A proposed systematic tendency of LLMs to collapse narrative meaning into abstract summary labels rather than preserving reconstructable inferential structure.