自動化システムに対する信頼感の推移:誤動作発生パターンへの依存性

抄録

This paper discusses the dynamics of operators' trust in automated systems. Previous research has shown that malfunctions in an automated system reduce trust of operators and that the trust depends upon the reliability of the automated system. This paper focuses on how the dynamics of trust depends on occurrence patterns of malfunctions as well as reliability of the system. Two experiments are conducted to clarify differences of effects of malfunctioning patterns on trust, where the total failure rate is constant. Firstly, we compare the dynamics of trust in which malfunctions occur at certain intervals with that in which malfunctions occur successively. Results of analyses on indicators, such as the elapsed time until decision and human intervention rate, show that successive malfunctions reduce trust, though malfunctions occurring at intervals have little effect on trust. The second experiment compares the dynamics of trust of operators who have poor experience with that of operators who have ripe experience after experienced successive malfunctions. The results show that trust of less experienced operators fluctuates and becomes stable gradually, and that richly experienced operators rapidly recover the trust that has declined due to the successive malfunctions. A regression model identifies the difference between them. Our results suggest that some occurrence patterns of malfunctions will reduce trust in an automated system even if it is highly reliable.