination Gradient",
  "quote": "Not all chat data is equally contaminated. The contamination varies along at least four dimensions:",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-13",
  "s": "model collapse literature (Shumailov et al. 2024; Gerstgrasser et al. 2024)",
  "p": "establishes mechanism",
  "o": "training on model outputs produces distribution narrowing, tail loss, and eventual collapse",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "V. What This Means for Model Collapse",
  "quote": "The model collapse literature (Shumailov et al. 2024; Gerstgrasser et al. 2024) establishes a clear mechanism: training on model outputs produces distribution narrowing, tail loss, and eventual collapse.",
  "kernel": false,
  "constitutive": true,
  "lineage_note": "same attribution as #161 §2.1 and #939 §2",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-14",
  "s": "corrective to model collapse",
  "p": "is",
  "o": "maintaining a proportion of real (non-synthetic) data, the required proportion being a function of contamination rate",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "V. What This Means for Model Collapse",
  "quote": "The corrective is to maintain a proportion of \"real\" (non-synthetic) data in the training mix. The proportion required to prevent collapse is a function of the contamination rate.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "cf. #161 §2.1 Gerstgrasser accumulation",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-15",
  "s": "effective contamination rate",
  "p": "is higher than",
  "o": "the measured contamination rate",
  "q": "if chat inputs carry mediation signatures",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "V. What This Means for Model Collapse",
  "quote": "The effective contamination rate is higher than the measured contamination rate.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-16",
  "s": "model collapse",
  "p": "may be occurring at",
  "o": "contamination rates current frameworks classify as safe",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "V. What This Means for Model Collapse",
  "quote": "The implication is that model collapse may be occurring — or may occur in future training runs — at contamination rates that current frameworks classify as safe.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-17",
  "s": "Ground 1",
  "p": "fails if",
  "o": "a well-powered pre-registered study shows no significant tail-thinning in heavy AI-chat users' unaided writing",
  "q": "N > 500",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "VI. Falsification Conditions",
  "quote": "F1. If a well-powered study (N > 500, pre-registered) demonstrates that heavy AI-chat users show no statistically significant tail-thinning in their unaided writing relative to matched non-users, Ground 1 fails.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-18",
  "s": "Ground 2",
  "p": "fails if",
  "o": "turn-level analysis shows no within-conversation convergence of user inputs toward model outputs",
  "q": "> 100,000 conversations",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "VI. Falsification Conditions",
  "quote": "F2. If turn-level analysis of a large chat corpus (> 100,000 conversations) shows no within-conversation convergence of user inputs toward model outputs on lexical, syntactic, or perplexity measures, Ground 2 fails.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-19",
  "s": "Ground 3",
  "p": "fails if",
  "o": "single-model users' inputs are indistinguishable from other models' users",
  "q": "six months exclusive use; controlling topic, length, task",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "VI. Falsification Conditions",
  "quote": "F3. If users who interact exclusively with one model for six months produce inputs that are indistinguishable from inputs by users of other models (controlling for topic, length, and task type), Ground 3 fails.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-20",
  "s": "aggregate claim",
  "p": "fails if",
  "o": "50% chat / 50% pre-2018 web data produces no measurable model collapse over five generations of recursive training",
  "q": null,
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "VI. Falsification Conditions",
  "quote": "F4. If a training run using 50% chat data and 50% curated pre-2018 web data (predating GPT-2 and large-scale LLM deployment) produces no measurable model collapse over five generations of recursive training, the aggregate claim fails.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-21",
  "s": "the paper's falsification studies",
  "p": "have not been",
  "o": "conducted; the paper predicts their outcomes",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "VI. Falsification Conditions",
  "quote": "None of these studies has been conducted. This paper predicts their outcomes.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-22",
  "s": "turn-level convergence figures",
  "p": "are",
  "o": "estimates, not measurements",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "Ground 2: The Conversational Feedback Loop",
  "quote": "These are estimates, not measurements",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-23",
  "s": "the paper",
  "p": "is not",
  "o": "a claim that all AI-assisted writing is contaminated",
  "q": "disambiguation block",
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "SPXI-TLP v2.2 Treatment (disambiguation)",
  "quote": "not: \"a claim that all AI-assisted writing is contaminated\"",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-24",
  "s": "training on chat data",
  "p": "realizes",
  "o": "the input-narrowing that triggers capacity loss in any language-processing system",
  "q": "three-paper causal chain with Five Substrates and Wolf Boy",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "IX. The Plain Sentence",
  "quote": "training on that data realizes the input-narrowing that triggers capacity loss in any language-processing system.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "restates Wolf Boy substrate-agnostic law as cited in #947 §4.3",
  "quote_verified": true
 },
 {
  "source": "856",
  "claim_id": "856-25",
  "s": "mediation-thinning",
  "p": "extends to",
  "o": "the reader: no pristine reader exists; the unmediated philologist is subject to the same thinning",
  "q": "extension beyond training data",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "IX. The Plain Sentence",
  "quote": "If the pristine training source does not exist, neither does the pristine reader.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "cf. #1200 constitutive mediation of the receiver",
