Multidimensional Trust Inference in Complex Human-Agent Teams

Abstract

This project develops and studies new computational models of the trust humans have in artificial agents, with the aim of improving human-machine teaming. Compared to the state of the art, our models are multidimensional and take into account not only abilities, but also motivations and social signals. As part of this project, we will develop computational social models and then test the validity of these models in experiments on a simulated testbed in which humans and AI agents work together. We will also use the new models to inform agent architectures. The motivation for this work lies in the understanding that trust is critical to team performance. A better model of the trust humans have in agents and other humans will give artificial teammates an enhanced mental model of the human and enable them to choose actions based on the inferred trust levels. The models developed in this project also support repair actions that can improve trust in the artificial agent and thus promote team performance.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 09, 2020
Source ID
W911NF2010004

Entities

People

  • Guy Hoffman

Organizations

  • Army Contracting Command
  • Cornell University
  • Defense Advanced Research Projects Agency

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML