{"path":"listening/2026-10-23-why-the-parts-of-the-strength-calculus-fit.md","content":"# Why the parts of the strength calculus fit together\n\n- **What it is**: a listening walkthrough of about twenty minutes, made the night of 22 to 23 October:\n  the piece *Why independent reasons add in log-odds*, made an hour earlier, rebuilt to explain the why\n  instead of the numbers. His request: *\"Make the walkthroughs explain the why like this, not numbers.\n  Rebuild the latest walkthrough with this in mind.\"*, then *\"Keep what still works in it ofc.\"*, after\n  *\"I'm more interested in WHY everything in our epistemic engine/machinery works together rationally and\n  elegantly than just having you burp numbers at me. WHY do these numbers result, and does it MAKE SENSE\n  given all that we know? You know.\"* and *\"Could do with some imagination-jogging through a few examples\n  I guess.\"* The first piece written under rule 61 of the method.\n- **Covers**: [../research/how-independent-reasons-add.md](../research/how-independent-reasons-add.md)\n  § 0 to § 9; [../research/how-independent-reasons-add/argumentation-builders-on-probability.md](../research/how-independent-reasons-add/argumentation-builders-on-probability.md);\n  [../research/strength-calculus-literature/computational-argumentation.md](../research/strength-calculus-literature/computational-argumentation.md)\n  § 0, D2, D8; [../research/strength-calculus-literature/bayesian-and-informal-logic.md](../research/strength-calculus-literature/bayesian-and-informal-logic.md)\n  § 2.1, § 2.6, § 3.3; [../research/what-relevance-is.md](../research/what-relevance-is.md) § 3f;\n  [2026-09-26-arc-p-the-april-conversation-audited.md](2026-09-26-arc-p-the-april-conversation-audited.md);\n  ledger Q46, Q41, Q42, Q43, AT8 ([../research/session26-open-rulings.md](../research/session26-open-rulings.md)).\n- **Chapter list for noting** (rule 33):\n  0. Opening: each part answers one question\n  1. What each factor is for: should strength aim at probability?\n  2. Why independent reasons multiply: a claim between for a shared story?\n  3. Why the field built something else: should a sum show its search?\n  4. How often, against how often anyway: rule the revised row?\n- **What was kept from the first version**: the four topics, the four decisions and predictions, his\n  quoted comment on AT8, the magnesium meta-analyst, the forecasting result in a line, Godden and Zenker,\n  vitamin C, red sky, strep, Carnap, every elder line but the first. **What changed**: each part now opens\n  on what a piece is for and why the result follows; the figures of the first version (nine-tenths, nine to\n  one, eighty-one to one, the bile-duct percentages, a tenth becoming a hundredth) are gone; new pictures\n  carry the why: false stories ruled out by witnesses at different windows (why odds multiply), a press\n  release in two newspapers (why a shared story counts once), a rumour against an inspection report (why\n  strength matters), Pearl's alarm and earthquake (why probability also flows backwards), and two roads to\n  one formula (why Potyka's logistic formula and Bayes meet).\n- **Objects from the graph** (rule 39): the Anthropic thresholds link (told in the walk of 22 October);\n  the two ACLU sentences supporting `claim_b8c24819db70` by `auto_connect` (*Modern evidence demonstrates\n  that innocent people are frequently convicted…*; *Scores of individuals wrongfully convicted of murder…*),\n  read here as one story. Read while preparing and not used: `claim_5d2fb2e40638` (Card and Krueger), whose\n  eight supports are critical-question poles (`cq_positive`) that the strength code already keeps on the\n  question channel rather than counting as reasons (`badge_inputs.py`, *a question's own pole … never lends\n  its own standing a second time*).\n- **Registers used** (rule 57): Lucy the spine; his clone on two of his sentences and four predictions;\n  Elliot on Jeffrey (`bayesian-and-informal-logic.md` § 2.1), Pearl's explaining away (Pearl 1988, § 3.4\n  there; the alarm and earthquake example told from memory) and Carnap (§ 3.3); Hope on which part was wrong and on red sky; Scotty\n  as a meta-analyst (the note § 5c) and as the padding critic (the field report § 4, § 6); Anna on Potyka's\n  logistic formula (the field report § 1); Myrddin in every part.\n- **Correction spoken** (rule 37): the 0.57 was the pairing of measuring from the middle with QEM's square,\n  not QEM alone (the note § 0 item 1, § 9).