Machine Trait Scales for Evaluating Mechanistic Mental Models of Robots and Computer-Based Machines

In previous work, we and others have used the Big Five Personality Scales and other human trait rating scales to evaluate anthropomorphism and social responses to machines (see our CHI short papers). However, we believe that most people correctly perceive that machines are inanimate and not " real " people. In other words, anthropomorphism is partial. Current research in cognitive science implies that when a machine, such as a robot, exhibits humanlike behavior (such as speech and ostensibly intentional movement), people retrieve cognitions associated with people and with machines from long-term memory. From these cognitions, people create a more or less coherent mental model of the machine they are observing. The purpose of this research was to create rating scales for robots (and other computer-based machines) that could be used to measure the extent to which people's mental model of the technology incorporates inanimate and mechanistic elements. We plan to use these scales in conjunction with measures used in studies of person perception, such as the Big Five personality scales and intellectual evaluation items. Whereas the latter measure animist and anthropomorphistic elements of a person's mental model of a machine, our new scales will measure the inanimate and mechanistic elements of the mental model. Method We first created a pool of 63 traits from several sources: Handbook of Human Factors Design (Woodson et al, 1992), a study of seven people who were asked to photograph and describe their favorite appliances, tools, and machines at home and at work, and through brainstorming in our research group. We then combined these traits with trait names from the Big Five Inventory (John et al, 1991) and intelligence evaluation items (Warner & Sugarman, 1986) to create two versions of a questionnaire. One version asked participants to rate the traits on 5-point scales in answer to the statement: " Please rate each of the attributes below and say