Using Robot Adaptivity to Support Learning in Child-Robot Interaction

Previous research has shown that if a robot invests physical effort in teaching human partners a new skill, the teaching will be more effective and the partners will reciprocate by investing more effort and patience when their turn to teach comes. In the current study, we extend this research to child-robot interaction. To this end, we devised a scenario in which a humanoid robot (iCub) and a child participant alternated in teaching each other new skills. In the robot teaching phase iCub taught participants sequences of movements, which they had to memorize and repeat. The robot then repeated the demonstration a second time: in the high effort (or Adaptive) condition, the iCub slowed down its movements when repeating the demonstration whereas in the low effort (or Unadaptive) condition he sped the movements up. In the participant teaching phase, children were asked to give the robot a demonstration of three symbols, and then to repeat it if the robot had not understood. The results reveal that children learned the sequences more effectively when the iCub adapted its movements to the learner, and that, when their turn to teach to the robot came, they slowed down and increased segmentation when repeating the demonstration.