\n- **Numbers kept**: a third and one in three (the red-sky rule's own words), one in six and about half\n  (strep, the lift the part is about), ten days (vitamin C), seven trials, ten thousand whispers, four\n  witnesses, 1999 (Sally Clark). Sally Clark's case is public record: convicted 1999, conviction quashed\n  2003; the expert's squared figure is told without its number.\n- **His words quoted**: 23 October, his report (*\"I'm more interested in WHY…\"*, the slash read as *or*);\n  22 October, his comment on AT8, typed the same night, so the record lookup does not find it yet.\n- **Predictions**: as in the first version (its header).\n- **Rendered**: 21.8 minutes (forecast 20:04 to 22:36; 149 words a minute while speaking, 106 overall),\n  5,768 voice credits, six chapter marks; audio in `.private/local-only/listening-audio/`.\n- **Status**: snapshot of 23 October; supersedes [2026-10-23-why-independent-reasons-add.md](2026-10-23-why-independent-reasons-add.md),\n  which stays as the first version.\n- Method: [how-to-write-for-listening.md](how-to-write-for-listening.md).\n\n---\n# Why the parts of the strength calculus fit together\n\n[[scene Lucy opens a late walk for Fredrik, just after midnight on 23 October. He asked for the why behind tonight's arithmetic, not the numbers, and for examples he can picture. She is calm, warm and unhurried, a teacher who enjoys showing how the pieces of a machine fit.]]\n\nWhy does tonight's arithmetic make sense?\n\n[[pause 5]]\n\nHere is the short answer. Each part of the calculus answers one question. And log-odds is the way of combining their answers that keeps every one of them honest.\n\nThe silences are deliberate. They are room to think.\n\n[[pause 4]]\n\nBefore listening to the last walk, you wrote:\n\n[[voice fredrik]]\n\n[dryly] I'm more interested in WHY everything in our epistemic engine or machinery works together rationally and elegantly than just having you burp numbers at me.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nSo this walk is rebuilt, with very few numbers. Each part asks what a piece is for, and shows it working in a case you can picture.\n\n[[pause 4]]\n\nThe map has three groups. First, what a single reason should do, and why. Second, why independent reasons multiply, and when they must not. Third, why the field built something else, and why how often an outcome follows is not enough.\n\n---\n\n# Part one. What each factor is for\n\n[[scene Lucy takes the part that decides the others: the three questions behind a single reason, and why the old curve silenced the answer. Elliot brings Richard Jeffrey. Hope asks which part was wrong. Myrddin speaks briefly. Mood: patient, a little playful.]]\n\nYou learn one new thing about a claim. How far should it move you?\n\n[[pause 5]]\n\nIt depends on three things. Where you stood before. How sure you are of the new thing. And how much it bears on the claim.\n\n[[pause 4]]\n\nStart with where you stood. I will call it the claim's start. If you know how often claims of this kind turn out true, that is the start. If you know nothing, the start is the middle: no evidence either way.\n\n[[pause 4]]\n\nThen how sure you are of the reason. A rumour says a bridge failed its inspection. Then the inspection report says the same. The report should move you far more.\n\n[[pause 4]]\n\nAnd a reason you can only guess at, a coin flip, should not move you at all.\n\nThe site already works this way. A reason nobody has examined lends its claim nothing.\n\n[[pause 4]]\n\nThen how much the reason bears. A true fact can bear fully, partly, or not at all. In the graph, Anthropic says it grows hard to tell whether its models have crossed certain thresholds. That supports a claim about every frontier company. How far? Only as far as one company's case speaks for the others.\n\n[[pause 4]]\n\nI will call that third factor relevance, and the second strength.\n\n[[voice elliot]]\n\nThe philosopher Richard Jeffrey put the three together in the sixties. Your belief moves from where it started toward what the reason would show. It moves as far as you trust the reason, and as far as the reason bears.\n\n[[pause 4]]\n\nYour two factors, ruled in September, are his rule exactly. That is why a fully trusted, fully relevant reason carries its claim all the way to its own strength. Common sense says the same.