Presence is not surveillance
The prototype begins only after a person explicitly starts a co-presence session. Camera and microphone status remains visible. The design prefers on-device extraction of limited events instead of continuous cloud upload of raw media.
A conservative default
The default action is silence or waiting. A single facial movement, pause, or posture change cannot trigger a claim about emotion. Candidate interventions require persistence, multiple signals, a suitable conversational gap, and no recent request for quiet.
Evidence boundary
The current work consists of an engineering prototype and a design specification. It demonstrates a feasible event-and-policy architecture; it does not yet establish benefit, comfort, or long-term acceptance in a representative user population.