Class 13: Human-in-the-Loop (HITL) - When to Trust AI and When to Intervene

<aside> 📥

Resources

📄 Lesson PDF

</aside>

Video Lesson

By the end of this class, students will be able to:

Explain why AI systems require human oversight based on their inability to operate beyond design parameters, and identify situations requiring human intervention (Understand)
Calculate expected AI performance thresholds using Watson concordance rates (81.52% overall) as benchmarks for operational decision-making (Apply)
Distinguish between automation bias errors of omission versus commission, and implement verification procedures to mitigate each failure mode (Analyze)
Evaluate operator training requirements to prevent dual trust failures caused by insufficient versus excessive familiarity with AI capabilities (Evaluate)
Design human-in-the-loop protocols that prevent cascading system failures like the USS Gettysburg incident where 15% integration time contributed to friendly fire (Create)