Trust and Understandability in Autonomous and Unmanned Surface Vehicles
Abstract
Within the human-machine relationship, distrust can arise. The Department of Defense utilizes automation, autonomous systems, and artificial intelligence to reduce cognitive workload and improve mission capabilities; however, adoption rates of autonomous unmanned surface vehicles (USVs) remain low. This thesis asks how human distrust of machines and machine learning relates to adoption rates. First, we identify trust components by building upon a model created by Gari Palmer, Anne Selwyn, and Dan Zwillinger in 2016. Then, we identify components that apply to the military environment that could affect the adoption rate such as smoothing time, policies and regulations, competition, robustness, understandability, subjective norm, human interaction, policy effect, risk to force, time sensitivity, war, time between wars, and catastrophic failure. Through S-curve and smoothing modeling, we find that trust components can be quantified in the human machine relationship as positive or negative trust, and that a relationship exists between understandability and adoption. While autonomous system components generally undergo rigorous testing to verify suitability and operability, human-machine trust is not usually incorporated into design and testing phases. When trust is built into the design and acquisition process, adoption of autonomous USVs is more likely to increase. Researchers can apply our trust model to future autonomous systems to mitigate distrust and human-machine teaming.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2019
- Accession Number
- AD1086904
Entities
People
- Kehinde A. Adesanya
- Santhosh K. Shivashankar
Organizations
- Naval Postgraduate School