{"path":"research/what-a-rule-rests-on/corpus-as-read-2026-10-08.md","content":"> **Kept as it came back.** This is the report of a read-only helper (a language model) made on 2026-10-08 for [what-a-rule-rests-on.md](../what-a-rule-rests-on.md). It surveyed this repository: documents, code and the founder's recorded words. Its findings are its own reading, with the depth of reading stated per item. One section, on the founder's private archive, is kept outside the public documents, and a local file path was removed. Nothing else was changed.\n\n# Corpus report: what Deliberus already holds on analogy, induction, cases, rules and likeness\n\nRead-only survey of the repository, 2026-10-08. Conventions: **[FULL]** read start to end; **[PART]** grep windows or line ranges only. \"Q:\" is an exact quote (30 words or fewer; markdown emphasis dropped). \"INFERENCE\" marks my own reading. No opinion on the question is given. All repo paths below are under docs/research/ unless stated; \"the 10-07 notes\" are philosophy-of-the-unwritten-step.md, writing-out-the-unwritten.md, what-a-link-rests-on.md, how-far-a-premise-is-from-its-claim.md.\n\n## A. Document by document\n\n**A1. philosophy-of-the-unwritten-step.md [FULL]** (2026-10-07; AD22 to AD29, all unruled) with its folder file philosophy-of-the-unwritten-step/what-each-lookup-returned.md [FULL] (fetch-model summaries; \"Claude did not read the pages\").\n- § 5 \"A step has a shape\". Q: \"From the rule and the case to the result is deduction. From the case and the result to the rule is induction.\" Hypothesis (later abduction) is rule + result to case; his 2011 squirrel is that third arrangement.\n- § 1, ancient logic: Aristotle, in the note's words, Q: \"counts the sign and the example as proofs beside the kind that runs from a rule\". Lookup #33 (a search summary) gives Peirce's bean bag: Q: \"Case.—These beans are from this bag. Result.—These beans are white. Rule.—All the beans from this bag are white.\"\n- § 5 table, third shape: Q: \"the earlier case, and the respect in which the two are alike\"; what stays to be asked: Q: \"are they alike in what matters\". Same section: Q: \"The reader built this evening is a deductive reader.\"\n- INFERENCE: induction (case + result to rule) is described in § 5 but is not one of the three shapes in the table or in AD23.\n- § 1, fifteen angles. Practice row: Q: \"Some links rest on a like case and not on a rule.\" Peirce and Mill row: Q: \"A rule's count across links is its register.\" Indian logic row: Q: \"A reason is good when it is found in like cases and never in unlike ones.\"\n- § 6 (Bolzano, Goodman, law, reference class, pragmatics). Q: \"A rule's count across links is its record, and a record is conservative.\" Entrenchment is \"the record of its terms in past projections\".\n- § 10. Q: \"an analogy is proposed together with its strongest counter-analogy\" (pointing to the-analogy-daemon.md). § 14: Q: \"No procedure.\" and Q: \"No verdict between the camps.\"\n- Lookups (second hand): Aristotle's example as \"rhetorical induction\" reasoning Q: \"from one particular to a similar particular if both fall under the same genus\" (#3); Mill, Q: \"General propositions are merely registers of such inferences already made, and short formulae for making more\" (#51); Kuhn, Q: \"similarity to the exemplar\" (#45); precedent, Q: \"analogies provide non-binding, variable-strength arguments grounded in underlying justifications, while rules presumptively settle cases.\" (#27). Unused search-result titles also sit in the raw lists: \"Reasoning by Analogy: A General Theory\", \"Hesse’s Condition for Transitivity of Probabilistic Support\", \"The Abuse of Casuistry\".\n\n**A2. The other 10-07 notes and the two 10-07 listening pieces [FULL].** No treatment of induction or analogy as such. Q: \"A written-out link is the first claim, what ties this case to a rule, and the rule.\" Rule tests: Q: \"a parallel case can test it\" (writing-out § 6). how-far § 3b: Q: \"(example, cause to effect, analogy, and so on)\" as the patterns links carry. Only support links were sampled.\n- what-a-link-rests-on/outside-reading-argumentation-theory.md [PART], Blair: Q: \"For defeasible arguments (analogy, induction, best explanation, conductive) a validity-ensuring premise such as no further evidence or all else equal misrepresents the force\".