Joint action with artificial agents: Human-likeness in behaviour and morphology affects sensorimotor signaling and social inclusion

Abstract

Sensorimotor signaling is a key mechanism underlying coordination in humans. The increasing presence of artificial agents, including robots, in everyday contexts, will make joint action with them as common as a joint action with other humans. The present study investigates under which conditions sensorimotor signaling emerges when interacting with them. Human participants were asked to play a musical duet either with a humanoid robot or with an algorithm run on a computer. The artificial agent was programmed to commit errors. Those were either human-like (simulating a memory error) or machine-like (a repetitive loop of back-and-forth taps). At the end of the task, we tested the social inclusion toward the artificial partner by using a ball-tossing game. Our results showed that when interacting with the robot, participants showed lower variability in their performance when the error was human-like, relative to a mechanical failure. When the partner was an algorithm, the pattern was reversed. Social inclusion was affected by human-likeness only when the partner was a robot. Taken together, our findings showed that coordination with artificial agents, as well as social inclusion, are influenced by how human-like the agent appears, both in terms of morphological traits and in terms of behaviour.