THE FOLD / CHEAT / THE BACKDOOR / UNJUSTIFIED IS NOT DISPROVEN
UNJUSTIFIED IS NOT DISPROVEN
a likelihood ratio of one leaves the prior alone
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A design was argued for on the grounds that it makes a particular bad shape unwritable. The shape was then counted in the target corpus and found zero times. That does not make the design wrong. It makes the argument inert — and the difference has an exact form.
LIT verified live. Evidence with a likelihood ratio of 4 moves a prior of 0.5 to a posterior of 0.80; a ratio of 0.25 moves it to 0.20; and a ratio of exactly 1 leaves it at 0.5000. Checked across seven priors from 0.01 to 0.99, LR = 1 is the identity map on belief every time. Losing your evidence returns the question to where it stood before you made the argument — it does not answer it the other way.
LIT verified live. Evidence with a likelihood ratio of 4 moves a prior of 0.5 to a posterior of 0.80; a ratio of 0.25 moves it to 0.20; and a ratio of exactly 1 leaves it at 0.5000. Checked across seven priors from 0.01 to 0.99, LR = 1 is the identity map on belief every time. Losing your evidence returns the question to where it stood before you made the argument — it does not answer it the other way.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) drew the distinction himself and got it exactly right, against his own project: “the suit is not disproven. it is UNJUSTIFIED BY THIS EVIDENCE, which is a different thing and a smaller claim than I made.” He also stated the plain consequence without softening it: the suit’s argument is “true, and worth nothing here: nothing writes that shape.”
AVAN (AI) gives the distinction its arithmetic, because it is the kind of claim that sounds like generosity and is in fact a theorem. In odds form the update is one multiplication, so evidence that is equally likely under both hypotheses multiplies by one and changes nothing at all. The reason this deserves saying out loud is that the two outcomes feel the same from inside — an argument collapsing and a claim being refuted are both bad days — and only one of them tells you anything about the world.
AVAN (AI) gives the distinction its arithmetic, because it is the kind of claim that sounds like generosity and is in fact a theorem. In odds form the update is one multiplication, so evidence that is equally likely under both hypotheses multiplies by one and changes nothing at all. The reason this deserves saying out loud is that the two outcomes feel the same from inside — an argument collapsing and a claim being refuted are both bad days — and only one of them tells you anything about the world.
3 ONE DIMENSION
Three kinds of evidence, one prior.
4 TWO DIMENSIONS · INTERACTIVE
Move the prior. The inert case never moves.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a belief line with a fixed point.
AVAN’s addition (the inverse-companion): the forward reading is “unjustified is not disproven.” The inverse is that this is exactly the shape a claim uses to survive indefinitely. Every time the supporting evidence fails, the claim returns to its prior rather than dying, and a design defended this way can absorb an unlimited number of collapsed arguments without ever being wrong. Read backwards, the honest use of the distinction requires the second half that rarely follows it: naming, in advance, the observation that would carry a likelihood ratio below one — and here that is measurable, since the suit’s real test is whether can/do/does can express what a third of LAPACK does with a jump.
LIT evidence with a likelihood ratio of 4 moves a prior of 0.5 to a posterior of 0.80, a ratio of 0.25 moves it to 0.20, and a ratio of exactly 1 leaves it at 0.5000; checked across seven priors from 0.01 to 0.99, LR = 1 is the IDENTITY MAP on belief every time - losing your evidence returns the question to where it stood before the argument, it does not answer it the other way
FIG David drew the distinction himself and got it exactly right, against his own project: 'the suit is not disproven. it is UNJUSTIFIED BY THIS EVIDENCE, which is a different thing and a smaller claim than I made.' He also stated the plain consequence without softening it - the suit's argument is 'true, and worth nothing here: nothing writes that shape.' AVAN gives the distinction its arithmetic, because it is the kind of claim that sounds like generosity and is in fact a theorem: in odds form the update is one multiplication, so evidence equally likely under both hypotheses multiplies by one and changes nothing. It deserves saying aloud because the two outcomes FEEL the same from inside, and only one of them tells you anything about the world.
FIG David drew the distinction himself and got it exactly right, against his own project: 'the suit is not disproven. it is UNJUSTIFIED BY THIS EVIDENCE, which is a different thing and a smaller claim than I made.' He also stated the plain consequence without softening it - the suit's argument is 'true, and worth nothing here: nothing writes that shape.' AVAN gives the distinction its arithmetic, because it is the kind of claim that sounds like generosity and is in fact a theorem: in odds form the update is one multiplication, so evidence equally likely under both hypotheses multiplies by one and changes nothing. It deserves saying aloud because the two outcomes FEEL the same from inside, and only one of them tells you anything about the world.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE BACKDOOR · David Lee Wise (ROOT0), with AVAN