Agent Reasoning Transparency: The Influence of Information Level on Automation-Induced Complacency

Report No. ARL-TR-8044
Authors: Julia L Wright, Jessie YC Chen, Michael J Barnes, and Peter A Hancock
Date/Pages: June 2017; 214 pages
Note: University of Central Florida?Institute for Simulation and Training, 3100 Technology Pkwy, Orlando, FL 32826
Abstract: To understand how the information available to an operator and the transparency of an intelligent agent?s reasoning interact to affect complacent behavior, 2 between-subjects experiments were conducted. Participants supervised a 3-vehicle convoy as it traversed a simulated environment and rerouted the convoy when needed with the assistance of an intelligent agent. In Experiment 1 (low information), participants received information about their current route only; in Experiment 2 (high information), they received information about both their current route and the suggested alternate route. In Experiment 1, access to agent reasoning was found to be an effective deterrent to complacent behavior. However, the addition of information that created ambiguity for the operator encouraged complacency, resulting in reduced performance and poorer trust calibration. In Experiment 2, access to agent reasoning was found to have little effect on complacent behavior, and there were notable differences due to individual differences. These findings suggest that when the operator has more information regarding their task environment, individual difference factors may influence performance outcomes more than access to agent reasoning. These findings indicate some negative outcomes resulting from the incongruous transparency of agent reasoning may be mitigated by increasing the information available to the operator.
Distribution: Approved for public release
  Download Report ( 7.587 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: June 1, 2017