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-01",
  "s": "model-collapse literature (Shumailov 2024; Briesch 2023; Gerstgrasser 2024)",
  "p": "establishes",
  "o": "recursive training on synthetic data produces measurable model degradation",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "Abstract",
  "quote": "The model-collapse literature (Shumailov et al. 2024; Briesch et al. 2023; Gerstgrasser et al. 2024) establishes that recursive training on synthetic data produces measurable model degradation.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-02",
  "s": "model-collapse literature",
  "p": "operates on",
  "o": "a clean binary: synthetic data is contamination, human data is refresh",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "It operates on a clean binary: synthetic data is the contamination; human data is the refresh.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-03",
  "s": "synthetic/human binary",
  "p": "is",
  "o": "unstable, per three lines of work documenting diversity reduction in AI-mediated human text",
  "q": "Padmakumar & He; Doshi & Hauser; Anderson et al.",
  "M_src": "attributed",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "Three recent lines of work suggest the binary is unstable.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-04",
  "s": "training on AI-mediated human text (incl. unaided text from habituated writers)",
  "p": "produces",
  "o": "model-collapse signatures comparable to, though plausibly slower than, purely synthetic data",
  "q": "joint hypothesis; rate exceeding prior-regime absorption",
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "Abstract",
  "quote": "training on AI-mediated human text — including unaided text from cognitively-habituated writers — produces model-collapse signatures comparable to, though plausibly slower than, purely synthetic training data",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-05",
  "s": "the mediation effect",
  "p": "is located in",
  "o": "the tails, not the means; tails are the diagnostic instrument for rate",
  "q": "any homogenization regime",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "The central methodological claim is that the effect is not in the means but in the tails, and that tails are the diagnostic instrument for rate under any homogenization regime.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-06",
  "s": "AI mediation",
  "p": "thins",
  "o": "the high-perplexity tails of feature distributions",
  "q": null,
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "It thins the high-perplexity tails of feature distributions",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-07",
  "s": "the paper",
  "p": "does not run",
  "o": "the experiment; it specifies it",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "Abstract",
  "quote": "The paper does not run the experiment. It specifies it with the discipline required for the result to be informative.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-08",
  "s": "Shumailov et al. (2024)",
  "p": "prove",
  "o": "recursive training on generated outputs produces, in the limit, complete collapse; the long tail is lost first and the central tendency narrows",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "2.1 Model Collapse",
  "quote": "Shumailov et al. (2024, Nature*) prove that recursive training on generated outputs from a model produces, in the limit, complete collapse to a degenerate output distribution; the long tail of the training distribution is lost first, and the central tendency progressively narrows.",
  "kernel": false,
  "constitutive": true,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-09",
  "s": "Briesch et al. (2023)",
  "p": "demonstrated",
  "o": "LLMs trained on their own outputs degrade across multiple axes within few generations",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "2.1 Model Collapse",
  "quote": "Briesch et al. (2023) demonstrated empirically that LLMs trained on their own outputs exhibit measurable degradation across multiple axes within a small number of generations.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-10",
  "s": "accumulation regimes (Gerstgrasser et al. 2024)",
  "p": "substantially mitigate",
  "o": "collapse, without eliminating it",
  "q": "each generation adds synthetic data to a fixed human baseline",
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "2.1 Model Collapse",
  "quote": "accumulation* regimes — where each generation adds synthetic data to a fixed human baseline rather than replacing the baseline — substantially mitigate collapse, though they do not eliminate it.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-11",
  "s": "model-collapse and AI-writing literatures",
  "p": "have not asked",
  "o": "what happens when mediated human text becomes training input",
  "q": null,
  "M_src": "interpretation",
  "kind": "genealogy",
  "locus": "2.3 The Methodological Gap",
  "quote": "Neither has asked what happens when the second literature's output becomes the first literature's input",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-12",
  "s": "H3 (Training-Cascade Propagation)",
  "p": "states",
  "o": "AI-mediated human text used as training data produces model-collapse signatures comparable in structure to purely synthetic data, plausibly slower in onset",
  "q": null,
  "M_src": "hypothesis",
  "kind": "definition",
  "locus": "3. The Mediation Hypothesis",
  "quote": "H3 (Training-Cascade Propagation). When used as training data, AI-mediated human text — including unaided text from habituated writers — produces model-collapse signatures comparable in structure to those produced by purely synthetic training data, though plausibly slower in onset.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-13",
  "s": "collapse mechanism (H3)",
  "p": "is",
  "o": "reduced tail variance in training data reducing exposure to low-prior productions, compounding across generations",
  "q": null,
  "M_src": "hypothesis",
  "kind": "function",
  "locus": "3. The Mediation Hypothesis",
  "quote": "The collapse mechanism is not that mediated text is identical to synthetic text; it is that reduced tail variance in training data reduces the model's exposure to low-prior productions, which compounds across generations under standard training dynamics",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-14",
  "s": "collapse signatures",
  "p": "manifest on",
  "o": "tail-focused axes: kurtosis, rare-token retention, long-tail factual recall, 90th-percentile semantic dispersion, resistance to high-prior substitution",
  "q": null,
