The authors argue that AI fairness and global accessibility cannot be evaluated solely through model-centric metrics. Instead ...