Design and user experience analysis of AR intelligent virtual agents on smartphones. (March 2023)

• Record Type: Journal Article
• Authors: Gan, Quehong; Liu, Zhen; Liu, Tingting; Zhao, Yumeng; Chai, Yanjie
• Is Part Of: Cognitive systems research. Volume 78(2023)
• Page Start: 33
• Page End: 47
• Publication Date: 2023-03
• Subjects: Intelligent virtual agents -- Augmented reality -- Cognitive architecture -- Emotional expression -- Uncertainty -- User experience; Cognition -- Periodicals; Cognitive engineering (System design) -- Periodicals; Artificial intelligence -- Periodicals; 153.05
• DOI: 10.1016/j.cogsys.2022.11.007
• Languages: English
• ISSNs: 1389-0417

Abstract

Abstract: Intelligent Virtual Agents (IVAs) can provide users with a friendly experience and have a wide range of applications in the era of artificial intelligence. However, most of existing IVAs are designed for personal computers. Design and user studies of IVAs on smartphones are uncommon. Therefore, developing IVAs for smartphones is an interesting topic. Considering Augmented Reality (AR) technology can provide more potential application value for IVAs, we mainly investigate users' experiences of AR IVAs on smartphones in this paper. To make an IVA more suitable for a smartphone, a lightweight IVA's cognitive architecture is proposed. To find out the factors that affect users' interaction experiences, the effects of humanoid embodiment and emotional expressions of IVAs on users' perceptions and experiences are explored. A museum is used as a specific task scenario to measure users' experiences. Three forms of AR agents are evaluated in this scenario: a voice assistant without an entity, a humanoid IVA without emotional expressions, and a humanoid IVA with emotional expressions. The results show that compared with the voice assistant, a humanoid embodiment can significantly improve the user's experience, and compared with humanoid IVA without emotional expressions, a humanoid IVA with emotional expressions is more welcome. Moreover, we use the cloud model to describe the uncertainty of IVAs' actions (blinking and body orientation). The results show that the uncertainty of actions can increase the believability of IVAs.