Future human-system interaction techniques to influence perceived trust

McCabe, Faye ORCID: 0000-0002-3614-9646 (2024). Future human-system interaction techniques to influence perceived trust. University of Birmingham. Ph.D.

Preview

McCabe2024PhD_Redacted.pdf
Text - Redacted Version
Available under License All rights reserved.
Download (2MB) | Preview

Abstract

The maritime defence domain presents unique challenges for the introduction of autonomous systems. With the Ministry of Defence predicting that intelligent information systems are the future of defence, recommending their integration into future systems in order to maintain a competitive advantage, the question is not “if”, but “how”.

The thesis presents a user-centred design approach to the development of automated sonar decision-support systems. It develops this approach through an understanding of Submersible Maritime Platforms (SMPs) as socio-technical systems, evaluating their informational and user requirements using context from the Trust-in-Autonomy and Human Factors literature. Requirements to produce trustworthy autonomy recommendations for the maritime defence domain are presented.

These requirements are then developed further through interviews with Subject Matter Experts, aimed to understand how they deal with informational uncertainty through the Critical Decision Method interview technique, to create a thorough understanding of the tasks involved in broadband sonar classification. This is built on through developing understanding of how they perform their cognitive classification process, in comparison to novices and other SMP crew members, using the Repertory Grid interview technique.

These literature reviews and interviews lead to the development of the VINAS: A Visual, Intelligent Narrative of Autonomous systems. This visualisation uses the cognitive constructs derived from the repertory grid to create an explanation behind an autonomous classifier’s decision-making processes, to allow an operator to evaluate its performance in a transparent and understandable way, in order to encourage appropriate trust calibration.

The VINAS visualisation is then evaluated through two experiments, which show that VINAS increases performance and trust when performing classifications utilising an autonomous classifier.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):