| Summary: | 碩士 === 國立清華大學 === 資訊系統與應用研究所 === 102 === Previous studies have depth explorations of agents’ realism. Most of researches look into external characteristics like appearance, emotion expression or body gesture and some discussions about collocation of verbal and nonverbal behavior but lack of research on characters’ traits. Character traits are also important elements of agents which may help tasks and improve relation between human and computer. Our study tries to evaluate if agents with character traits can influence users’ cognition and perception.
This study was a 2 x 2 between-subjects design. The independent variables were task complexity (simple vs. complexity) and character behavior of agents (realism vs. unrealism). The dependent variables were participants’ cognitive load, performance and perception. Participants should try to read and find key information of tasks and all information was provided from agents.
The results found participants who faced complex tasks ignored the character behavior of agents, so they could concentrate on their tasks. It helps participants’ task performance and cognitive load. There are no differences of task performance and cognitive load between two complexity groups. But participants faced complex tasks showed that they like realistic agent more than unrealistic agent even if character behavior would increase their cognitive load.
Participants who faced simple tasks could find the character behavior of agents and they enjoyed alone with agent which showed character behavior. They got close to agent and hope to interact again. However, some participants didn’t like extreme behavior of agent.
But participants who faced simple task and unrealistic agent didn’t like agent significantly. They complained that agent without behavior was cold and it was too hard to get close. They believed that agent with realistic behavior can improve relationship between them and task performance.
Character traits of agent may improve human-computer interaction, but also increase users’ cognitive load. This study suggests that application of agent should think out task complexity and users’ preferences. Designers should try to balance quantity of task information and behavior from agents.
|
|---|
