<b>Dialogue with Socrates: LLM-Powered Autonomous Agents in Healthcare</b>

Abstract

Autonomous agents powered by large language model (LLM) will bring paradigm shift to healthcare. Equipped with memory and tool use, agent has the capabilities of perceiving the environment, learning from experience, making informed decisions, and taking appropriate actions to achieve specific healthcare goals. Enabling agents to interact with each other in a collaborative or competitive manner, multi-agents can achieve advancement through teamwork or adversarial interactions. Drawing inspiration from clinical multidisciplinary teams (MDTs), we introduce the M3-agent (Medical Multimodal Multi-Agent), an LLM-driven, multimodal system tailored specifically for the medical domain. Comprising multiple expert agents, the M3-agent engages in discussions and debates, leveraging techniques such as Socratic Method, to diagnose challenging diseases. Furthermore, we explore diverse potential applications of agents in medicine, emphasizing the role of LLM agents as facilitators between doctors and patients, enhancing communication and patient advocacy. Additionally, we identify specific opportunities and challenges for the implementation of agents in medicine, including ensuring interpretability, fostering human-agent collaboration, maintaining ethical consistency, and ensuring output robustness. The application of LLM agents in medicine holds immense potential to revolutionize healthcare by offering enhanced decision support while prioritizing patient well-being.

Ähnliche Arbeiten