Advice from Robots: Would You Choose a Robot that Looked More or Less Human?

Abstract

Humans are increasingly turning to non-human agents for advice. Therefore, it is important to investigate if human-likeness of a robot affects advice-seeking. In this experiment, participants chose robot advisors with different levels of human-likeness when completing either social or analytical tasks, and the task was either known or unknown when the robot advisor was selected. In the agent first condition, participants chose the advisor before receiving their task assignment, and in the task first condition participants received their task assignment before choosing the advisor. Results indicated that task type did not play a role in agent selection in either condition. However, in the agent first condition, more human-like robots (Nao and Kodomoroid) were selected at a higher rate than machine-like robots (Cozmo) and, in the task first condition, Nao was selected at a higher rate than Cozmo or Kodomoroid. These results should be considered when designing robots for giving advice to improve human-robot interaction.