ABSTRACT
This study proposes a classification system that classifies voice intelligent agents (VIAs) into four levels (general, parental guidance, adults-only, and restricted levels) based on both negative dimensions (six content labels and three interaction behaviors) and positive dimensions (three moral intelligence items). An experiment was conducted, and the results validate that the system can predict the classification of the VIA. Furthermore, we also found that different interaction methods have a significant influence on the classification results. The complexity of the interaction significantly lengthens task completion time, and improved evaluator satisfaction leads to different classification results. Therefore, we suggest that evaluators should use text interaction when evaluating the dimension of moral intelligence but should employ the same interaction style as the real scenario when evaluating content labels and interaction behaviors.
Additional information
Funding
Notes on contributors
Xiang Ji
Xiang Ji received her B.S. degree in from Mechanical Manufacture and Automation in Tsinghua University, China, in 2013. She is currently pursuing the Ph.D. degree in Department of Industrial Engineering, Tsinghua University. Her research interests include human–computer interaction, voice intelligent agent, user-centered design, and usability engineering.
Pei-Luen Patrick Rau
Pei-Luen Patrick Rau received the Ph.D. degree in industrial engineering from Perdue University, USA. He is a Professor with the Department of Industrial Engineering, Tsinghua University, China. His research areas include human factors engineering, human–computer interaction, cross-cultural design, design for older people, user experience, human–robot interaction, and service design and evaluation.