A configuration of a human-AI workflow in which final decision authority is retained by the human rather than transferred to the AI system, characterised by the human's capacity to set, confirm, or override any AI-proposed action across delegated tasks.
STATE (graded). The degree of retained human decision authority is rated per workflow on a 1-5 ordinal scale, anchored 1 = AI executes consequential actions without a human confirmation point, 5 = every consequential action requires explicit human confirmation. Rating is based on documented control points in the workflow, not on user self-perception.
Proposed measurement protocol (not yet empirically validated): Workflow audit by trained raters who map each consequential action to its control point and assign the 1-5 authority-retention score; agreement across raters could be reported as ordinal ICC (two-way, agreement). Complementary objective indicator: proportion of consequential actions gated by a human confirmation step.
https://andreasehstandlicenseofclarityloc.github.io/augmanitai-periodic/#topic-meta-reflection