Editing Openai/6973e373-2b8c-800d-84b7-d7001cfc6bac (section)

===== Axiom 1 — Text is behavior, not just content. =====
A thread is a sequence of actions: answering, dodging, conceding, specifying, reframing, straw-manning, updating, etc.

Axiom 2 — Single snapshots are weak; sequences are strong.
One comment can be miscalibration. Repeated cycles under scrutiny are diagnostic.

Axiom 3 — Analyst skill matters.
The strength of inference depends on the analyst’s ability to track premises, repairs, substitutions, and patterns across time—so “text alone can’t tell” is often an overgeneralization.

Axiom 4 — Confidence is not evidence.
Confidence can correlate with truth or with miscalibration; it is not a truth-maker.

Axiom 5 — Tone is not a truth-maker.
Tone may be data, but it cannot substitute for the repair/engagement record.

Axiom 6 — Intent is often unknowable; function is often clear.
Therefore, function-based classification (“effective”) is the disciplined default.

Axiom 7 — The strongest inference comes from repair opportunities.
When a person is offered a low-cost, specific chance to repair—and repeatedly refuses—the causal space collapses rapidly.

Axiom 8 — “Not measured” is not “not present.”
But “not measured” is also not a valid premise for certainty claims.