WORK


Kinoko vs. Takenoko : Which Side Are You On? AI "MOTHER"

Web

OVERVIEW

AI facial analysis to determine snack preference

This web and mobile interactive experience powered by AI determines whether a user belongs to the Kinoko or Takenoko camp, based on Meiji's iconic chocolate snacks, "Kinoko no Yama" and "Takenoko no Sato" (also known as "Chocorooms" and "Chococones").

A global survey conducted for the 50th anniversary of Kinoko no Yama revealed a striking contrast: while 14 out of 20 overseas countries favored Kinoko, 46 out of 47 prefectures in Japan favored Takenoko. Inspired by these findings, which highlighted how taste preferences are shaped by environment, regional culture, and preconceived notions, this project was conceived as an AI experience designed to reveal an individual's latent preferences, independent of external influences.

By submitting a facial photo via PC or smartphone, the user's facial features are analyzed by the AI across multiple parameters. Benchmarked against an extensive database, the system determines whether the user aligns more closely with Kinoko or Takenoko attributes, delivering an engaging experience complete with an allegiance score, personalized commentary, and a multifaceted breakdown of their characteristics.

RESPONSIBILITIES

Technical direction for an assessment experience integrating facial analysis and generative AI

BASSDRUM provided the overall technical direction for this project, covering everything from initial concept planning and system design to implementation, AI integration strategy, and compliance management.

The core assessment system utilizes proprietary statistical data based on facial images and responses to a 100 question survey from hundreds of Meiji employees. We built a mechanism that converts facial landmarks extracted from the user's image into numerical data, which is then cross-referenced with this statistical model to determine the result. Additionally, we integrated a generative AI system to produce personalized commentary, adding narrative depth to the assessment results.
 
During the development process, we conducted a fundamental architectural redesign. While the initial plan involved sending facial images directly to an external AI, we revised this approach to mitigate the risks of cross-border personal data transfers and to comply with corporate regulations. By designing the system to transmit only quantified facial data points rather than raw images to the AI, we ensured the safety, ethical standards, and high-level security required for a major corporate promotional campaign.
 
Furthermore, we repeatedly refined the template designs and conducted extensive prompt tuning to ensure the generated results were completely free from appearance-based biases (lookism). By comprehensively integrating technologies across multiple domains, including statistical modeling, image analysis, generative AI, and security design, we elevated this project into a highly engaging entertainment experience perfectly suited for a corporate promotion.


THE CLIENT AND OUR TEAM

  • Client: Meiji Co., Ltd.
  • UX Director: Qanta Shimizu
  • Tech Director: Ryuki Nagahora
  • Web Director: Yusuke Takai (Puzzle)
  • Production Managers: Miho Yamashita (Puzzle) / Ayano Fukuda (Puzzle)
  • Designer: Yukari Teramto (Puzzle)
  • Design Engineer: Yota Odaka (BASSDRUM / evenrain Inc.)
  • Backend Engineer: MountPosition
©BASSDRUM inc.