State Partnership Program

Video by Kevin D Schmidt

Dr. Mica Endsley

Air Force Research Laboratory

March 31, 2023 | 53:43

Description:
In this edition of QuEST, President of SA Technologies and former Chief Scientist of the Air Force Dr. Mica Endsley discusses Situation Awareness in AI Human and Machine Teaming.
Key Moments in the video include:
Outlining the need for ‘effective oversight of AI and Autonomous Systems’
Synergistic Human-Autonomy Integration
Team Situation Awareness and the importance of Transparency and explainability
Team Situation Awareness and Shared Situation Awareness
Real-time information and mental models of a system both ‘feed’ Situation Awareness
Shared Situation Awareness needs in Human-AI teams
Taskwork and SA
Design principles for Situation Awareness
Tesla autopilot example
Measuring SA

Audience questions:
What about commander’s intent? What role can transparency and explainability play in ensuring that the machine has the correct interpretation of commander’s intent?
I have a question about confidence and meta-awareness, and the differences or similarities - confidence for the task SA and meta-awareness for the agent SA - are those the same things or are there differences between them?
How do we deal with the fact that both parties, agents and humans, are both learning?
How do we know what is being explained by the human or the agent - does a human know that there are certain things that a machine can do so well, such that it develops a belief or confidence in the machine? How much are we banging on the door of being able to explain ourselves to each other - machine and human?
What happens when you have more than one AI/AI system?
I’ve applied GDTA and I’ve found that you get a lot of information requirements for an operator - how do you pare that down to just what the operator needs at that time and not overwhelm them?