Most large language models lack empathy for the socioeconomically underprivileged, highlighting a concerning bias.
