SBIR-STTR Award

This proposal is to develop a methodology and an Artificial Intelligence (AI)-based system for calibrating operator trust in the ethical behavior of lethal robots. The purpose is to enable the human-robot team to function at its highest level of effectiveness from the standpoint of ethical behavior as well as technical capability. We have done extensive work on the calibration of operator trust in the technical capabilities of automated systems. A paper including our former employee and present consultant Dr. Ewart de Visser, “Measurement of Trust in Human-Robot Collaboration” (de Visser et al, 2007), is a seminal work in the field and is frequently referenced by other researchers and developers. The present proposal builds on our previous work, as described in Section 3. In Phase I of the present project we will thoroughly review the recent literature on ethics of lethal robots while finding an AF customer to assist in refining our technical approach for Phase II; in Phase II we will develop a prototype system for operator ethical trust calibration and validate it in a proof-of-concept demonstration; in Phase III we will adapt our system to selected Air Force and other service applications and begin our commercialization activiti

This proposal is for Phase II continuation of our project to develop a software framework for Calibration of Ethical Trust for Lethal Autonomous Weapons (CETLAWS) that can be used by Air Force, other DoD organizations and commercial enterprises. The mission impact of this project on the Department of the Air Force will be an enhanced understanding of the roles of ethics and trust in lethal autonomous systems and the availability of new methods to facilitate calibrated trust in human-autonomy teaming scenarios. Ethical trust calibration will increase military users’ willingness to collaborate with lethal robotic technology and will also overlap with key commercial applications in trusted artificial intelligence (AI) and social computing. In Phase I of the present project, we demonstrated and validated an initial CETLAWS using a simulation of a swarm of autonomous lethal unmanned aerial vehicles (UAVs) and a set of ethical equations for making autonomous Shoot/No-Shoot decisions. In Phase II we will implement a complete CETLAWS prototype using both ethical equations and Bayesian modeling for trust calibration, validate this prototype; and begin our transition and commercialization activities with an initial focus on insertion into DAF wargames and test events with the help and collaboration of our USMC end-user