AI

We leverage machine learning, generative AI, agents, and other technologies to support item search and decision-making that better understand user intentions and preferences and to improve transaction reliability. We also explore new marketplace models based on AI and conduct research and development for next-generation C2C platforms.

Research Topics

Development of high-precision search and recommendation models for secondhand markets

We are developing a generative AI-based model that improves the precision of searches for “one-of-a-kind” items in secondhand markets. By training models on an integrated dataset combining Mercari’s vast transaction data with review data synthesized using LLMs, we aim to transform the search experience in secondhand marketplaces.

AI support for multilingual and multicultural communication

In communication between users with different languages and cultures, we analyze trust-building processes and work on designing AI interventions to reduce the anxiety and contextual discrepancies caused by language and cultural differences. We develop tools and platforms that facilitate mutual understanding among people from diverse cultural backgrounds and improve the efficiency of crossborder communication.

Development of marketplace world models

We construct world models of C2C marketplaces themselves. Within these models, we train various AI agents and verify the interactions that emerge between them. We also use these models to examine marketplace structures and design marketplaces that can grow sustainably.