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nSo why did the site's code barely move the claim? Two good parts met in the wrong combination.\n\n[[pause 4]]\n\nThe site measures every reason from the middle. That is right: a coin-flip reason should push nothing. But it means pushes stay small. Even a very strong reason pushes less than half as hard as the curve expects from a full argument.\n\n[[pause 4]]\n\nThen the code fed those pushes to a curve from the field, the quadratic energy model. Nico Potyka built it to score debates. There, an argument's strength starts at zero and grows, and faint arguments should only whisper. So the curve squares a small push, which shrinks it further.\n\n[[pause 4]]\n\nSo every reason whispered, even a strong one.\n\n[[pause 4]]\n\nThat is my correction. Earlier tonight I blamed the curve alone. Fed full strengths, as Potyka designed it, the curve lets a strong reason speak. The silence came from the pairing.\n\n[[voice hope]]\n\nSo which part was wrong?\n\n[[voice lucy]]\n\nNeither, alone. Measuring from the middle is right for us. Squaring is right for his debate scores. Together, they cancel the reason.\n\n[[pause 4]]\n\nAnd here is why that matters beyond one curve. Log-odds is right only if a claim's strength means how likely the claim is. If strength meant how the debate stands, a curve that quiets faint arguments could be defended.\n\n[[pause 4]]\n\n[[voice myrddin]]\n\nA tool is only as good as the question it was made to answer. Before you blame the tool, ask whose question it was built for.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nSo the decision underneath this part. Should a claim's strength aim at how likely the claim is, given the record? The weight beside it, how much of the record has been examined, was ruled tonight.\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[firmly] Yes, rule it. I want a stranger to trust a claim's strength as its strength in the world. The weight beside it says how much of the world was looked at.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, a strong reason barely moves its claim on the site. Built, a claim would say what its record supports, and each reason's share would be plain to see.\n\n---\n\n# Part two. Why independent reasons multiply\n\n[[scene Lucy explains why two independent reasons multiply rather than average, with witnesses, a newspaper story and a trial that went wrong. Scotty borrows a confident meta-analyst's voice. Anna brings forecasting. His clone gives the prediction. Mood: vivid, then sober.]]\n\nTwo witnesses say the same thing. When does the second one count?\n\n[[pause 5]]\n\nPicture every way the world could be, as a crowd of stories. In some the claim is true. In others it is false.\n\nA good witness who saw the event rules out most of the false stories. Not all: she could be mistaken. But most.\n\n[[pause 4]]\n\nNow a second witness, standing at another window. She rules out most of the false stories too. Including most of those that survived the first witness.\n\nEach witness removes a share of what is left. A share of a share. That is why their evidence multiplies.\n\n[[pause 4]]\n\nOn the scale called log-odds, multiplying turns into adding. So each independent witness simply adds her share of evidence. The order does not matter. And nothing runs out of room near certainty.\n\n[[pause 4]]\n\nAveraging would treat the second witness as worth nothing new. That is right in one case only: when she saw nothing the first did not.\n\n[[inhales]] Picture two newspapers printing the same press release. Two papers, one story. If the release was wrong, both are wrong together. The second paper rules out no false story the first had not.\n\n[[pause 4]]\n\nThat is tonight's ruled question. Suppose the claim is false. What most likely explains each reason? If one story explains both, the second adds nothing.\n\n[[pause 4]]\n\nThere is a pair like that in the graph today. Two sentences from the ACLU. Innocent people are frequently convicted. And people wrongly convicted of murder have been sentenced to death.\n\nSuppose such errors were rare. What would explain both sentences? One story: one organisation reading one record of exonerations. So the pair would count as one reason.\n\n[[pause 4]]\n\n[takes a slow breath] Getting this wrong has ruined lives. In 1999 Sally Clark, a British mother, was convicted after two of her babies died. An expert squared the chance of a single cot death. He treated the two deaths as strangers to each other.