Agent Transparency for Intelligent Target Identification in the Maritime Domain, and its impact on Operator Performance, Workload and Trust
Date
2017
Authors
Chan, Gloria Wai Yee
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Thesis
Citation
Statement of Responsibility
Conference Name
Abstract
Objective: To examine how increasing the transparency of an intelligent maritime target
identification system impacts on operator performance, workload and trust in the intelligent
agent.
Background: Previous research has shown that operator accuracy improves with
increased transparency of an intelligent agent’s decisions and recommendations. This can be
at the cost of increased workload and response time, although this has not been found by all
studies. Prior studies have predominately focussed on route planning and navigation, and it is
unclear if the benefits of agent transparency would apply to other tasks such as target
identification.
Method: Twenty seven participants were required to identify a number of tracks based on
a set of identification criteria and the recommendation of an intelligent agent at three
transparency levels in a repeated-measures design. The intelligent agent generated an
identification recommendation for each track with different levels of transparency
information displayed and participants were required to determine the identity of the track.
For each transparency level, 70% of the recommendations made by the intelligent agent were
correct, with incorrect recommendation due to additional information that the agent was not
aware of, such as information from the ship’s radar. Participants’ identification accuracy and
identification time were measured, and surveys on operator subjective workload and
subjective trust in the intelligent agent were collected for each transparency level.
Results: The results indicated that increased transparency information improved the
operators’ sensitivity to the accuracy of the agent’s decisions and produced a greater tendency
Agent Transparency for Intelligent Target Identification 33
to accept the agent’s decision. Increased agent transparency facilitated human-agent teaming
without increasing workload or response time when correctly accepting the intelligent agent’s
decision, but increased the response time when rejecting incorrect intelligent agent’s
decisions. Participants also reported a higher level of trust when the intelligent agent was
more transparent.
Conclusion: This study shows the ability of agent transparency to improve performance
without increasing workload. Greater agent transparency is also beneficial in building
operator trust in the agent.
Application: The current study can inform the design and use of uninhabited vehicles and
intelligent agents in the maritime context for target identification. It also demonstrates that
providing greater transparency of intelligent agents can improve human-agent teaming
performance for a previously unstudied task and domain, and hence suggests broader
applicability for the design of intelligent agents.
School/Discipline
School of Psychology
Dissertation Note
Thesis (M.Psych(Organisational & Human Factors)) -- University of Adelaide, School of Psychology, 2017
Provenance
This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses.
Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the author of this thesis and do not wish it to be made publicly available, or you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at: http://www.adelaide.edu.au/legals
Description
This item is only available electronically.