Adaptive Automation / Augmented Cognition (DARPA, NASA)
Status: FOUNDATION with critical difference. Closest existing engineering.
What they built
Real closed-loop systems: EEG-driven workload regulation demonstrated (Prinzel et al. 2000, 2002). Cognitive state monitoring + real-time adaptation is established engineering. Physiological indicators (EEG, ECG, EOG, respiration) used as basis for dynamic task allocation.
Critical differences
- Their machine: Adaptive controller that reallocates functions based on workload signals. No understanding of human values, context, or narrative.
- Their actuator: Function reallocation (machine takes over tasks when human is degraded).
- Their objective: Mission performance. Workload/SA are instrumental to performance. "Capacity for self-directed decision-making" as formal objective: NOT FOUND.
- Their design philosophy: "Computer management of adaptive task allocation produces superior performance to human delegation" (Kaber & Prinzel 2006). Machine decides what the human should do. Explicitly machine-as-controller.
What transfers
Closed-loop physiological state monitoring is engineering-feasible. Real-time adaptation based on estimated cognitive state works. Multi-sensor fusion for state estimation is established.
What is different
Everything about what happens after state estimation. Their system replaces the human. Our system stabilizes the human. Their system optimizes mission success. Our system optimizes human capacity. Their machine doesn't understand anything about the human beyond workload.
Positioning statement
Augmented cognition demonstrated that closed-loop cognitive state monitoring with physiological sensing is engineering-feasible. Their design choice was to replace the human when degraded. Our design choice is to stabilize the human so they can keep deciding. The sensing and adaptation infrastructure transfers. The objective function, actuator logic, and machine intelligence requirements are categorically different.
Sources
- Kaber & Prinzel 2006 (NASA/TM-2006-214504)
- Prinzel et al. 2000, 2002