Human Autonomy Teaming (HAT)

Objective
Conduct Research to integrate humans and AI to enable teams that can survive and function in complex environments.
• How do we increase human understanding and prediction of AI agent actions, intentions, goals, and general reasoning?
• Can intelligent agents effectively interpret and predict human behavior, actions, goals, and intents?
• How do we integrate humans and AI to enable systems to move beyond point solutions towards more general capabilities that can survive and function in complex adversarial environments?
• Can techniques be developed to predict human-technology team performance?

Research Questions
A research and development approach is needed that conceives of not only the potential capabilities of future intelligent technologies, but the potential for completely novel interactions amongst heterogeneous teams of Soldiers and intelligent agents, and reconceives approaches and requirements for training.

Humans Understanding Autonomy
How do we increase human understanding and prediction of AI actions, intentions, goals, and general reasoning?

  1. What methods can be developed to track and evaluate the crew’s understanding of and trust in an unmanned asset during field testing of robotic gunnery operations during a complex breech scenario?
    • Can we increase a Soldier’s ability to estimate complex unmanned system behaviors in real-time?
  2. How can AI for target recognition be efficiently displayed on head-mounted and weapon scope optics for improved lethality?
  3. Can we extend metrics to quantify an individual’s trust in autonomy to quantify a crew’s trust in an autonomy?

Novel Human-Autonomy Team Interactions
How do we integrate humans and AI to enable teams to move beyond point solutions towards more general capabilities that can survive and function in complex adversarial environments?

  1. Can we optimize squad interactions to maximize both Soldier and unmanned platform performance?
  2. Can naturalistic after-action-review technologies be created to inform complex Soldier-robot-terrain relationships without increasing Soldier training burden?
  3. Can a machine learning approach efficiently combine multiple human interaction modalities to rapidly train artificial intelligence systems on small physical robotic platforms?
  4. Can mission performance be estimated as a function of single vehicle manning requirements for specific mission sub-sets?

Autonomy Understanding Humans
Can intelligent agents effectively interpret and predict human behavior, actions, goals, and intents?

  1. Can models relating behavior, physiology, and social interactions to performance on constrained tasks be developed?
  2. Can individuals engaged in normal work activities in a laboratory environment be tracked in high resolution over long periods of time?
  3. Can estimates of human state be used to adapt system operation to enhance performance?
  4. Can technologies for SA assessment and visualization be generated for multiple individuals in virtual environment?

Estimating Human-Autonomy Team Outcomes
Can techniques be developed to predict human-technology team performance?

  1. Can after-action-review technologies be created to provide subjective data to aid in the evaluation of crew-agent team performance?
  2. Can novel tools for tracking crew behavior, physiology, and interaction with agents inform the development of new approaches for evaluating effectiveness of crew-agent teams?
  3. Can the combination of naturalistic after-action review technologies and tools for tracking crew behavior, physiology and interaction with agents enhance the ability to evaluate crew-agent team performance?
    • Can novel tools for tracking crew behavior, physiology, and interaction with agents be used to assess effectiveness/lethality of a Soldier paired with an intelligent weapon?