\n\n[[pause 4]]\n\nSuppose she was innocent. What explains two deaths in one home? Whatever made the first likely there made the second likely too.\n\n[[pause 4]]\n\nOne story, not two. Her conviction was overturned years later.\n\n[[voice scotty]]\n\n[confidently] Seven small trials of magnesium after a heart attack, all pointing the same way. Pool them, and the benefit is large and certain. Who needs a big trial?\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThose were my words, in a meta-analyst's manner. A trial far larger than all of them combined later found no benefit at all.\n\n[[pause 4]]\n\nThe shared story: small trials that found a benefit got published, and the others mostly did not.\n\n[[pause 4]]\n\nOn the real cases tried tonight, the picture held. Tests with different mechanisms added almost exactly. Signs of one illness added a little less, because they share a cause. The site's code was the furthest off every time.\n\n[[voice myrddin]]\n\nAgreement is cheap when it comes from a single source. Ask of every chorus how many throats it has.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe decision. When the check finds a single story, the reasons could go under a claim of their own, which supports the conclusion once. The trials would sit under: small trials found fewer deaths with magnesium. And under that link: these trials were not selectively published. Should that be ruled?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[curious] Yes. Then the shared story sits on the map, where someone can attack it, instead of hiding in a strength. And a crowd of reasons right under a claim tells me a claim is missing in between.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, every reason on the site adds, shared story or not. Built, the check would ask the rival question of each pair, and a pile of echoes would count once.\n\n---\n\n# Part three. Why the field built something else\n\n[[scene Lucy turns to the researchers who build argument-strength formulas, and asks why their choices differ from tonight's. Anna reads the builders' papers. Scotty voices their sharpest worry in a borrowed voice. Elliot brings Judea Pearl's alarm. Myrddin speaks. His clone gives the prediction. Mood: curious and fair.]]\n\nWhy would careful researchers build a formula that is not about probability?\n\n[[pause 5]]\n\nBecause they set out to score a debate, not to estimate the world. A debate needs no start, no base rate, no check that two arguments are separate. A probability needs all three.\n\n[[pause 4]]\n\nSo most authors leave probability out of the result. Leila Amgoud and colleagues put it, if anywhere, in an argument's starting weight.\n\n[[pause 4]]\n\nThey chose principles instead: supporters should raise a claim, attackers lower it.\n\n[[voice anna]]\n\nAnd yet Potyka himself later wrote a second formula. He read a neural network as an argument graph. A claim's log-odds there is its starting log-odds, plus what its supporters add.\n\n[[pause 4]]\n\nHe chose it because its results read simply, between zero and one. His paper does not speak of probability. But it is exactly how independent evidence combines.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nTwo roads to one formula. One road wanted a tidy scale. The other followed what evidence does. When two roads meet, that is a reason to trust the meeting point.\n\n[[voice scotty]]\n\n[irreverent] Fine. But then piling up weak reasons makes any claim climb. Potyka and Booth warned about exactly that. Ten thousand whispers make a shout.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThose were my words, in a critic's manner. The worry is fair, and part two answers half of it. Whispers from one story are one whisper.\n\n[[pause 4]]\n\nThe other half is the witness picture. Four weak witnesses, each at a different window, each rule out a different slice of the false stories. Godden and Zenker show such reasons: none alone makes a claim likelier than not, together they do. That is not padding. That is evidence.\n\n[[pause 4]]\n\nWe found no published formula that pairs log-odds with an independence check. That pairing is ours.\n\n[[voice elliot]]\n\nOne difference stays, and Judea Pearl's alarm shows it. Your burglar alarm rings: a burglary becomes more likely. Then the radio reports an earthquake. The burglary becomes less likely again, because the earthquake already explains the alarm.