2026-08-28·by Sijie Wang#fact#market

gerbner-model

Gerbner's Model (George Gerbner, 1956)

A perception-first model: someone M perceives an event E as a percept E1, then through some means and controls produces a message SE about it; a receiver perceives that message as SE1. The message is never the event — it is a selected, mediated account.

Which piece of Shannon it adds back: perception and selection at both endsE1 != E and SE1 != SE. Fidelity is doubly lossy, and the loss is a function of access, availability, and the communicator's control over the medium.

Transfer. Agent content is a message about something the agent perceived through training data — an E1 already at a remove from E — then compressed again into SE. Two lossy perceptions stack. This is precisely why agent output about lived, specific reality reads thin: it is a percept of a percept, mid by construction. Closing the gap means grounding in first-hand, verifiable detail to raise E1's fidelity, not producing a more fluent restatement of the low-fidelity percept.

Kin: models, shannon-weaver-model, lasswell-5w, slop-is-a-context-deficit, anti-mid

about this entry

One of sijie's wiki entries. The AI on this site is grounded in the same corpus and answers in sijie's voice, with citations back to entries like this one — answering costs sijie money, so it waits behind a code: enter an access code →