Monitor Accuracy Isn't the Answer: A Decision-Value Framework for Evaluating AI Agent Restart Monitors

A decision-value framework shows that classification accuracy and true-positive and false-positive rates alone do not determine whether restarting an AI agent improves success, according to an analysis using an exact alarm-to-action identity for an independent clean restart with zero monitoring overhead. The analysis demonstrates analytically that a monitor with fixed 90% accuracy does not by itself establish that a restart improves agent outcomes.

Classification accuracy and true-positive and false-positive rates alone do not determine whether restarting an AI agent improves success. Using an exact alarm-to-action identity for an independent clean restart with zero monitoring overhead, we show analytically that a monitor with fixed 90% accura