\n\n[[pause 4]]\n\nIn probability, news travels back and sideways like that. In the graph, strength flows only forward, from reasons to claims. That is owed to the build.\n\n[[voice myrddin]]\n\nA formula is a promise about what its strengths mean. Keep the promise, and the formula follows from it.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nA worry stays from the September walk. A claim where only its supporters showed up still climbs. Strength given the record is only as good as the record.\n\n[[pause 4]]\n\nSo the weight sits beside the strength. And an open row goes further. Beside a sum of independent reasons, it shows what the sum is really worth. And how the search went on each side. The case against, never sought. Should it be ruled?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[earnest] Yes. A sum without its search is the participation problem all over again. Show me what nobody looked for.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, the site uses the field's debate-scoring curve. Built, it would use the field's own logistic formula, paired with an independence check, a pairing we found nowhere else.\n\n---\n\n# Part four. How often, against how often anyway\n\n[[scene Lucy takes the rule another session revised tonight. Hope tries a sentence about red sky. Elliot gives Carnap's distinction. His clone reads his comment from tonight and the prediction. Myrddin closes. Mood: light, a small discovery.]]\n\nMost people who take vitamin C are over their cold within ten days. Is that a reason to take it?\n\n[[pause 5]]\n\nNo. Most colds are over by then anyway.\n\n[[pause 4]]\n\nAnother session proposed a rule tonight. In a case, a link's relevance would be the rule's strength, times how often its outcome follows. Its example: about a third of red-sky mornings bring rain.\n\nYou wrote, tonight:\n\n[[voice fredrik]]\n\n[warmly] Outcome-tracking definitely makes sense to me, seems bayesian enough...\n\n[[voice lucy]]\n\n[[pause 4]]\n\nIt is a subtraction short. Think of the stories again. A reason matters by how many false stories it rules out. How often rain follows red sky says nothing about that, until you know how often rain comes anyway.\n\n[[pause 4]]\n\nAs first written, the rule would have made rain likelier than not on a red-sky morning. Yet the rule itself says one in three.\n\n[[voice hope]]\n\nSo red sky means rain is likely?\n\n[[voice lucy]]\n\nOnly if rain is rarer on other mornings. If rain comes just as often anyway, red sky tells you nothing.\n\n[[pause 4]]\n\nNow a case where a modest chance is big news. Among sore throats in general, strep is the minority, about one in six.\n\n[[pause 4]]\n\nWith all four signs of the clinical score, about half are strep. Half is not a high chance. It is a big lift.\n\n[[voice elliot]]\n\nRudolf Carnap drew this distinction in the sixties. Two meanings hide in the word confirm. Firmness: how probable a claim is, given the evidence. And increase in firmness: how much the evidence raised it.\n\n[[pause 4]]\n\nA stated proportion is firmness. A reason's relevance is the increase. Vitamin C has firmness and no increase. The throat signs have modest firmness and a large increase.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nSo the repair. The proportion sets what the reason alone would make of the claim. Relevance is how far that moves the claim from its start. Below the start, the link becomes an attack.\n\n[[pause 4]]\n\nThe comparison is the claim's start: the base rate, wherever one is known. Rules with named exceptions need no change.\n\n[[voice myrddin]]\n\nA sign is worth what it adds to what you already expected. The rest is the weather.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nThe row's own session revised it this way, just after midnight. It is open. The decision: should the rule, as revised, be ruled?\n\n[[pause 5]]\n\nMy prediction.\n\n[[voice fredrik]]\n\n[thoughtful] Yes, rule it. That's what I meant by outcome-tracking. How much more often it follows, not just how often.\n\n[[voice lucy]]\n\n[[pause 4]]\n\nToday, no rule claim in the graph carries a proportion yet. Built, the pass that writes a rule would also bring the claim's start.\n\n---\n\n[[scene Lucy closes the walk, brief and warm.]]\n\nThat is the walk. One word a chapter, if you have one: muddled, agreed, or disagreed and why.\n"}