On November 8, 2026, BAKEBAKE member Riri Ikebe of Japan Women’s University presented our yōkai-generating apparatus and its exhibition practice at the 11th Annual Meeting of the Japan Society for Digital Archive, held at Kindai University’s Higashiosaka Campus.
More and more digital archives are open to the public, yet searching and using them still assumes some prior knowledge and interest. The talk proposed starting from people’s everyday experiences, so that non-specialists find their own way into an archive. Our subject is the birth of yōkai. For centuries, people have attributed strange events they could not explain to yōkai, named them, and passed the stories on as shared knowledge.
Visitors describe a strange everyday experience or an unease that is hard to put into words. The apparatus then finds similar stories among about 30,000 records in the Kaii-Yokai Folklore Database maintained by the International Research Center for Japanese Studies. Guided by those stories, generative AI suggests a name, a description, and an image, and visitors choose or revise them, following how a yōkai is born from their own experience.
At our exhibitions, visitors created 66 yōkai, and 37 participants answered our questionnaire. Vague experiences became concrete once they had a name and a form, and some participants took the activity as a way to express their own anxieties and fears, not only as a way to make an image. The results suggest that the experience partly helped visitors understand yōkai as a way of naming experiences that are hard to put into words.
At the same time, a generated result alone does not show which parts come from folklore and which from AI. The talk argued that AI output should be tied to past records rather than simply displayed. Next, we will show the relationship and provenance between each generated yōkai and its source stories, let users trace it back to the records it drew on, and explore adjusting how detailed the folklore data is for each participant.
Talk title: Developing a Yōkai Generation Interface to Describe Inexplicable Experiences via Digital Archives
Practical Challenges in Constructing a Platform with Generative AI and a Yōkai-Kaii Database
Presenter: Riri Ikebe
Co-authors: Hiroki Ichikura, Mirai Hoshikawa, Kosuke Shimizu
The proceedings paper appears in the Journal of the Japan Society for Digital Archive, Vol. 10, No. s2, and is free to read on J-STAGE (in Japanese). Read the paper (J-STAGE)
Project BAKEBAKE