\n- writing-out-the-unwritten/outside-reading-explication-with-checks.md [PART], on Hitchcock: tested by Q: \"refutation by logical analogy\", i.e. Q: \"find a parallel argument, built by substituting content expressions, with true premises and a false conclusion.\"\n\n**A3. the-squirrel-and-the-strength-calculus.md [FULL]** (2026-10-08) and its helper report the-squirrel-and-the-strength-calculus/strength-rulings-and-code-as-read-2026-10-08.md [FULL]. § 4 maps shape to arithmetic: rule + case = \"jointly necessary parts\", weakest governs; sign = \"they add, and a rival that explains them counts against all at once\"; like case = Q: \"not in the review\" and \"open\". Q: \"So the all-necessary rule fits the first shape only, and a sign is weighed for what it is worth.\"\n\n**A4. Analogy machinery (August to September).**\n- the-analogy-daemon.md [FULL]: status Q: \"designed on paper here, nothing built\". Q: \"no analogy claim without its strongest counter-candidate attached.\" Q: \"an analogy is the highest-leverage vehicle for plausible-wrong\". § 7 (2026-09-29): Q: \"kinship means a shared problem and a shared family of solutions, a shared shape alone being a metaphor.\"\n- perceiving-possibility-and-the-analogy-organ.md [FULL] (2026-09-02): the only place analogy-as-cognition appears. § 4: Hofstadter and Sander, Q: \"every concept is built by analogy\", \"claim-sameness stated from cognitive science\"; Copycat codelets, Slipnet, temperature. Own caveat: references are Q: \"from memory of well-known work, not re-read this session\".\n- convictions-across-scales.md [FULL] (2026-08-18; update 2026-08-30): the four checks (Recurrence: Q: \"Does the pattern appear at two or more independent levels, named by someone not arguing for this project?\"; Mechanism identity; Adversary; Substrate condition). Update: Q: \"the generic CQs were induced from everyday argument\". A fractal prior is Q: \"an analogy-expressing MOTHER claim (this pattern recurs at N scales with mechanism M) decomposing into per-scale instance subclaims\". Cross-scale recurrence is a prior; Q: \"priors of this kind have a specific failure history\".\n- the-irit-harvest.md § 3 [FULL]: Amgoud's four considerations Q: \"Relevance, Number, Diversity, Disanalogy\"; Q: \"every distinct dissimilarity weakens the analogy\"; Q: \"semantics choice is argument-type-relative\". Proposed per-scheme layer, unbuilt.\n- fractal-priors-for-success.md [FULL]: Q: \"these are priors about the CLASS, never predictions about this instance\". the-reuse-flywheel-prior.md [FULL]: seven-scale table, legal precedent Q: \"case law accumulates; new cases cite holdings rather than re-deriving them\", identity \"judged\"; Bar-Haim: seven key points cover 72.5% of 6,515 arguments. assembly-theory-and-the-reuse-mechanism.md [FULL]: Q: \"a borrowed category earns its place by yielding a move\".\n- retrieval-instruments-beyond-cosine.md [FULL] and dogfood-run-6-israel-palestine-cross-domain.md [FULL]: Q: \"Different words, matching shape.\" Q: \"Two claims can share an argument structure while sharing almost no vocabulary, and claim-level cosine measures vocabulary.\" The only \"case-based\" is a source title.\n- docs/convergence.md [PART]: Q: \"Decomposition is only one resolution operator: the typed leaf is where reframing, analogy transfer to already-settled horizons, scope-narrowing, and the handoff from analysis to attunement get their exact coordinates.\"\n\n**A5. Scheme taxonomy and the inductive label.**\n- scheme-set-exhaustiveness.md [FULL]: Q: \"The set is a species catalogue built by induction from examples\". Kienpointner sorts schemes into rule-using, rule-establishing (\"inductive example argumentation\") and neither (\"illustrative example, analogy, authority\"). Macagno and Walton: one causal relation gives five moves including \"inductive generalization\". Unbuilt proposal: reasoning direction (deductive, abductive, inductive) as a closed-layer coordinate.\n- scheme-bounded-decomposition-and-evidence-as-subgraph.md [FULL], § Deductive vs Inductive: Q: \"You don't need separate systems for deductive vs inductive.\" Analogy decomposes as source case, similarity mapping, relevant differences, counter-analogies; example as instance, generalization, representativeness, counter-examples.\n- walton-argument-schemes.md [FULL]: analogy is Q: \"Case C1 is similar to case C2; A is true in C1; therefore A is true in C2.