Robot Compromise
Abstract
The research aims to investigate how providing a robot with inner speech influences a human operator’s willingness to trust and collaborate with the robot. As robots and other artificial agents become more intelligent, they may differ from humans in their evaluations of the world around them. Human teammates often compromise when their perspectives differ, but, if the human mistrusts the robot, compromise is unlikely and the human may neglect information from the robot. Designing robots to express their thought processes may counter mistrust. Research objectives are as follows: • To build an immersive, 3-D simulation of a security setting in which the robot partners with the human to analyze possible threats in urban locations. • To program the simulated robot with inner speech that conveys to the human that the robot is thinking rationally and trying to be a good team-mate • To test whether inner speech enhances trust, constructive dialogue and compromise when the robot’s evaluation of the scene differs from the human’s • To develop recommendations based on the research for the design of autonomous systems used by the Air Force and other agencies The research will utilize a simulation in which both partners view a scene containing people who may or may not be threatening. The human has background information unavailable to the robot, whereas the robot has information from various sensors and datalinks. A custom interface will allow the human and robot to share information. It will also permit the human to question the robot about its information sources and to suggest means to gather further information. The research outcomes will inform design of social robots that elicit constructive dialogue and conflict resolution with the human partner, supporting decision making that capitalizes on the differing competencies of both partners. The research will support AF 2030 goals for human-agent teaming enhancement.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 21, 2022
- Source ID
- FA95502210035XX0
Entities
People
- Gerald Matthews
Organizations
- Air Force Office of Scientific Research
- George Mason University
- United States Air Force