EXPLORING COLLABORATION WITH AN INTELLIGENT VIRTUAL AGENT

Building on advanced technologies, including artificial intelligence and virtual reality, researchers have utilized intelligent virtual agents to simulate humans in virtual environments. Although researchers have focused on various attributes of virtual agents and their impacts on human perception and user experience during interaction with humans, the understanding of how the intelligence of the virtual agent influences human perception and user experience has remained underexplored. Also, while researchers have employed collaborative tasks to design interactions between virtual agents and humans, prior studies examining their impact on human-agent interaction have been limited. Therefore, I employed jigsaw puzzle co-solving activities to design collaborative tasks and implemented a virtual reality application where participants can solve a puzzle with an intelligent virtual agent. Based on this application, I conducted three studies exploring how the design of virtual agents and collaborative tasks affects human perception, user experience, and human behavior metrics.

In the first study, I focused on the intelligence levels of virtual agents and how these levels can influence human perceptions of virtual agents and user experiences. Through a within-subjects study, I found that the intelligence levels of virtual agents affected our participants' perceptions of the virtual agent, their user experiences, and their behavior metrics during collaborative tasks.

In the second study, I expanded the first study by employing self-correction behaviors. To do so, I conducted a study with five experimental conditions encompassing the intelligence level and self-correction accuracy. The experimental results indicated that self-correction behaviors could affect our participants' perceptions of the virtual agent and improve user experiences.

Last, I introduced task complexity driven by the number of jigsaw puzzle pieces and their size. Specifically, I designed a 2 × 2 study comprising task complexity and the intelligence level of the virtual agent. I found significant results on our participants' ratings of the perceptions of the virtual agent and user experiences, not only by the intelligence level of the virtual agent but also by task complexity.