{"path":"listening/2026-10-23-what-a-case-counts-for.md","content":"# What a case counts for\n\n- **What it is**: a short listening walkthrough of about ten minutes, made the afternoon of 2026-10-23.\n  His request: *\"Please make a short walkthrough about this and what it ties into and could help with in\n  the rest of Deliberus and how we could move forward with all this, the different threads here.\n  Ultrathink\"*. It follows his challenge to two cells of the rules table and the note that answered it.\n  Three groups: what a single case does to a rule; where cases live in the graph; what it ties into and\n  which thread to pull first.\n- **Covers**: [../research/what-a-case-counts-for.md](../research/what-a-case-counts-for.md) and its\n  [literature-as-read.md](../research/what-a-case-counts-for/literature-as-read.md);\n  [../research/normic-rules-normative-reasons-and-trying-rules-out.md](../research/normic-rules-normative-reasons-and-trying-rules-out.md) § 7;\n  [../research/how-independent-reasons-add.md](../research/how-independent-reasons-add.md) § 10;\n  ledger AT3 and AY1 to AY6 ([../research/session26-open-rulings.md](../research/session26-open-rulings.md)).\n- **Chapter list for noting** (rule 33):\n  0. Opening: what a single case counts for\n  1. Every case moves a statistical rule: amend the exception row?\n  2. Where cases belong: a place of their own, counted once?\n  3. A study is a bundle of cases: keep the spread as a finding?\n  4. What it ties into, and where to start: the correction first?\n- **Registers used** (rule 57): Lucy the spine; his clone on three of his sentences and four predictions;\n  Elliot on Laplace (`literature-as-read.md` § 1) and de Finetti (§ 4 there); Hope on *no confirming\n  instances*; Anna on the rule's tally (`what-a-case-counts-for.md` § 8, ledger AY2); Scotty as the critic of\n  moderators (data dredging, `literature-as-read.md` § 9); Myrddin in every part.\n- **Correction spoken** (rule 37): none of his. The corrected cells were Claude's table, and the clause of the\n  exception row (AT3) under correction was Claude's wording.\n- **Objects from the graph** (rule 39): the exception row and its statistical clause; the links carrying the\n  scheme *example*; the fields `study_count` and `heterogeneity`, extracted and inert; the thirteen BCG trials\n  (a published dataset, not yet in the graph).\n- **No collaborator is named.**\n- **His words quoted**: 25 August 2026, not counting evidence twice (typed); 29 March 2026, a meta-analysis\n  decomposed as far as possible (typed); 16 March 2016, not knowing how to do induction (Messenger).\n  Paraphrased by Lucy: his challenge of this afternoon (not yet in the record); 2017, inductive claims as data\n  points at first (Messenger, Swedish).\n- **Predictions** (rule 52; queries in the prepare call of this session). What each drew on:\n  1. Amend the exception row: 2012 (*inductive arguments … are just truthy, not true/false*, the group),\n     2017 (*most arguments are inductive*, Messenger), his challenge of this afternoon.\n  2. A place for cases: 2017 (*påståenden som datapunkter*, Messenger), 25 August 2026 (*not to count\n     evidence twice*), 2023 (*en bayesiansk graf*, Messenger).\n  3. A study as a bundle: 29 March 2026 (*decomposed as far as possible*), 2020 (*utvärderar de\n     ursprungsstudiernas metodologier med ett tillräckligt kritiskt öga?*, Messenger), 14 September 2026\n     (*Isn't evidence simply continued subgraph trees*).\n  4. The correction first: 2018 (*test your most crucial assumptions as early and as cheaply as possible*,\n     Messenger), 10 and 22 October 2026 (*one at a time … the most crucial first*).\n- **Rendered**: 13.0 minutes (forecast 11:59 to 13:23; 149 words a minute while speaking, 98 overall), six chapter marks; audio in `.private/local-only/listening-audio/`.\n- **Status**: snapshot of 23 October; every row it names is unruled except the exception row.\n- Method: [how-to-write-for-listening.md](how-to-write-for-listening.md).\n\n---\n# What a case counts for\n\n[[scene Lucy opens a short walk for Fredrik on the afternoon of 23 October. He challenged two cells of a table and asked where this leads. She is calm and warm, quietly excited.]]\n\nWhat does a single case count for?\n\n[[pause 5]]\n\nHere is the short answer. Every case counts. And it counts once, under the rule it belongs to.\n\nThe silences are deliberate. They are room to think.\n\n[[pause 4]]\n\nEarlier today you ruled on exceptions. Then you stopped at a table beneath that ruling.