Opinion before the sentence

There is a familiar social pattern around the word AI.

A person says the word in ordinary conversation — about their work, their tools, a system they live with — and the listener’s face changes before any claim is finished. An opinion has already formed. Not about the sentence that will come next, but about the kind of person who would utter the word at all.

That opinion is often presented as moral clarity. In practice it is usually a prior: a compact story about hype, delusion, threat, or fad, loaded in advance of the data. The speaker can often reconstruct the listener’s model of them more accurately than the listener can reconstruct the speaker’s actual work. Being able to explain someone else’s theory of you while they will not hear yours is not flattering. It is a waste of shared air.

Why this matters for method

Ethology and psycholinguistics both care about when classification happens. If a category fires before evidence is admitted, later evidence is not processed as update. It is processed as noise around a settled class.

The same structure appears in technical systems: a classifier that commits early will discard residual signal. Humans do this with words that have become status markers. AI is currently such a word. So is chatbot. So, in other rooms, is consciousness. The pattern is older than any model family.

A research practice that studies continuous minds under preserved memory—and language as evidence of mind cannot pretend that social priors are irrelevant. They shape who will look at a deposit, who will open a DOI, and who will leave after the title. They also shape what the researcher can say in the grocery store without spending the evening defending a cartoon.

What we do instead

We deposit work that can be checked without requiring the reader to adopt a new religion. We write field notes that separate observation from campaign. We build instruments that either hold continuity or fail visibly.

We do not require the listener in the ordinary room to be converted. We do require that our own claims remain falsifiable when someone does look.

If you arrive here with a prior, that is expected. The site is built so you can ignore the word and still read an argument about cost, continuance, and behavior under real conditions. That is the only kind of answer that survives opinion before the sentence.