\" inductive-reasoning-research.md [FULL] and technical-direction.md [PART]: Q: \"Most real-world arguments are inductive — based on evidence, statistics, probability, analogy, generalization.\" conceptual-threads.md [FULL] Thread 5: Q: \"(deductive, inductive, abductive, from analogy, from authority)\". science-lineage-philosophy.md [PART]: Peirce only as community inquiry and convergence.\n- academic-foundations.md [FULL] and depth.md [FULL]: nothing on analogy, induction, cases or Hofstadter.\n\n**A6. Evidence dossier and strength literature.**\n- what-counts-as-evidence.md [FULL], what-counts-as-evidence/philosophy-of-evidence.md [FULL], corpus-map-founder-voice.md [FULL], README.md and research-briefing.md [FULL]; the other six strand files [PART, windows around hits]. Q: \"it worked there\" is not \"it will work here\"; proposed (Q2/Q11, unruled) a generalises inference carrying external-validity questions; Q: \"The current schemes (cause-to-effect, analogy, example) each partly cover this\"; Q: \"Diverse evidence beats repeated evidence\" is false in general; Whewell's consilience of inductions; reflective equilibrium, Q: \"principles (mother claims) and case judgments (children) adjusted until they cohere\"; Bradford Hill's analogy, Q: \"parallel (consistency, replication, analogy)\"; Wigmore's inductive generalizations, Q: \"that provide a warrant or a license for reasoning from one node to another\".\n- weakest-link-arithmetic-and-the-merge-hunch.md [FULL], § 9b: Q: \"replication can trump variety\"; Q: \"the dial is proposed, unruled\".\n- strength-calculus-literature/bayesian-and-informal-logic.md [PART: lines 1-330 and ranges]: link kinds listed are Q: \"cause-like, sign-like, an independent sufficient reason, a jointly necessary part\"; no like-case kind. computational-argumentation.md [PART]: no analogy. starting-values-and-the-problem-of-the-prior.md [PART]: Q: \"Probability is two-place: the reference class problem is everyone's\"; the choice of class Q: \"is still a judgement\".\n- core-periphery-and-the-concept-layer.md [FULL]: a sense is Q: \"A sense is a triple\" (exemplar core, criteria list, per-position weighting). INFERENCE: this rule-versus-exemplar structure is not linked anywhere to link-level likeness.\n- what-the-walkthroughs-and-the-reviews-are-for.md [FULL] (2026-10-08): AD23 pictured as Q: \"a map that reads a step from a rule, from a sign, and from a like case\".\n\n## B. The founder's own statements in the repo (verbatim; date; file)\n\nNo recorded statement of his says that induction rests on analogy, or that a step's rule comes from induction over cases (search result; see E).\n\n1. 2011-08-29, writing-out-the-unwritten.md § 1 (a saved discussion thread): Q: \"Just to clarify I mean simply to argue that \"The squirrel is dead.\" from \"The squirrel smells of decomposition.\" along with \"Dead squirrels smell of decomposition.\"\"\n2. about 2012, simplenote-archive-analysis.md § 1 (Leverage notes): Q: \"Isomorphisms between paths to conclusion? Or just isomorphism between conclusions?\"\n3. Jan 2012: Q: \"Logic is universal.\" (chat-logs-analysis.md § 3). technical-direction.md L72 gives his 2012 sentence as Q: \"Deductive arguments are simpler than inductive arguments — dealing with probability estimates for all claims can get hairy.\" corpus-map-founder-voice.md flags the surrounding \"icky issues\" list as paraphrase.\n4. 2026-03-29, scheme-bounded-decomposition-and-evidence-as-subgraph.md: Q: \"But of course there's also inductive arguments, right? And the whole thing about what constitutes a good bayesian inductive argument and so on.\"\n5. 2026-04-05, session26-open-rulings.md P2 (OPEN): Q: \"if there's inductive evidence for any claim, that's what determines the probability assessment\"\n6. 2026-08-18, convictions-across-scales.md: Q: \"there might be a strong expectation that I hold for it based on analogies to similar patterns in the world\". Same day, assembly-theory doc, quoted as \"flagged by its holder\": Q: \"theory, analogies, and fractal evolutionary priors can enlighten the corpus\".\n7. 2026-08-29, the-reuse-flywheel-prior.md § 6: Q: \"people don't have that many different claims and concepts that they throw around\"\n8. 2026-08-30, convictions-across-scales.md update: Q: \"represented within Deliberus as well-evidenced analogies/claims that express analogies, simply. Is there an analogy scheme??\"\n9. 