Tag
This study examines how large language models judge urban safety based on neighborhood names versus coordinates, revealing that names carry both crime signal and demographic bias, with bias scaling with geographic knowledge.