The effect of an interaction does not end when the interaction ends. This follows directly from how biological generative models update their priors.
- Bayesian parameter updating. An interaction does not affect only the immediate instant. The sensory data you provide forces the other node's brain to update its internal generative model. The update persists.
- Post-interaction hysteresis. If an interaction leaves a node in high ΔF, its system continues to loop on that prediction error long after you have walked away. An animal stays on high alert; a non-verbal partner's felt-sense matrix remains flooded; a plant exhibits altered physiological stress responses.
- The residual wave. Conversely, an interaction characterised by deep safety, validation and zero friction leaves a residual wave of coherence. The other node continues to operate in a low-energy, harmonised state space long after the event.
The operational consequence is that the unit of responsibility is not the interaction; it is the interaction plus its residue. A sharp exchange that is "over" in ten seconds is not over for the node that received it. Equally, ten seconds of clean presence continues to work in the other system after you have left the room.
The residue is easiest to observe in animals, whose interfaces do not narrate and therefore do not mask. Notice the posture of a companion animal in the period after a tense exchange between humans in the same room, compared with the period after a calm one. The animal was not addressed in either case; it was in the field.