Scientific Liquidity Agents: A Multi-Agent Framework for Modeling Knowledge Markets in Scientific Ecosystems

This project builds a computer model that simulates how scientific knowledge changes over time. The model uses many agents—each powered by a large language model—to act like individual researchers. By watching how these agents behave, the model shows how their combined actions drive scientific progress in a simulated “knowledge market.” The agents follow the Belief‑Desire‑Intention (BDI) cognitive framework. Each researcher agent has its own skill set, typical behavior, and set of incentives, so the population is diverse. On a larger scale, the system keeps a constantly updating knowledge graph and a collaboration network. These structures record how information and ideas move across different fields. A special type of agent, called

an Epistemic Arbitrageur, links otherwise separate domains and cuts down on “knowledge friction.” The simulation measures things like how productive each agent is, how roles differ, and what publication patterns emerge as the model runs through many steps. Overall, the framework gives a clear way to study how ideas spread, how researchers work together, and what structural factors affect the flow of knowledge in science. It can be used for computational social science, mapping knowledge in systems biology, and evaluating research policies.

Message: Hi FE, I need help with “Scientific Liquidity Agents: A Multi-Agent Framework for Modeling Knowledge Markets in Scientific Ecosystems” in “Computational Biology / Systems Biology / Multi-Omics Integration”