2026-08-30, the-analogy-daemon.md § 1: Q: \"whose job it is to discover and evaluate analogy claims\"; Q: \"seeing the broad strokes and intuiting what belongs or doesnt\".\n10. 2026-08-30, the-harness-conviction.md: Q: \"sloppy analogies and forgotten premises and just general AI slop\"; Q: \"the most accurate priors, evidence, inferences and analogies that humanity can muster\". § 8.2: Q: \"I have no clue what the graveyard spiral is so maybe not the best pedagogical analogy\".\n11. 2026-09-02, the-analogy-daemon.md § 7 (human-seeded channel): Q: \"Very useful, document this idea.\"\n12. 2026-09-05, session29-the-derivability-arc-and-the-conservative-design.md Part III: Q: \"Help me see this clearly: what's the load-bearing idea/s? One bridging analogy. Competing perspectives if relevant.\"\n13. 2026-08-17, retrieval-instruments-beyond-cosine.md: Q: \"What IS the right instrument for that population then, or for search in Deliberus in general, if not cosine over embeddings?\"\n\n## C. Ruled versus unruled\n\n**Ruled (his words), and shared by every scheme because the arithmetic does not branch on scheme:** B9/B10 (2026-08-31, \"Go with option C, both.\"; row B10 still prints \"Awaiting founder confirm\" although the provisional blessing is on record, helper F2); Q33 (2026-09-28, \"About the decision: your proposal is blessed, make it so.\"; a link carries supporter strength times link relevance; how the factors combine stays open); Q34 (2026-09-30, for now: a refuted supporter adds nothing); Q31 (a link's strength is a reading); Q32 (retire `confirmed`; still read in code per helper F9).\n\n**Open, Claude's proposals, none ruled:** AD23 (\"three ways to walk\": rule + case; sign; like case; \"rec: adopt; until the other two shapes have checks, mark such links by shape and do not force a rule\"); AD22, AD24, AD26, AD27, AD28 (Sunstein, Toulmin), AD29; AD25 (rules in the plural; Goodman's entrenchment; \"a record is conservative\"); AD13 (its row still says the squirrel already has the three-part form; AD23 says it does not, helper F7). Other open rows: P2 and P3 (his positions on inductive evidence and \"fuzzy gooey subjectivity\"), Q5 (Keynesian weight), Q17 (inductive risk), Q22 (Schum's five combination patterns), S6 (kinship needs a shared problem), T2, O13, F3 (an imported taxonomy brings \"induction corpus plus its unresolved disputes\"). S5 is SUPERSEDED. Q28 (OPEN) keeps its layer 2 behind the 2026-08-30 hold on self-hosted fractal priors.\n\n**No ledger row exists** on: the analogy daemon, the four-check test as the analogy scheme's CQ enrichment, Hofstadter or Copycat, induction as the source of a step's rule, or per-scheme aggregation (grep of all 766 lines).\n\n**CLAUDE.md:** L758 (convictions-across-scales entry: Q: \"Do NOT re-derive these as open.\"; self-hosting held, under Q: \"2026-08-30 update + HOLD\"); L799 (Hofstadter \"never cited before\"; human-seeded channel founder-endorsed); L822 (the four-check test Q: \"ruled the same day as the analogy scheme's cross-scale CQ enrichment\"); L651 (research docs are \"NOT a specification\"); L904 (his phrase \"A wiki where every word has been vetted deductively and inductively\", 2026).\n**TODO.md:** L938 (open: a check for a step from a sign and from a like case; until then \"marked by shape and left unwritten\"); L1747-1762 (Hofstadter into the sameness lineage and the Academic Foundations list \"needs your blessing\"); L1966-1969 (per-scheme aggregation, design note, unruled); L1998 (analogy daemon \"designed, awaiting founder pacing\").\n\n## D. The code facts (I read these in full: deliberus/extraction/schemes.py, cq_fallbacks.py, graph/qbaf.py, relevance.py, support_semantics.py; graph/badge_inputs.py L39-255)\n\n**Where the list lives.** deliberus/extraction/schemes.py: `ArgumentScheme` (46 members; the docstring says \"Walton's 96\"), `SchemeCluster` (11 values), `WALTON_SCHEMES` with the CQ templates, `CQ_BEARS_ON` (L593-662), `SCHEME_BADGE_TERMS` (L673-685). deliberus/cq_fallbacks.py `FALLBACKS` (L31-177) gives a slot-free sentence per template, e.g. relevant_differences: \"There is no relevant difference between the two cases that undermines the analogy\"; precedent_similar: \"The past case is sufficiently similar to this one in the respect that matters\". There is **no scheme named induction or generalisation** (the only \"induct\" in deliberus/*.py is the docstring at support_semantics.py L28-29); `example` carries the case-to-pattern move.