Agent personification in trust development

To determine how trust is affected by an interface containing the anthropomorphic representation of an intelligent agent, 593 visitors to a popular automotive Web site were surveyed on their use of an online agent-like system. The type of Personification (plain text-only, text with an icon representing the agent, text with an icon and a biography about the agent) and Message Style (Structured and Anthropomorphic) were manipulated. A significant interaction was found for all three derived factors: Decision Support, F(2, 587) = 5.79, p ? .01; Implicit Trust, F(2, 587) = 6.08, p ? .01; and System Competence, F(2, 587) = 3.29, p ? .05. The lack of main effects indicates that simple manipulations may not have the large impact typically assumed in the literature. Interface Designers wishing to enhance the acceptance of systems that include intelligent agents should ensure that the anthropomorphic characteristics as a whole, including the communication style and visual representation, should form a strongly cohesive and consistent presentation to the user.