Children’s Beliefs About Robot Knowledgeability Vary With Their Ascriptions of Mind in Robots

• In Event: Poster Session 06
  In Poster Session: PS 06 Topic - Social Cognition

Abstract

To distinguish which agents know more than others can be a challenging task for young learners, as children may overwhelmingly rely on certain features of the agents (e.g., size, age, familiarity) to make judgements (e.g., Miller, 2000; VanderBorght & Jaswal, 2009). Children likely develop such strategies through their extensive interactions with human agents during their everyday activities. With the ubiquity of new technology, children regularly come in contact with intelligent technologies that appear to “know” things, such as humanoid robots. It is unknown, however, how children may perceive the knowledgeability of robots compared to human agents and what factors may contribute to the development of such perceptions. Here, we examined how children’s ascriptions of mental capabilities to robots may affect the extent to which they view robots as being knowledgeable vis-à-vis human agents of varying knowledgeability.

Fifty-two 4- to 6-year-old children (22 girls; Mage=5.36 years, range=4.01-6.23) participated in the study. In a Who-Knows-More task, children were presented with four agents (i.e., robot, teacher, child, baby) in pairs and asked to choose “which one knows more” for each pair. Children’s choices of the robot (versus other agents) were summed as indicative of their relative perception of knowledgeability of the robot in relation to other agents. To measure ascriptions of robots’ mental capabilities, children were asked to answer 5 yes-no questions assessing whether or not they attribute mental states to a robot (e.g., “Do you think a robot can think?”). As an exploratory investigation, children (n=43) also completed a Draw-A-Robot task examining their visual representation of robots. Children’s drawings were scored using a 10-point scale (adapted from Goodenough, 1926), with credits being given for humanlike features (e.g., body parts, clothing).

Children viewed robots as being knowledgeable to varying degrees (see Figure 1). In particular, 46.2% of the children (n=24) perceived the robot to be more knowledgeable than the other three human agents, even more knowledgeable than a teacher who is assumed to be the most knowledgeable of the human agents. Seventeen children (32.7%) viewed the robot as knowing more than a baby but less than a preschool-age child and a teacher. Six children (11.5%) reported that the robot knows more than a preschool-age child but not a teacher, while the remaining five (9.6%) believed all three human agents know more than the robot. An ordinal regression predicting perception of robot knowledgeability scores revealed a significant effect of mind ascriptions (OR=1.48, p=.016) and no significant effect of children’s age (OR=1.02, p=.97). Specifically, a tendency to ascribe mental capabilities to robots predicted a higher probability that children perceive robots to be more knowledgeable than the other human agents. Scores in the Draw-A-Robot task did not predict children’s judgments of robot knowledgeability. These results suggest that children’s attribution of a humanlike “mind” (rather than a humanlike “body”) to robots may be a key contributing factor to whether they would perceive robots as knowledgeable agents. We discuss the relevance of these findings to the power of mind ascriptions on children’s evaluation of technological agents.

Authors

• Xiaoqian Li, Singapore University of Technology and Design
  Presenting Author

• W. Quin Yow, Singapore University of Technology and Design
  Non-Presenting Author