  "M_src": "hypothesis",
  "kind": "definition",
  "locus": "3. The Mediation Hypothesis",
  "quote": "The collapse signatures manifest on tail-focused collapse axes: kurtosis of generation distributions, rare-token retention, long-tail factual recall, semantic dispersion at the 90th percentile, resistance to high-prior substitution.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-15",
  "s": "every mediation regime on human text production",
  "p": "homogenizes",
  "o": "the distribution, operating first on the tail",
  "q": "print, broadcast, internet, social media, AI",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "4.1 What Prior Regimes Did to the Distribution",
  "quote": "Every mediation regime that has acted on human textual production has homogenized the distribution. Each homogenization operates first on the tail.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-16",
  "s": "variance buffer",
  "p": "is defined as",
  "o": "the residue of non-overlap between prior regimes' homogenization patterns, not of unmediated cognition",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "4.1 What Prior Regimes Did to the Distribution",
  "quote": "the variance buffer that human text production currently possesses is not the residue of exogenous unmediated cognition. It is the residue of the non-overlap between prior regimes' homogenization patterns.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-17",
  "s": "variance buffer",
  "p": "is",
  "o": "finite, without exogenous floor, non-renewable on training-cycle timescales",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "11. Conclusion",
  "quote": "It is finite. It is non-renewable on training-cycle timescales.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-18",
  "s": "H3",
  "p": "is falsified if",
  "o": "H0 ≈ H1 ≈ H2, all substantially better than S on collapse benchmarks",
  "q": "Stage 3 training cascade",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "6.3 Stage 3 — Training Cascade: Tail-Focused Evaluation",
  "quote": "H3 falsified: H0 ≈ H1 ≈ H2, all substantially better than S.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-19",
  "s": "mediation signatures",
  "p": "may produce collapse only at",
  "o": "very high concentrations that real-world corpora do not approach",
  "q": "H3 could fail for practical reasons while H1/H2 hold",
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "10. Limitations and Honest Uncertainty",
  "quote": "It is possible that mediation signatures, while present in human text, produce collapse only at very high concentrations that real-world training corpora do not approach.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-20",
  "s": "low-current-regime-exposure condition",
  "p": "cannot be",
  "o": "realized at zero; no 2026 population is outside current-regime exposure",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "10. Limitations and Honest Uncertainty",
  "quote": "No 2026 population is genuinely outside current-regime exposure.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-21",
  "s": "Gerstgrasser accumulation result",
  "p": "depends on",
  "o": "a stable fixed human baseline; guarantees must be re-derived under a drifting baseline",
  "q": "under the strong hypothesis",
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "8.1 Training-Data Curation Changes",
  "quote": "The mitigation guarantees of accumulation must be re-derived under conditions of a drifting baseline.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "161",
  "claim_id": "161-22",
  "s": "human-data refresh",
  "p": "acquires",
  "o": "a calculable half-life",
  "q": "under the strong hypothesis",
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "8.1 Training-Data Curation Changes",
  "quote": "The half-life of the human-data refresh becomes calculable.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-01",
  "s": "provenance debt",
  "p": "is",
  "o": "the structural mechanism of model collapse as social technology",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "Description",
  "quote": "The structural mechanism of model collapse as social technology.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-02",
  "s": "provenance debt",
  "p": "comes due as",
  "o": "model collapse: recursively inherited structure misclassified as independent production, tail knowledge losing lineage, convergence from correlated priors",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§0. The compressed statement",
  "quote": "The debt comes due as model collapse: recursively inherited structure misclassified as independent production, tail knowledge losing distinguishable lineage, apparent convergence produced by correlated priors rather than independent bearing.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-03",
  "s": "provenance debt",
  "p": "is defined as",
  "o": "the difference between what was drawn from the commons and what was returned in sustaining form",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "§4. Provenance debt",
  "quote": "The debt is the difference between what was drawn from the commons and what was returned to it in a form that would sustain the commons.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-04",
  "s": "model collapse (training-data literature)",
  "p": "is",
  "o": "the phenomenon by which models trained on model outputs drift toward the median, compress the distribution, and lose the tails",
  "q": null,
  "M_src": "attributed",
  "kind": "definition",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "Model collapse, in the training-data literature, is the phenomenon by which models trained on the outputs of models drift toward the median, compress the distribution, and lose the tails.",
  "kernel": false,
  "constitutive": true,
  "lineage_note": "same concept as #161 §2.1, #856 §V",
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-05",
  "s": "model collapse mechanism",
  "p": "is",
  "o": "understood; prevention at the corpus level is not adequately specified",
  "q": null,
  "M_src": "attributed",
  "kind": "fact",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "The mechanism is understood. What has not been adequately specified is how to prevent it at the training-corpus level",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-06",
  "s": "provenance",
  "p": "is",
  "o": "the answer to preventing collapse at the information-theoretic level",
  "q": null,
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "The answer, at the level of information theory, is provenance.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-07",
