AI Bibliography

This report (Volumes I and II) documents the successful completion of Phase II of the Defense Advanced Research Projects Agency (DARPA) Improving Warfighter Information Intake Under Stress (IWIIUS) program. Volume I discusses the Phase II integration of cognitive state gauges into operational systems that could demonstrate the viability of augmenting cognition into an operational context. Volume II (Appendices) describes the successful results of four industry teams building Closed-Loop Integrated Prototype (CLIP) systems that demonstrate how the limitations of human cognition can be addressed by augmenting cognition with advanced cognitive state sensors that provide input to complex computational systems. A cognitive bottleneck was identified by each of the four development teams along with empirically testable goals. The bottlenecks were operationally defined in terms of the application environments addressed by each of the four teams.

The Honeywell® research team addressed Attention as their primary bottleneck. Attention was operationally defined in the context of a ground infantry soldier in the U.S. Army’s Future Force Warrior (FFW) program. The FFW will have more information available (via physical sensors, communications, etc.) and processing that information in a dynamic and lethal environment requires careful management of limited attentional resources. During their Concept Validation Experiment (CVE) they demonstrated a 380% performance improvement where attention resources were required, exceeding the 100% improvement goal set by DARPA. Honeywell® was able to correctly classify attention state changes more than 98% of the time, in less than 300 ms, again exceeding the Phase II operational goals. Finally, they were able to change mitigations in well under the 1-minute deadline established by DARPA as adequate for operational testing.

DaimlerChrysler Corporation (DCC) met all their bottleneck performance goals. They addressed the Sensory Input bottleneck that had significant cognitive impact in the context of a Light Armored Vehicle (LAV) operator. A future LAV operator will have multiple cognitive tasks (e.g., communica- tion, planning, command and control) to manage. Sensory input resources are required by these cognitive tasks as well as by the primary job of maneuvering the vehicle. DCC investigated the sensory bottleneck in the auditory and visual modalities, with multiple mitigation strategies. Their experimental results under real-world driving conditions showed that the sensory bottlenecks could be improved by as much as 108% with an accuracy of up to 98%, depending on the modality being examined. The sensory bottleneck status could be detected in as little as 200 ms, and mitigations could be invoked in as little as 0.2 sec, depending on the mitigation used. The DCC team met all DARPA-defined performance criteria for Phase II of the IWIIUS program.

The Lockheed Martin® Advanced Technology Laboratory (LMATL) addressed Working Memory as their primary bottleneck within the Tactical Tomahawk Weapons Control System (TTWCS) simulation environment. They developed their CLIP for a Tactical Strike Coordinator (TSC), who must manage a number of missiles, targets, and shipboard launch platforms, and dynamically reassign the missiles to targets as critical targets pop-up, missiles fail, etc. The operator must recall and recognize far more information than can be maintained in working memory to perform the task effectively. They demonstrated an improvement in working memory throughput by at least 500%

in the context of a U.S. Navy command and control task. LMATL used an Intelligent Sequencing mitigation strategy to strategically present related information about specific missile-target pairings when the Working Memory bottleneck was saturated, achieving a 642% performance improvement, well in excess of the Phase II goal for this bottleneck. Working memory status (high or low load) was correctly identified in over 90% of the trials. The sequencing mitigation took place in less than 500 ms, again meeting all Phase II development objectives demonstrating a successful mitigation for the Working Memory bottleneck.