Autonomous Agents and Human Interpersonal Trust: Can We Engineer a Human-Machine Social Interface for Trust?

摘要

There is a recognized need to employ autonomous agents in domains that are not amenable to conventional automation and/or which humans find difficult, dangerous, or undesirable to perform. These include time-critical and mission-critical applications in health, defense, transportation, and industry, where the consequences of failure can be catastrophic. A prerequisite for such applications is the establishment of well-calibrated trust in autonomous agents. Our focus is specifically on human-machine trust in deployment and operations of autonomous agents, whether they are embodied in cyber-physical systems, robots, or exist only in the cyber-realm. The overall aim of our research is to investigate methods for autonomous agents to foster, manage, and maintain an appropriate trust relationship with human partners when engaged in joint, mutually interdependent activities. Our approach is grounded in a systems-level view of humans and autonomous agents as components in (one or more) encompassing metacognitive systems. Given human predisposition for social interaction, we look to the multi-disciplinary body of research on human interpersonal trust as a basis from which we specify engineering requirements for the interface between human and autonomous agents. If we make good progress in reverse engineering this "human social interface," it will be a significant step towards devising the algorithms and tests necessary for trustworthy and trustable autonomous agents. This paper introduces our program of research and reports on recent progress.