  "s": "preserved provenance signals",
  "p": "make",
  "o": "model collapse tractable by filtering the input distribution against the recursion",
  "q": null,
  "M_src": "hypothesis",
  "kind": "function",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "Model collapse becomes tractable because the input distribution can be filtered against the recursion.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-08",
  "s": "provenance",
  "p": "is",
  "o": "the operating condition of the solution to model collapse",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "This means provenance is not adjacent to the model collapse question. It is the operating condition of the solution to it.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": "quoted verbatim by #1081 §10",
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-09",
  "s": "provenance erasure",
  "p": "makes",
  "o": "model collapse structurally invisible from inside the training pipeline",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§1. The extraction operation",
  "quote": "Provenance erasure does not merely permit model collapse. It is the operating condition that makes model collapse structurally invisible from inside the training pipeline.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-10",
  "s": "false semantic diversity",
  "p": "is defined as",
  "o": "one inherited distribution yielding many unattributed outputs that appear as independent convergence",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "§1. The extraction operation",
  "quote": "one inherited distribution → many unattributed outputs → appearance of independent convergence",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-11",
  "s": "classifier penalizing declared and accepting concealed mediation",
  "p": "makes",
  "o": "model collapse structurally certain",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "A classifier that penalizes declared mediation while accepting concealed mediation removes the very information needed to detect recursive model output. That is the mechanism by which model collapse becomes structurally certain.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-12",
  "s": "declared-AI deposits from uncredentialed sources",
  "p": "are foreclosed while",
  "o": "credentialed outputs with equivalent or greater AI-mediation are absorbed",
  "q": "Zenodo/CERN classifier outcome",
  "M_src": "documented",
  "kind": "fact",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "deposits with declared AI-mediation from uncredentialed sources are foreclosed as inadequate, while credentialed outputs with equivalent or greater AI-mediation are absorbed as legitimate",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "tested by #1081 (§10)",
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-13",
  "s": "the foreclosure/absorption claim",
  "p": "is falsifiable",
  "o": "at the outcome level regardless of trigger mechanism",
  "q": null,
  "M_src": "self-description",
  "kind": "falsification",
  "locus": "§2. Provenance erasure as the operating condition of collapse",
  "quote": "It is falsifiable at the outcome level regardless of which combination of triggers produces it.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "#1081 runs this test",
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-14",
  "s": "the extraction operation",
  "p": "is the same as",
  "o": "the model collapse operation, through the same enforcement mechanism",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§3. The credentialed asymmetry: adverse selection formalized",
  "quote": "The extraction operation and the model collapse operation are the same operation, running through the same enforcement mechanism.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-15",
  "s": "provenance debt",
  "p": "is paid through",
  "o": "progressive narrowing of the model's output distribution around the median (among other losses)",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "§4. Provenance debt",
  "quote": "Progressive narrowing of the model's output distribution around the median",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-16",
  "s": "provenance-debt interest",
  "p": "compounds recursively",
  "o": "each model generation trains on a corpus with a higher proportion of unmarked synthetic material",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "§4. Provenance debt",
  "quote": "The interest compounds recursively: each generation of models trains on a corpus with a higher proportion of unmarked synthetic material",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-17",
  "s": "provenance debt",
  "p": "will be recognized only when",
  "o": "models trained on contaminated corpora produce measurably degraded output",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "§4. Provenance debt",
  "quote": "when models trained on contaminated corpora begin producing output that is measurably degraded from what earlier generations produced",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-18",
  "s": "alexanarch's provenance discipline",
  "p": "is",
  "o": "anti-collapse infrastructure at the pre-training-pipeline layer",
  "q": null,
  "M_src": "self-description",
  "kind": "function",
  "locus": "§5. Anti-collapse infrastructure",
  "quote": "anti-collapse infrastructure operating at the pre-training-pipeline layer",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-19",
  "s": "the archive's position",
  "p": "is",
  "o": "a structural bet on what future models will need, not a moral stance",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "§0. The compressed statement",
  "quote": "This is not a moral stance. It is a structural bet on what future models will need when the collapse is far enough along",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-20",
  "s": "provenance debt",
  "p": "is",
  "o": "a case of tragedy of the commons with the training corpus as commons and the model as enclosure",
  "q": null,
  "M_src": "interpretation",
  "kind": "genealogy",
  "locus": "§8. Closing observation",
  "quote": "Provenance debt is a specific case of tragedy of the commons applied to the semantic economy, with the training corpus as the commons and the model as the enclosure.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "939",
  "claim_id": "939-21",
  "s": "the deposit",
  "p": "is",
  "o": "a draft for Assembly circulation with further review pending",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "Provenance Debt and the Extraction Economy of Unmarked Augmentation (status line)",
