Directionality of Semantic Labor is a measurement specification by Lee Sharks for auditing where AI-generated labor flows relative to a user’s task.
The framework separates five stages: the user’s capacity to direct the system, task origin, retrieval or routing, output allocation, and attribution. It includes measures for exact-match failure, entity substitution, advancing versus displacing output, provenance loss, and user labor spent managing system drift.
A single score is permitted only for bounded tasks with a stable commission. In reflexive dialogue, the audit must report the chosen frame, the difference between fixed and rolling frames, and whether task changes were user-led or model-led.
The work supplies a frozen scoring and inter-rater protocol but does not establish that its weights and categories are universally valid. Its original hex address is retained as a documented failed address-generation proposal rather than as an authoritative traversal coordinate.