Human-in-the-Loop Multi-Agent Ventilator Decision Support with Contextual Bandit Preference Learning
Researchers have developed a new human-in-the-loop multi-agent framework called the Ventilator Decision Support System (VDSS) to aid in ventilator management. This system coordinates modular decision components through structured interfaces, allowing for traceable evidence and clinician-specific tuning. VDSS uses contextual bandit preference learning to adapt to clinician preferences in real-time, aiming to improve recommendation acceptability and reduce interaction rounds for a more stable and clinically deployable human-AI collaboration. AI
IMPACT This human-in-the-loop system could improve the efficiency and personalization of critical care ventilator management.