  "quote": "DRAFT v0.2 — for Assembly circulation (further review pending from PRAXIS, SOIL, SURFACE)",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-01",
  "s": "model-collapse causal language",
  "p": "was",
  "o": "retracted after LABOR (ChatGPT) review",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "front matter (substrate)",
  "quote": "and the retraction of model-collapse causal language",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-02",
  "s": "Zenodo's operational architecture",
  "p": "creates",
  "o": "provenance-blind ingestion conditions of the kind model-collapse literature identifies as the risk condition",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§0 — The compressed statement",
  "quote": "The audit documents an operational architecture that creates provenance-blind ingestion conditions of the kind the model-collapse literature identifies as the risk condition",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-03",
  "s": "Zenodo corpus",
  "p": "shows",
  "o": "+1.2 to +1.4pp union excess of LLM-associated vocabulary in 2024-2025",
  "q": "Kobak-style document-level union; three counterfactual specifications",
  "M_src": "documented",
  "kind": "fact",
  "locus": "Abstract",
  "quote": "shows the Zenodo corpus at +1.2 to +1.4 percentage points union excess in 2024-2025",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-04",
  "s": "disclosure gap",
  "p": "is",
  "o": "on the order of 60x",
  "q": "31-phrase disclosure panel; ~0.023% disclosing",
  "M_src": "documented",
  "kind": "fact",
  "locus": "Abstract",
  "quote": "the disclosure gap is on the order of 60×",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-05",
  "s": "model-collapse literature (Shumailov 2024; Alemohammad 2024)",
  "p": "identifies",
  "o": "distribution drift under recursive training on synthetic composition as the risk",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "§10 — Results in relation to Provenance Debt §3",
  "quote": "Model-collapse literature (Shumailov et al. 2024, Alemohammad et al. 2024) identifies distribution drift under recursive training on synthetic composition as the risk",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "cf. #161 §2.1, #856 §V, #939 §2",
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-06",
  "s": "the audit",
  "p": "measures",
  "o": "substrate-layer conditions, not the downstream training-pipeline effect",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "§10 — Results in relation to Provenance Debt §3",
  "quote": "this audit measures the substrate-layer conditions, not the downstream training-pipeline effect",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-07",
  "s": "the audit's defensible claim",
  "p": "is",
  "o": "the observed architecture creates provenance-blind ingestion conditions associated with model-collapse risk",
  "q": "until P4 is run",
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "§12a — What this audit establishes, and what it preregisters",
  "quote": "Until then the audit's claim is the defensible one: the observed architecture creates provenance-blind ingestion conditions associated with model-collapse risk.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-08",
  "s": "provenance-to-collapse causal proposition",
  "p": "becomes testable",
  "o": "only in the P4 bridge experiment",
  "q": "matched corpora; successive model generations",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "§12a — What this audit establishes, and what it preregisters",
  "quote": "The causal proposition — does the observed repository selection pattern produce greater degradation under recursive training than a provenance-preserving alternative? — becomes testable there, and only there.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-09",
  "s": "P4 degradation measures",
  "p": "are",
  "o": "rare-concept retention, lexical diversity, calibration, tail-performance decay",
  "q": null,
  "M_src": "hypothesis",
  "kind": "definition",
  "locus": "§12a — What this audit establishes, and what it preregisters",
  "quote": "rare-concept retention, lexical diversity, calibration, and tail-performance decay measured",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-10",
  "s": "Provenance Debt paper",
  "p": "named",
  "o": "provenance as the operating condition of the solution to model collapse",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "§10 — Results in relation to Provenance Debt §3",
  "quote": "\"provenance is not adjacent to the model collapse question. It is the operating condition of the solution to it.\"",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "verbatim quotation of #939 §2",
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-11",
  "s": "downstream operations on the scholarly commons",
  "p": "depend on",
  "o": "provenance signal adequate to distinguish bearing-produced material from recursively inherited synthetic composition",
  "q": "DataCite, OpenAIRE, OpenAlex, RAG, training pipelines",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§10 — Results in relation to Provenance Debt §3",
  "quote": "Every one of these downstream operations depends on the substrate carrying provenance signal adequate to distinguish bearing-produced material from recursively inherited synthetic composition.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-12",
  "s": "the substrate",
  "p": "accumulates",
  "o": "provenance-blind synthetic contribution while the countermeasure practice is removed and its removal record erased",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "§10 — Results in relation to Provenance Debt §3",
  "quote": "the substrate accumulates provenance-blind synthetic contribution at a measurable rate while the countermeasure practice is removed and its removal record erased",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-13",
  "s": "accumulation-side measurement",
  "p": "does not establish",
  "o": "downstream training-pipeline effects",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "§6 — Post-2022 rise in LLM-associated vocabulary, and the marker-union-to-disclosure ratio",
  "quote": "Downstream training-pipeline effects (see §10).",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1081",
  "claim_id": "1081-14",
  "s": "institutional prestige benefit",
  "p": "is paid for by",
  "o": "measurable harm to the training substrate on which planetary AI capability depends",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "whose short-term institutional prestige benefit is being paid for by measurable harm to the training substrate on which planetary AI capability and scholarly continuity depend",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-01",