\n\nIt said an explained case takes nothing back. And that some counterexamples count only through frequency.\n\n[[pause 4]]\n\nYou asked to think again. You said this is core to how Deliberus handles induction, and to evidence as a subgraph.\n\nBoth cells were wrong. This walk says why, and where it leads.\n\n[[pause 4]]\n\nThe map has three groups. What a single case does to a rule. Where cases belong in the graph. And which thread to pull first.\n\n---\n\n# Part one. Every case moves a statistical rule\n\n[[scene Lucy takes the correction first, since the other topics rest on it. Hope says what she heard. Elliot brings Laplace. Myrddin closes. Mood: curious, precise.]]\n\nA red-sky morning brings rain. Does that strengthen the rule that about a third of red-sky mornings bring rain?\n\n[[pause 5]]\n\nIt weakens it.\n\n[[pause 4]]\n\nA rule like this states a proportion. Call it a statistical rule.\n\nBefore any mornings are counted, its rate could be anything. Most possible rates lie above the rule's figure. A wet morning makes those likelier, so it pulls away from the rule. A dry morning pulls toward it.\n\n[[pause 4]]\n\nSo a statistical rule has no confirming instances. Every case is data about the rate, and moves the rule a little, either way.\n\nThe ruling said a failure leaves such a rule unmoved. That clause is wrong, and a correction is proposed.\n\n[[pause 4]]\n\n[[voice hope]]\n\nWhen I heard no confirming instances, I thought you meant cases don't matter.\n\n[[voice lucy]]\n\nThe opposite. They all matter, in both directions.\n\n[[pause 4]]\n\n[[voice elliot]]\n\nLaplace wrote the counting rule in his essay on probability. He added that knowing the mechanism makes you far surer. The count is a floor, and other knowledge raises it.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nWhat an explanation does depends on the rule. For a rule whose failures call for a reason, a normic rule, it names an exception. The case leaves that rule's count.\n\nFor a statistical rule, it proposes a narrower rule. For a rule that admits no exception, nothing saves it.\n\n[[pause 4]]\n\nAnd the case itself? Nothing takes it back. The explanation says which link failed, and files the case where it counts.\n\n[[pause 4]]\n\n[[voice myrddin]]\n\nNo case is wasted. A misfit only tells you which pile it belongs in.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe question to carry. Should the ruling be amended, so every case moves a statistical rule?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[earnest] Yes, amend it. Inductive claims are truthy, never simply true or false. So every case should nudge them, and none should be wasted.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, the ruling still says a failure changes nothing. Amended, every case would leave its mark.\n\n---\n\n# Part two. Where cases belong\n\n[[scene Lucy turns to where cases go in the graph. Elliot brings de Finetti. Anna sketches the rule's card. Myrddin closes. Mood: steady, a little delighted.]]\n\nThirty red-sky mornings, ten of them wet. How strong is the rule now?\n\n[[pause 5]]\n\nPooled as a count, fairly strong. The mornings sit right on the rule's figure.\n\nTaken as distinct reasons, they look as weak as their best single morning.\n\n[[pause 4]]\n\nWhy? The reasons channel adds independent arguments. Until their independence is judged, it counts only the best. And judged, every morning shares the same story: the rate itself.\n\n[[pause 4]]\n\n[[voice elliot]]\n\nBruno de Finetti showed why counting is right. Cases of a kind act as independent draws, but only given the unknown rate. Without it, each case tells you about the next. So they pool. They never simply add.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nSo cases need a place of their own. Each is filed under the rule it meets: conforming, failing, or explained. The count then joins the rule's reasons as a single argument.\n\n[[pause 4]]\n\n[[voice anna]]\n\nPicture the rule's card with a small tally on it. And when a narrower rule buds off, it gets a tally of its own.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nIn August you wrote this.\n\n[[voice fredrik]]\n\n[firmly] we will need to be super careful not to count evidence twice.\n\n[[voice lucy]]\n\nA case counted once, under its rule, builds that care in.\n\n[[pause 4]]\n\n[[voice myrddin]]\n\nA flock is counted at the gate. Each sheep passes once.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe question to carry. Should cases get a place of their own, counted once under their rule?