REV. 2026-09

What we mean by human state

A signal is not a human state.

Skin conductance is not stress; a paused cursor is not hesitation. The state is an inference built on the signal, and the inference is where the claims — and the responsibility — begin. This page fixes the terms the rest of the site uses.

The six principles

A signal is not a human state.

A measurement is an input, never an answer. Skin conductance, eye movement, voice pitch, choice latency: these are signals, and the human state is an inference built on top of them. Copy that lets a signal masquerade as a state is wrong before it is inaccurate.

In practice: every human-state term on this site resolves to a construct, a signal, and a stated inference boundary — or it is not used as a state.

Uncertainty is data.

A confidence band, a contested link, an honest null — these are results, not embarrassments. A system that hides its uncertainty is asking to be trusted rather than checked.

In practice: uncertainty renders in the same visual weight as measurement, and conditional null findings print at full size alongside their tested scope and response variability.

Certainty about evidence, not certainty about the mind.

We do not promise to know what someone feels. We promise you can check what a claim about them rests on: the construct, the signal, the evidence, the boundary. That is the only certainty on offer, and it is the one that survives scrutiny.

In practice: the word “certainty” appears on this site only as the name, or inside this sentence.

Human state deserves first-class standing.

Inferences about people enter intelligent systems as inputs, whether or not anyone declared them. First-class standing means declared constructs, named signals, stated evidence, explicit boundaries — not unlabeled side effects of a model.

In practice: any human-state inference discussed on this site is discussed with its declaration attached.

The loop is the unit of work: Sense → Model → Calibrate → Adapt → Verify.

Measurement is not done when the number appears. It is done when the system's response to the number has been verified. Adaptation without verification is drift.

In practice: the five stages appear as one unbroken loop, never as a feature list.

Claim boundaries are part of the contribution.

A result without its edge is an overclaim in waiting. Stating what a work does not show is not a disclaimer appended at the end; it is part of the result.

In practice: every work mentioned on this site ships with its boundary — “✓ Established / ✗ Explicitly not claimed” — at equal visual weight to its findings.

How claims are labeled here

Every empirical sentence on this site carries exactly one of five tier labels, adjacent to the claim:

  • Established background

    settled science, independent of us. Context, never proof.

  • Supported company research

    findings from this program that passed review or are published with stated evidence.

  • Current work

    the live build and active manuscripts, with explicit status. A status is not a result.

  • Hypothesis

    a testable expectation, stated with the test that could falsify it.

  • Long-term vision

    direction we choose to publish. Aspiration; implies no existing capability.

Labels are never upgraded because a sentence sounds better. If a sentence only works one tier up, the sentence goes.