\n\n**Analogy cluster, CQ templates verbatim (L321-343).**\n- analogy: relevant_differences \"There are no relevant differences between {source_case} and {target_case} that undermine the analogy\"; source_true \"{claimed_property} is actually true in {source_case}\"; counter_analogy \"There is no stronger counter-analogy — a case similar to {target_case} where {claimed_property} is false\".\n- example: example_representative \"{example_case} is representative of the general pattern claimed about {general_domain}\"; example_true \"The claimed outcome actually occurred in {example_case}\"; counter_example \"There are no counter-examples where {claimed_outcome} did NOT occur in similar cases\".\n- precedent: precedent_similar \"{past_case} is sufficiently similar to the current case regarding {relevant_dimension}\"; precedent_changed \"Circumstances have not changed since {past_case} in ways that undermine the precedent\".\n\n**Sign and cause patterns (causal cluster).**\n- sign (L288): sign_reliable \"The correlation between {indicator} and {indicated} is strong enough to be reliable\"; sign_alternatives \"There are no other factors that could explain {indicator} besides {indicated}\".\n- cause_to_effect: causal_strength \"The causal generalization that {cause} leads to {effect} is well-supported by evidence\"; cause_present \"There is adequate evidence that {cause} is present in this case\"; causal_interference \"There are no other factors that could prevent {effect} despite {cause} being present\".\n- effect_to_cause: alternative_causes \"There are no alternative causes that could explain {effect} besides {proposed_cause}\"; base_rate \"The base rate of {proposed_cause} occurring is high enough to make it a plausible explanation for {effect}\"; mechanism_plausible \"There is a plausible mechanism by which {proposed_cause} would produce {effect}\".\n- correlation_to_cause: confounders \"The correlation between {variable_a} and {variable_b} is not explained by confounding variables\"; temporal_order \"{variable_a} precedes {variable_b} temporally, supporting a causal direction\"; dose_response \"There is a dose-response relationship between {variable_a} and {variable_b}\".\n\n**Nearest to the other shapes.**\n- abductive cluster: evidence_to_hypothesis: explanatory_power \"{hypothesis} satisfactorily explains {evidence}\"; competing_hypotheses \"No competing hypothesis explains {evidence} better than {hypothesis}\"; search_thorough \"The search for alternative hypotheses to explain {evidence} has been thorough\". abductive: best_explanation \"{hypothesis} is the best available explanation for {observation}\"; alternatives_eliminated \"Alternative explanations for {observation} have been adequately considered and found wanting\".\n- rule cluster, established_rule: rule_applicable \"The rule about {rule_content} applies to this specific case\"; rule_exceptions \"This case does not fall under any recognized exception to the rule about {rule_content}\"; rule_authority \"The authority establishing the rule about {rule_content} is legitimate\". Also verbal_classification, exception_to_rule, composition/division (property_transfers \"The property {property} of the parts validly transfers to {whole}\"), causal_slippery_slope and precedent_slippery_slope.\n- Fallback for any link with no fit, `default_inference` and `does_not_fit` (L199-204): premise_acceptable \"{premise} is itself acceptable — true, well-supported, or reasonably presumed\"; warrant_acceptable \"IF {premise} holds, {conclusion} genuinely follows — the connecting assumption survives being stated out loud\"; exception_absent \"No exceptional circumstance applies here under which {premise} holds but {conclusion} fails\"; rebuttal_absent \"There is no stronger independent case FOR the opposite of {conclusion}\".\n\n**Where the \"cross-scale\" or \"four-check\" enrichment lives: in documents only.** convictions-across-scales.md (2026-08-30 update), the-analogy-daemon.md § 3, CLAUDE.md L758 and L822. What the documents say it does: the shipped CQs are \"the generic tier\"; Recurrence turns example_representative into a count of independent instances; Mechanism identity turns relevant_differences from absence-of-difference into presence-of-same-mechanism; Adversary and Substrate condition are the two classes of difference that break cross-scale transfers. In code: no match for four-check, cross-scale, transfer test, mechanism identity, substrate condition, recurrence or fractal in deliberus/, tests/, config/ or web/src; tests/test_schemes.py L88-95 pins the analogy scheme to exactly three CQs. Git: commit 2d0d6c7 (2026-08-30) calls it an \"unruled mapping\".