Paper Type
Complete
Paper Number
PACIS2025-1267
Description
Artificial intelligence (AI) is poised to transform augmented reality processing and natural voice interactions, reshaping human collaboration with AI agents. Unlike human-human digital collaborations, which foster mutual understanding, human-AI collaborations emphasize optimizing instruction clarity. Building on the instructional design literature, this study investigates two augmented reality features for progressive, multi-round human-AI diagnosis, verbalized instructions and visual illustrations, and their effectiveness and efficiency. Integrating dual-coding theory, this study examined the interaction and mediating roles of cognitive effort and confused emotions. Two laboratory experiments using augmented reality devices and neuroscience techniques revealed that dominance-toned verbalizations paired with schematic illustrations improved efficiency, whereas those with structural illustrations enhanced effectiveness. Cognitive effort and confused emotions jointly mediated performance effects, with confused emotions being especially relevant to positive attitudes toward AI.
Recommended Citation
Hong, Zeyuan (Stephen); Choi, Ben; and Boh, Waifong, "Demystifying Human Reliance on AI through a Dual-Coding Perspective: An Augmented Reality Experiment with Eye-tracking and Facial Expression Analysis" (2025). PACIS 2025 Proceedings. 3.
https://aisel.aisnet.org/pacis2025/emerg_tech/emerg_tech/3
Demystifying Human Reliance on AI through a Dual-Coding Perspective: An Augmented Reality Experiment with Eye-tracking and Facial Expression Analysis
Artificial intelligence (AI) is poised to transform augmented reality processing and natural voice interactions, reshaping human collaboration with AI agents. Unlike human-human digital collaborations, which foster mutual understanding, human-AI collaborations emphasize optimizing instruction clarity. Building on the instructional design literature, this study investigates two augmented reality features for progressive, multi-round human-AI diagnosis, verbalized instructions and visual illustrations, and their effectiveness and efficiency. Integrating dual-coding theory, this study examined the interaction and mediating roles of cognitive effort and confused emotions. Two laboratory experiments using augmented reality devices and neuroscience techniques revealed that dominance-toned verbalizations paired with schematic illustrations improved efficiency, whereas those with structural illustrations enhanced effectiveness. Cognitive effort and confused emotions jointly mediated performance effects, with confused emotions being especially relevant to positive attitudes toward AI.
Comments
Technologies