CCC, V Point Marketing, and Matsuo Institute Joint Research Paper Accepted at the 2026 Annual Conference of the Japanese Society for Artificial Intelligence

Culture Convenience Club Co., Ltd. (Headquarters: Yokohama, Kanagawa; President and CEO: Takanori Takahashi; hereinafter "CCC"), V Point Marketing Co., Ltd. (Headquarters: Yokohama, Kanagawa; Representative Director and President: Seigo Hirota; hereinafter "V Point Marketing"), and Matsuo Institute, Inc. (Headquarters: Bunkyo-ku, Tokyo; Representative Director: Nobuyoshi Kawakami; hereinafter "Matsuo Institute") are pleased to announce that their joint research paper, titled "Integration of Multi-Domain Behavior Logs and Behavior Prediction Using Large Language Models," has been accepted at the 2026 Annual Conference of the Japanese Society for Artificial Intelligence (the 40th). The research results are scheduled to be presented at the conference on Monday, June 8, 2026.

In recent years, the values and behaviors of consumers have been rapidly diversifying, making it difficult to capture their ever-changing needs using conventional approaches alone. To address this challenge, CCC and V Point Marketing, in collaboration with the Matsuo Institute, have been conducting joint research since September 2025, aiming to gain a more three-dimensional and multifaceted understanding of each individual consumer by combining "language data" such as surveys with "actual behavioral data" such as purchasing behavior. This joint research paper represents a theoretical achievement that serves as a first step toward the social implementation of a "human-centric AI persona agent" based on this deep understanding.

Demonstrating the Effectiveness of Multi-Domain Data for Consumer Understanding (User Modeling)

This research aims to advance "user modeling," which structurally captures the behavior and values of each individual consumer. Previously, behavior analysis using large language models (LLMs) primarily assumed the use of history within a single domain (a specific business type or service). However, many consumers operate across multiple business types and services, and fragmented data alone cannot capture their overall picture. In this study, the researchers investigated and demonstrated the effectiveness of a method that uses LLMs to extract and integrate consistent, textual persona information from multi-domain (multiple business types) behavioral histories. In the application of AI technology, the diversity and volume of data used for learning and analysis determine the resolution of its output. The study academically confirms that higher-precision user modeling is possible through AI technology. This is expected to lead to individually personalized customer experiences in the future. Specifically, by verbalizing the values and decision-making backgrounds of consumers—which cannot be fully grasped from "actual behavioral data" like purchase history alone—as persona text and analyzing it in combination with "language data" such as surveys, it becomes possible to design deeper, more realistic customer profiles than ever before. This is expected to lead to communication and service design that is more attuned to the context of each individual consumer.

Going forward, the companies will consider implementing these findings into services and products that capture the latent needs of consumers with higher resolution. CCC and V Point Marketing, together with the Matsuo Institute, will continue to implement "AI that starts with people" and work to create new value that benefits daily life and society.