\n\n**How a link's strength treats these schemes: one arithmetic for all.** qbaf.py `compute_claim_badge` (L292-482): each argument link adds relevance x (acceptability - 0.5) (relevance.py L108-119; zero if the supporter has no standing or is below 0.5), signed by SUPPORTS or ATTACKS, summed; plus the energy of DECOMPOSES_INTO parts; then `qem_strength(0.5, energy)` with impact x^2/(1+x^2); then capped by the weakest required part. Several links into one claim are separate terms; I found no grouping or independence discount (the fair-share discount in evidence_routing.py I did not read; the helper report says only slicing_stability.py uses it). Scheme enters in three ways only: (a) an edge counts only if it carries a non-empty scheme (badge_inputs.py L50-51); (b) the cluster picks the headline term, e.g. \"STRONG ANALOGY\" or \"WEAK ANALOGY\", \"WELL-EVIDENCED\" or \"UNSUPPORTED\", \"BEST EXPLANATION\" or \"CONTESTED\", \"VALID\" or \"INVALID\", and only when the direct-edge channel dominates (qbaf.py L422-440); (c) `CQ_BEARS_ON` sorts each answered question: relevant_differences, example_representative, precedent_similar, precedent_changed, sign_reliable, sign_alternatives and the abductive and rule questions are \"connection\" (they move the link's relevance); source_true, example_true and cause_present are \"premise\" (they move the supporter's acceptability); counter_analogy and counter_example are \"rebuttal\" (they act on the target). Relevance comes from an argued relevance claim, else a confident machine proposal, else a presumption: 0.5 for extraction, human or split provenance, 0.0 for any other named provenance such as auto_connect (relevance.py L64-89). support_semantics.py (necessary, corroborative, alternative, undecided) governs DECOMPOSES_INTO parts only; its docstring says Q: \"deductive versus inductive is about the INFERENCE and is already absorbed by the scheme on the edge\", but no code path reads the scheme to change a sum. Per-scheme aggregation (counting semantics for analogy) is proposed (TODO L1966-1969) and unbuilt.\n\n**Built like-case finder.** deliberus/extraction/candidate_generation.py `structural_kinship_candidates` (L128-224) scores two claims by weighted Jaccard over the multiset of (direction, edge type, scheme) on their incident edges (minimum mass 4). It proposes pairs only: Q: \"neither writes an edge, moves a strength, or merges anything\". Measured as the weakest of three generators (the-reuse-flywheel-prior.md § 2).\n\n**Live counts** (the-harness-conviction.md, 2026-08-30, not re-measured by me): Q: \"558 analogy-cluster edges (413 example / 112 analogy / 33 precedent) of 3,666 scheme-bearing\".\n\n## E. Searched for and NOT found (a zero matters)\n\n- **No hit at all** in docs, specs, deliberus, scripts, TODO.md, README.md, SESSION-HISTORY.md or CLAUDE.md: Gentner; Holyoak; Bartha; \"John Wisdom\"; overhypothes*; Tenenbaum.\n- **Hits that are not the thing sought:** \"structure.mapping\" (one match, the phrase \"question-to-structure mapping\"); Hesse (one search-result title about transitivity of probabilistic support, not analogy); Norton (one publisher URL; no John Norton); casuist* (one search-result title); \"case-based\" (one source title); \"paradigm case\" (non-technical uses, never Wisdom's); \"prototype\" (product prototypes only, no prototype theory).\n- **Present but never on this topic:** Hume (is-ought, testimony reductionism; no problem of induction, whose only trace is a Stanford citation in starting-values-and-the-problem-of-the-prior.md § 2.1); Mill (only the Peirce/Mill row and the lookup quotes); Goodman (grue, entrenchment, and \"Goodman 1955 for induction\" as reflective equilibrium; nothing on analogy); Keynes (weight versus balance only); Sunstein (the four failures of deliberating groups; incompletely theorized agreements, AD28); Peirce (rule-case-result only in the 10-07 philosophy note and ledger AD23; elsewhere community inquiry); Nyaya, Dignaga, \"Indian logic\" (only that note, its lookups, AD23); Kuhn's exemplars (only lookup #45).