Human-like behavioral variability blurs the distinction between a human and a machine in a nonverbal Turing test

Area of Science:

• Human-Robot Interaction
• Cognitive Science
• Robotics

Background:

• Behavioral variability is a key feature of biological systems, including humans.
• Limited research explores how variability influences human perception of robots.

Purpose of the Study:

• To investigate if reaction time variability (standard deviation and distribution shape) affects humanness ascription to robots.
• To assess human sensitivity to robot behavioral variability in a joint action task.

Main Methods:

• Participants performed a joint Simon task with an iCub robot.
• The robot's reaction times were manipulated to fall within human ranges but with non-human distribution shapes.
• A nonverbal Turing test variant was used to evaluate humanness perception.

Main Results:

• Participants correctly identified the human agent above chance level.
• The iCub robot, with reaction times within the human range but non-human distribution, passed a nonverbal Turing test.
• The range of variability, not its distribution shape, appeared to influence humanness ascription.

Conclusions:

• The range of behavioral variability is a significant cue for observers when ascribing humanness to humanoid robots.
• Future research should explore other variability features in human-robot interaction.