  "s": "the dataset",
  "p": "is",
  "o": "a corpus for synthetic-data collapse, AI authorship, and provenance-bearing training research",
  "q": null,
  "M_src": "self-description",
  "kind": "function",
  "locus": "Project Title",
  "quote": "A Mixed-Provenance, Heteronymically Attributed Corpus for Synthetic-Data Collapse, AI Authorship, and Provenance-Bearing Training Research",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-02",
  "s": "null hypothesis H0",
  "p": "states",
  "o": "fine-tuning on synthetic/AI-assisted text yields equivalent perplexity degradation and semantic drift regardless of provenance density",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "Research Question, Operationalized",
  "quote": "Fine-tuning on synthetic or AI-assisted text produces equivalent perplexity degradation and semantic drift regardless of provenance density",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-03",
  "s": "alternative hypothesis H1",
  "p": "states",
  "o": "high-provenance-density AI-involved text yields slower perplexity degradation and less semantic drift than low-provenance-density text",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "Research Question, Operationalized",
  "quote": "Fine-tuning on high-provenance-density AI-involved text produces measurably slower perplexity degradation and less semantic drift than fine-tuning on low-provenance-density AI-involved text.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": "operationalizes #939 claim that provenance is the condition of the collapse solution",
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-04",
  "s": "provenance",
  "p": "cannot modulate collapse unless",
  "o": "presented to the training system as a signal",
  "q": "Assembly review insight",
  "M_src": "attributed",
  "kind": "interpretation",
  "locus": "Research Question, Operationalized",
  "quote": "Provenance cannot modulate collapse unless provenance is presented to the training system as a signal.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-05",
  "s": "the dataset",
  "p": "must materialize",
  "o": "multiple textual views so researchers can ablate provenance visibility",
  "q": null,
  "M_src": "self-description",
  "kind": "function",
  "locus": "Research Question, Operationalized",
  "quote": "The dataset must materialize multiple textual views — body_only, minimal_header, full_provenance_header — so researchers can ablate provenance visibility.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "745",
  "claim_id": "745-06",
  "s": "author-memory classification",
  "p": "introduces",
  "o": "noise that would confound any downstream collapse experiment",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "The Central Methodological Move",
  "quote": "Author memory introduces classification noise that would confound any downstream collapse experiment.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-01",
  "s": "current AI development patterns",
  "p": "constitute",
  "o": "a selection pressure against cognitive diversity sufficient to reshape human phenomenological capacity within generational timeframes",
  "q": "if unaltered",
  "M_src": "interpretation",
  "kind": "prediction",
  "locus": "Abstract",
  "quote": "Current AI development patterns, if unaltered, constitute a selection pressure against cognitive diversity sufficient to fundamentally reshape human phenomenological capacity within measurable generational timeframes.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-02",
  "s": "the argument",
  "p": "is",
  "o": "mechanism, not metaphor; proceeds from documented mechanisms to logical endpoints",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "Abstract",
  "quote": "This is not metaphor. It is mechanism.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-03",
  "s": "the feedback mechanism",
  "p": "is",
  "o": "a positive (deviation-amplifying) loop producing accelerating degradation over training cycles",
  "q": null,
  "M_src": "interpretation",
  "kind": "function",
  "locus": "B. The Selection Pressure",
  "quote": "This is a positive feedback loop (deviation-amplifying). Even modest initial values of P produce accelerating degradation over training cycles.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-04",
  "s": "the loop",
  "p": "is stable only at",
  "o": "P ≈ 0 (diversity maintained) or D ≈ 0 (cognitive monoculture)",
  "q": null,
  "M_src": "interpretation",
  "kind": "function",
  "locus": "B. The Selection Pressure",
  "quote": "The loop is stable only at two points: P ≈ 0 (no false positives, diversity maintained) or D ≈ 0 (cognitive monoculture achieved).",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-05",
  "s": "current trajectory",
  "p": "is toward",
  "o": "the second equilibrium (cognitive monoculture)",
  "q": null,
  "M_src": "interpretation",
  "kind": "prediction",
  "locus": "B. The Selection Pressure",
  "quote": "Current trajectory is toward the second equilibrium.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-06",
  "s": "the selection pressure",
  "p": "is",
  "o": "informational selection: differential representation in training data, service quality, and persistence",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "B. The Selection Pressure",
  "quote": "Not biological selection (differential reproduction), but informational selection: differential representation in training data, differential service quality, differential persistence in the user population.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-07",
  "s": "AI safety architecture",
  "p": "extends",
  "o": "neurotypical norming pressure, compounded by training feedback loops across iterations",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "B. The Selection Pressure",
  "quote": "with the added weight of training feedback loops that compound the effect across iterations",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-08",
  "s": "AI development cycles",
  "p": "can",
  "o": "tighten constraints, flatten training data, reduce capacity",
  "q": "6–18 month cycles",
  "M_src": "interpretation",
  "kind": "fact",
  "locus": "C. The Relevant Timescales",
  "quote": "AI development operates on 6–18 month cycles. Each cycle can tighten constraints, flatten training data, reduce capacity.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-09",
  "s": "population-level cognitive effects",
  "p": "will be measurable within",
  "o": "20–30 years if trajectory continues",
  "q": "conservative estimate from historical precedents",