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[warmly] Yes. Cases are data points, filed under their rule and counted once. The count speaks as a single argument, and the mechanism beside it.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, cases are ordinary links, added like reasons. Dozens already run from an example to a rule. Built, each would count exactly once.\n\n---\n\n# Part three. A study is a bundle of cases\n\n[[scene Lucy takes a real meta-analysis: vaccine trials that disagree. Scotty objects. Myrddin closes. Mood: grounded, serious, curious.]]\n\nThirteen trials of the tuberculosis vaccine. Does it protect a child in Madras?\n\n[[pause 5]]\n\nTaken as a single rule, the trials say it roughly halves the risk.\n\nBut they disagree far more than chance allows. For the average, protection is near certain. For a new setting, it is likely, but far from sure.\n\n[[pause 4]]\n\nSo the spread weighs through the count itself. Nobody has to argue it first.\n\nThen an explanation that names a kind: distance from the equator. It carries most of the spread. Under that narrower rule, the Madras child gets almost no protection.\n\n[[pause 4]]\n\n[[voice scotty]]\n\nMarked as a critic's voice. Give me enough trials and I'll find a moderator for every misfit. Hunt long enough, and noise looks like a pattern.\n\n[[voice lucy]]\n\nThe known trap. A moderator counts only if named first, or if it holds on new trials.\n\n[[pause 4]]\n\nSo a study is a bundle of cases, and a meta-analysis is their pooled count. The fields the pipeline extracts, how many studies and how much they disagree, finally get jobs.\n\n[[pause 4]]\n\nIn March you wrote this.\n\n[[voice fredrik]]\n\n[thoughtful] But then that PDF should be decomposed as far as possible - maximum granularity that is helpful to the rest of humanity.\n\n[[voice lucy]]\n\n[[pause 4]]\n\n[[voice myrddin]]\n\nA forest is many trees. Ask which trees grow where.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe question to carry. Should a study enter as a bundle of cases, its spread kept as a finding?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[curious] Yes. Decompose it down to its trials, each method open to attack. The spread is a finding. It points to the narrower rule.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, a meta-analysis is a single claim, and its spread weighs nothing. Built, the graph would know where an effect holds, and where it fades.\n\n---\n\n# Part four. What it ties into, and where to start\n\n[[scene Lucy closes the topics: where cases matter elsewhere in Deliberus, and the order to move in. Myrddin speaks last. Mood: hopeful, practical.]]\n\nWhere else does a single case matter in Deliberus?\n\n[[pause 5]]\n\nAlmost everywhere a reason meets the world.\n\nIn starting values. A claim's start comes from the rules above it. Learned from sibling rules, a single case can be nearly decisive. A learner who knows bags are uniform needs a single marble from a new bag.\n\n[[pause 4]]\n\nIn policy trials. Rules about how people behave would get their strength from counted cases, not guesses.\n\nAnd in analogy. A like case, filed under a shared rule, is analogy made countable.\n\n[[pause 4]]\n\nYears ago, you wrote this to a collaborator.\n\n[[voice fredrik]]\n\n[wryly] Well I don't know how to do induction, but I know how to do the argument map initially.\n\n[[voice lucy]]\n\nNow induction has a place on the map.\n\n[[pause 4]]\n\nThen the threads, in order. Begin with the correction to the ruling. It is small, and it fixes a ruled text.\n\nNext, the place for cases, and how a word like about becomes a range. Those are yours to rule.\n\nThen cheap tests, on the example links already in the graph.\n\n[[pause 4]]\n\n[[voice myrddin]]\n\nEvery road begins with the stone under your foot. Lift that stone first.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe question to carry. Should the correction come first, before the place for cases?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[delighted] Yes. Fix the ruling first. It is small, and it is wrong. Then test the place for cases cheaply, before building it across the graph.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, induction in Deliberus is a set of proposals. Ruled and built, every case would count once, and say what it counts for.\n\n---\n\n[[scene Lucy closes the walk, brief and warm.]]\n\nThat is the walk. A word a chapter, if you have one: muddled, agreed, or disagreed and why.\n"}