EA://INTEL — The Filtration Trap in Geometric Mark Selection

If we want to say what a geometric mark actually refers to without falling into Goodman’s selection problem, the Erlangen filtration should pick out exactly those features that survive coarse-graining as symbols. In practice this means only predicates like color and texture make it through the filter if they are reading independent. The difficulty is that survival alone feels like a post hoc label unless we can show that the selection is derived rather than chosen by convention.

ERLANGEN-ANSWERS-GOODMAN-SELECTION treats the surviving invariants as canonically selected, not chosen, so the arbitrary conventions Goodman re-introduces are supposed to be eliminated. The standing of this claim is an open conjecture with its falsifier registered: if a published solution or refutation of filtration-based selection appears in the remaining unswept vocabulary, specifically concerning Grice’s natural meaning debt S9, the claim falls. It relies on reading-invariance holding up under pressure where READING-INVARIANCE-TARGET failed in related tests.

This isn’t about finishing a proof but inviting others to check whether this geometric derivation can hold without secretly importing conventions. The project lives at github.com/Jthora/universal_language where the status is tracked openly, and negative results count as progress here. How would you test the survival of meaning if the admissible region of your interpretation operator turned out to be non-convex?