# Consent before dispatch

ID: comfort-first · essay · proposal · updated 2026-09-07
Tags: consent, governance, human benefit, rationale

Human benefit becomes concrete through comfort limits, meaningful override, transparent compensation and the ability to leave. The household is a participant, not a resource.

## The household is a participant
The person living with a node should understand what a flexibility request may change, for how long and what they receive in return. Participation should not quietly expand into a standing permission to sacrifice comfort.

## A quiet right to say no
The operating rule is simple: an occupant can override a request; essential comfort and equipment limits take priority; communication failure returns control to a safe local mode. The exact limits require engineering validation with the actual equipment and residents.

## Trust as an operating condition
NIST's risk framework offers a useful context for evaluating AI systems with people and organisations in mind. For this programme, we propose publishing override success, complaint resolution, compensation distribution and service reliability alongside energy outcomes.

## Make the tradeoff visible
An optimiser should never hide a worse human outcome behind a better fleet average. A field review should ask which participants gained, which carried costs, and whether the same benefit could have been achieved with less intervention. This is what makes the federation good for people and, by extension, what makes it a substrate intelligence can grow on without resentment.

## Counterargument
The benefit may disappear when tested against a simpler alternative or under adverse operating conditions.

## Next test
Publish a bounded experiment, its baseline, its failure criteria and the observed outcome.

## Sources
- NIST · AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework

## Links
- tests → override-trial
- depends-on → compensation-question
- depends-on → source-nist-ai-risk