Agent Transparency for an Autonomous Squad Member

Report No. ARL-TR-7298
Authors: Michael W Boyce; Jessie YC Chen; Anthony R Selkowitz; Shan G Lakhmani
Date/Pages: May 2015; 84 pages
Abstract: The ability to understand the reasoning behind an intelligent agent's actions can help to increase operator performance as the use of human-agent teams for military operations grows. This experiment tested the effect of display design to convey environment and intelligent agent information in a simulation-based unmanned ground vehicle monitoring task. Three groups were tested with visual displays representing 1 of 3 types of information: current status only (group 1); current status with reasoning information (group 2); and current status, reasoning information, and projected information (group 3). Performance measures included comprehension of situation awareness probes, operator trust based on 3 different surveys, workload, and system usability. Results indicated a significant interaction between conditions and pre- and postadministration of a trust survey modified from Jian et al. (2000), with only group 2 increasing in trust preexperiment in comparison with postexperiment. The situation awareness probes failed to yield any significant differences among the conditions. No significant effects of operator workload or individual difference factors were observed across conditions. This research demonstrates the potential of agent transparency displays to improve Soldier trust and situation awareness.
Distribution: Approved for public release
  Download Report ( 2.420 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: May 1, 2015