\n- **Hofstadter, Copycat, \"Surfaces and Essences\", \"Fluid Concepts\":** only perceiving-possibility-and-the-analogy-organ.md and its mirrors (the-analogy-daemon.md, session29..., TODO.md, README.md, SESSION-HISTORY.md, CLAUDE.md L799; Copycat once in major-transitions-heredity-and-evolvability.md). Not in sameness-merge-and-split-across-fields.md, docs/academic-foundations.md or CLAUDE.md's Academic Foundations section, where the 2026-09-02 note proposed adding him.\n- \"variety of evidence\" as an exact phrase matches 3 files; allowing the hyphenated form \"variety-of-evidence\" it matches 12 (weakest-link § 9b among them). No document states that induction rests on analogy, or the reverse.\n\n## F. Where documents disagree, and gaps no document fills\n\n**Disagreements (observed).**\n1. *Status of the four-check enrichment.* \"ruled\" (CLAUDE.md L822; the-analogy-daemon.md § 3, \"already ruled\") against \"load-bearing-but-unratified\" (convictions-across-scales.md), \"unruled mapping\" (commit 2d0d6c7), no ledger row, and no code (the analogy scheme still has exactly three CQs).\n2. *Hofstadter.* The 09-02 note's \"rg -i hofstadter docs/ returns nothing\" and CLAUDE.md L799's \"never cited before\" are now stale; the proposal to add him to the sameness lineage and the Academic Foundations list is still waiting (TODO L1761), and the note's own Hofstadter claims are \"from memory\".\n3. *The squirrel.* Ledger AD13 still says it has the three-part form; AD23 and the 10-07 philosophy note say it is a step from a sign (helper F7).\n4. *Stale build status.* dogfood-run-6 calls scheme-level matching unbuilt; structural_kinship_candidates shipped 2026-08-20.\n5. *How induction and analogy relate.* inductive-reasoning-research.md, technical-direction.md and conceptual-threads.md list analogy beside generalization as kinds of inductive argument. Kienpointner (scheme-set-exhaustiveness.md) files \"inductive example argumentation\" under rule-establishing and analogy under \"neither\". Aristotle's example is \"rhetorical induction\" by likeness (lookup #3). Peirce (§ 5) makes induction one arrangement of rule, case and result. No document reconciles these or says which is basic.\n6. *Counts.* The module docstring says \"Walton's 96\" and a comment says ten clusters; the code has 46 schemes and 11 clusters. precedent is filed under both rule-based and analogy in walton-argument-schemes.md; in code only under analogy.\n\n**Gaps (INFERENCE from what is and is not written).**\n- The three-shape list has deduction, a sign and a like case but no induction shape. The rule in shape one gets a provenance only as a \"register\" (Mill) or \"record\" (Goodman) and as one of five kinds of stopping point, \"a generalization from experience\" (§ 4); nothing says what the earlier inferences in that register consist of.\n- Strength for the like-case shape is \"open\" (A3). The code adds links by the same sum for every scheme, with no independence discount. The number of cases, their diversity and their independence are treated separately in the IRIT harvest (Number, Diversity), in weakest-link § 9b (variety of evidence) and in the four-check Recurrence test; no document joins them.\n- No check exists for the like-case shape (TODO L938); the trial's blind reader is deductive. Testing a rule by a parallel case or \"logical analogy\" appears in writing-out § 6 and a helper report but not among the trial's four checks.\n- The concept layer's exemplar core and criteria list, and the link layer's rule versus like case, are two unconnected treatments of rules versus examples.\n- Analogy as the core of cognition appears once, as the justification for claim-sameness (perceiving-possibility § 4); no document draws on it for how a link or a rule is formed.\n\n## G. The founder's private archive\n\nSearched as well. What it returned is kept apart from the public documents.\n\n## Read-depth disclosure\n\nNot read in full though on your list: six strand files in what-counts-as-evidence/ (corpus-map-code, corpus-map-research-docs, formal-and-computational-models, grading, law-intelligence, testimony; windows around hits only) and the two strength-calculus-literature files (lines 1-330 plus ranges; headers and tables). CLAUDE.md, TODO.md, session26-open-rulings.md, README.md and SESSION-HISTORY.md were searched by grep and the matching lines or rows read. I did not query the live graph, so every live count above is a document's.\n"}