Naoya Iwata Presented Research to Two ROIS Councils

On June 17 and 18, 2026, Naoya Iwata (Nagoya University; cross-appointed to the Digital Content and Media Sciences Research Division at the National Institute of Informatics) gave a research presentation titled “What Should AI Trust? Sources and Authority from the Perspective of Classical Studies” to the Management Council and the Education and Research Council of the Research Organization of Information and Systems (ROIS).

Generative AI is adept at producing fluent, plausible prose, but when used on its own it does not reliably show where its claims come from and may generate false information that looks factual. For medicine, law, scholarship, and every other domain in which the quality of a conclusion depends on the provenance of its evidence, this is a foundational problem. The presentation argued that simply waiting for better models will not solve it: the key lies in designing the context — what is given to the AI, and in what form. Drawing on classical studies, a field that has grappled with sources and authority for 2,400 years, Iwata introduced the retrieval-augmented generation and citation mechanisms implemented in the Humanitext project.

ROIS is the inter-university research organization that includes the National Institute of Informatics (NII), where Iwata holds a cross-appointment. The two sessions offered an opportunity to report the current state of this research to the members of both councils.