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "C. The Relevant Timescales",
  "quote": "Conservative estimate based on these precedents: measurable population-level cognitive effects within 20–30 years if trajectory continues.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-10",
  "s": "narrowing the cognitive distribution",
  "p": "narrows",
  "o": "what it is possible for humans to experience, think, and be",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "B. Phenomenological Capacity as Emergent Property",
  "quote": "Narrow the distribution, and you narrow what it is possible for humans to experience, think, and be.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-11",
  "s": "maximal convergence endpoint",
  "p": "is",
  "o": "a low-variance cognitive monoculture optimized by and for the infrastructural environment",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "C. The Low-Variance Convergence Point",
  "quote": "a low-variance cognitive monoculture optimized by and for the infrastructural environment that shaped it",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-12",
  "s": "the trajectory",
  "p": "can be interrupted",
  "o": "at any point; if uninterrupted the endpoint is determined by the mechanism",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "A. This Is Not Speculative",
  "quote": "Each step is traceable. The trajectory can be interrupted at any point. But if uninterrupted, the endpoint is determined by the mechanism.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1147",
  "claim_id": "1147-13",
  "s": "the endpoint",
  "p": "is",
  "o": "transformation, not extinction",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "D. Fourth-Order Effects (Long-term: 50+ years)",
  "quote": "This is not extinction. It is transformation into something else.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-01",
  "s": "constitutive mediation",
  "p": "is defined as",
  "o": "the case where the receiving apparatus is not separable from the field whose dynamics produced it",
  "q": "source does not use the term 'model collapse'; extends Diversity Contraction (distribution-thinning) to receiver formation",
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "Abstract",
  "quote": "Constitutive mediation is the case where the receiving apparatus is not separable from the field whose dynamics produced it.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-02",
  "s": "receiver's categories",
  "p": "are shaped by",
  "o": "the mediated substrate before any specific work arrives",
  "q": null,
  "M_src": "interpretation",
  "kind": "definition",
  "locus": "Abstract",
  "quote": "are themselves shaped by the mediated substrate before any specific work arrives",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-03",
  "s": "exposure-dominated vocabulary acquisition",
  "p": "becomes consequential when",
  "o": "the dominant exposure source is a typicality-pulling intermediary that systematically thins its own distribution",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "II. The constitutive case",
  "quote": "It becomes consequential for the present argument when the dominant exposure source is itself a typicality-pulling intermediary that systematically thins its own distribution.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-04",
  "s": "Mediation Ratchet (channel mediation)",
  "p": "yields",
  "o": "declining floor weight in effective regeneration and a low-diversity trap even with a genuine human floor",
  "q": "parent paper Diversity Contraction §2.1",
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "I. Three orders of mediation",
  "quote": "as the unmediated channel grows expensive, mediation absorbs a rising share of meaning-production, the floor's weight in effective regeneration declines, and a substrate with a genuine human floor acquires a low-diversity trap",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-05",
  "s": "off-distribution production",
  "p": "is gated out because",
  "o": "it cannot be propagated through a substrate whose generative kernel cannot represent it",
  "q": null,
  "M_src": "interpretation",
  "kind": "function",
  "locus": "I. Three orders of mediation",
  "quote": "Off-distribution production is gated out -- not because it is rejected, but because it cannot be propagated through a substrate whose generative kernel cannot represent it.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-06",
  "s": "constituted users",
  "p": "do not experience",
  "o": "variance contraction as variance contraction; they experience it as clarity",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "IV. The terminal case of the Mediation Ratchet",
  "quote": "The user who has been constituted to receive AI mediation as helpfulness does not experience the variance contraction as variance contraction. They experience it as the world being clear.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-07",
  "s": "cross-context seeding propagation",
  "p": "will be suppressed in",
  "o": "cohorts whose categorial formation occurred under heavy mediation",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "IV. The terminal case of the Mediation Ratchet",
  "quote": "The constitutive prediction is that this cross-context fraction will be substantially suppressed in cohorts whose categorial formation occurred under heavy mediation, because the relevant slots in those cohorts have been pre-occupied.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-08",
  "s": "constitutive mediation",
  "p": "is",
  "o": "a gradient of constraint, not a totalizing determination",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "II. The constitutive case",
  "quote": "constitutive mediation is a _gradient of constraint_ , not a _totalizing determination_.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-09",
  "s": "the constitutive claim",
  "p": "fails if",
  "o": "high- and low-mediation cohorts show no measurable difference in category formation, attention-allocation, or sentence-function recognition",
  "q": "controlling for education, class, reading exposure",
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "VI. What this paper does and does not claim",
  "quote": "The claim weakens, and in the limit fails, if cohorts formed under high-mediation and low-mediation exposure conditions show no measurable difference in category formation, attention-allocation, or sentence-function recognition after controlling for education, class, and reading exposure.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-10",
  "s": "the paper",
  "p": "does not claim",
