Definition
The rate at which adverse or unexpected outcomes are traced to the interaction of two or more independently validated agronomic models (irrigation scheduler with nitrogen engine, climate optimizer with fertigation and lighting controllers, harvest timing with marketing advisor), per 100 decision-pairs where both models were active.
The Measurement Gap
There is no metric that tells you how often two independently sound models (irrigation and nitrogen; climate, lighting, and fertigation) interacted to produce an outcome neither predicted, which means each model is validated alone and the system is never validated.
Formula
ACMIR = N_interaction_incidents / N_active_pairs × 100, where an interaction incident is an adverse outcome whose root cause review attributes it to the joint action and not to either model's own error
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Which model pairs need a shared constraint or a coordinator before both are allowed to act automatically on the same field or house.