A Mixed Analysis of Influencing Factors for Trust in a Risk-Aware Autonomy

摘要

To study the factor of controllability for humans’ trust in autonomy in decision making, 73 participants used a risk-aware, autonomous planner to navigate an underwater robot in a software-simulated, resource- limited, and risky environment. The experiment examined the association of controllability of the leg size of the path planner with participants’ trust, as well as the underlying reasons. The quantitative analysis showed no significant effect of controllability on trust. However, a verbal data analysis method, which systematically coded and quantified participants’ reasons for choosing their trust levels at different intervals during the experiment, showed that the dimension execution and control was the most frequently mentioned among four emerging practical dimensions of influencing factors for their trust. The other three included risk evaluation; training and learning; and general attitudes. The findings suggested further research on more aspects of controllability, especially the ones that support personal plans, from a human- autonomy interface design perspective.