Correlating Humans' Trust to System's Action and Inspector's Intervention Behavior in a Hybrid Inspection System

摘要

Recent advancements in computer technology have motivated the automation of various functions of inspection. However, purely automated inspection does have limitations that humans can compensate for given their superiority in pattern recognition, rational decision-making, and adaptability to new circumstances. To achieve best performance and quality of inspection in a hybrid system, functions should be properly allocated between human and machines. Although the classical issues of speed and accuracy have been considered in this cooperative situation, the subjective measure of human trust in automation has not. This topic is an important one, impacting on the use of automation, especially in the context of function allocation. This paper, therefore, investigates the variation of trust as a function of a system's action, specifically the type of errors made by the system using a printed circuit board inspection task. Most significantly, this study found that a linear regression model best fits the human trust in the system as a function of the number of times the human overrides the system's response.