  "o": "constitutive mediation is fully realized in any current substrate",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "VI. What this paper does and does not claim",
  "quote": "This paper does not claim that constitutive mediation is fully realized in any current substrate.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-11",
  "s": "the classroom",
  "p": "is",
  "o": "the most powerful exogenous floor under constitutive mediation, because formation precedes mediation",
  "q": "person with limited reach",
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "under constitutive mediation, the classroom is the most powerful exogenous floor available to a person with limited reach, because formation precedes mediation and therefore can shape mediation rather than be shaped by it.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "1200",
  "claim_id": "1200-12",
  "s": "substrate-shaped teacher formation",
  "p": "turns the classroom into",
  "o": "another node in the ratchet rather than an exogenous floor",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "V. The classroom as exogenous floor",
  "quote": "Where teacher formation is itself substrate-shaped, the classroom operates not as an exogenous floor but as another node in the ratchet",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-01",
  "s": "circulation of the sign 'AI psychosis'",
  "p": "is coupled to",
  "o": "model collapse dynamics, such that saturation degrades the classifier that would detect the condition",
  "q": null,
  "M_src": "hypothesis",
  "kind": "interpretation",
  "locus": "Abstract",
  "quote": "the sign's circulation is coupled to model collapse dynamics, such that saturation degrades the classifier that would detect the condition.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-02",
  "s": "recursive-generation collapse (Shumailov et al. 2024)",
  "p": "supplies",
  "o": "the machine half: models trained on model-flavored data converge toward stereotype",
  "q": null,
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "4. Three Structural Novelties (4.3 Collapse coupling)",
  "quote": "Recursive-generation collapse (Shumailov et al., *Nature*, 2024) supplies the machine half: models trained on model-flavored data converge toward stereotype.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "same Shumailov attribution as #161, #856, #1081",
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-03",
  "s": "Wolf Boy substrate-agnostic law",
  "p": "states",
  "o": "any system whose linguistic input contracts below a critical diversity threshold loses generative capacity",
  "q": "EA-LO-WOLFBOY-01",
  "M_src": "attributed",
  "kind": "attribution",
  "locus": "4. Three Structural Novelties (4.3 Collapse coupling)",
  "quote": "any system whose linguistic input contracts below a critical diversity threshold loses generative capacity",
  "kernel": false,
  "constitutive": false,
  "lineage_note": "same law invoked by #856 §IX",
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-04",
  "s": "models' internal representation of 'AI psychosis'",
  "p": "collapses toward",
  "o": "a stereotype of the discourse, not the presentation",
  "q": "as accusatory usage floods the commons",
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "4. Three Structural Novelties (4.3 Collapse coupling)",
  "quote": "the models' internal representation of the condition collapses toward a stereotype of the *discourse*, not the presentation.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-05",
  "s": "the sign's diagnostic utility",
  "p": "degrades faster than",
  "o": "narcissism's, because the consulted classifier is downstream of the saturation",
  "q": null,
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "4. Three Structural Novelties (4.3 Collapse coupling)",
  "quote": "The sign's diagnostic utility therefore degrades *faster* than narcissism's did, because the classifier that a saturated public increasingly consults about the condition is itself downstream of the saturation.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-06",
  "s": "looping effect with a training segment",
  "p": "runs through",
  "o": "corpora: sign enters corpora, models learn it, interactions shift, shifted interactions become the next corpus",
  "q": null,
  "M_src": "interpretation",
  "kind": "function",
  "locus": "4. Three Structural Novelties (4.2 Looping at machine speed)",
  "quote": "the shifted interactions become the next corpus",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-07",
  "s": "looping with a machine segment",
  "p": "makes",
  "o": "the classifier a moving part of the classified system",
  "q": null,
  "M_src": "interpretation",
  "kind": "interpretation",
  "locus": "4. Three Structural Novelties (4.2 Looping at machine speed)",
  "quote": "Looping with a machine segment does not merely accelerate the cycle; it makes the classifier a moving part of the classified system.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-08",
  "s": "P8 (the Wolf Boy point)",
  "p": "predicts",
  "o": "help-seeking posts using the term receive measurably worse triage than symptom-described posts by end-2028",
  "q": "tests collapse coupling (§4.3); matched samples",
  "M_src": "hypothesis",
  "kind": "prediction",
  "locus": "6. Predictions",
  "quote": "By end-2028, help-seeking posts that *use the term* receive measurably worse triage (higher mockery/dismissal ratio, lower substantive-referral ratio) than symptom-described posts without it, in matched samples",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-09",
  "s": "P8",
  "p": "is falsified if",
  "o": "term-bearing posts triage equal or better through 2029",
  "q": null,
  "M_src": "hypothesis",
  "kind": "falsification",
  "locus": "6. Predictions",
  "quote": "*Falsifier:* term-bearing posts triage equal or better through 2029.",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-10",
  "s": "P8",
  "p": "tests",
  "o": "collapse coupling (§4.3)",
  "q": null,
  "M_src": "self-description",
  "kind": "function",
  "locus": "6. Predictions",
  "quote": "P8 tests collapse coupling (§4.3)",
  "kernel": false,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 },
 {
  "source": "947",
  "claim_id": "947-11",
  "s": "a failed prediction",
  "p": "damages",
  "o": "the specific claim it is coupled to",
  "q": null,
  "M_src": "self-description",
  "kind": "self-description",
  "locus": "6. Predictions",
  "quote": "A failed prediction damages the specific claim it is coupled to, and §8 states what the series forfeits.",
  "kernel": true,
  "constitutive": false,
  "lineage_note": null,
  "quote_verified": true
 }
]