Associate Professor Hayashi gave a research presentation at the 12th International Conference on Computational Social Science (IC2S2 2026), held in Burlington, Vermont, USA, from July 28 to 31, 2026.

This study built a data market simulation incorporating large language models (LLMs) as autonomous buyer and seller agents, and analyzed how regulatory interventions and market structures affect the growth, stagnation, and decline of data markets.

The presentation examined two institutional design scenarios. The first was a "domain-specific trading regulation," in which trading of data in a particular domain is prohibited partway through the simulation. The second compared a "general-purpose market" handling data across diverse domains with an "industry-specific market" limited to a particular domain.

The analysis showed that prohibiting trade in a specific domain does not necessarily lead to an immediate contraction of the overall market; buyers may instead shift their demand to data in adjacent domains while maintaining their original analytical objectives. It also found that industry-specific markets tend to grow rapidly in their early stages, since buyers' objectives are more likely to align with the data on offer, but tend to saturate relatively quickly. General-purpose markets, by contrast, grow more slowly at first, but as diverse demand continues to emerge, they eventually surpass industry-specific markets in trading volume over the long term.

These results indicate that institutional design for data markets needs to account not only for short-term trading efficiency, but also for participants' substitution and adaptation of demand, domain diversity within the market, and long-term market dynamics.

Through this research, we will continue to clarify how regulation and market structure affect data distribution, and to advance research toward the institutional design of sustainable data markets and data ecosystems.

Title: Do Data Markets Grow or Decline? Institutional Dynamics Explored through LLM-Based Multi-Agent Simulation

Authors: Jun Sashihara, Yukihisa Fujita, Kota Nakamura, Masahiro Kuwahara, Teruaki Hayashi