Could some of the interactions between a young child and his or her caregiver (such as a parent) be a useful model source for some of the interactions between a robot and a person? This paper reports on some experiments like that, done with a Sony AIBO robot that can learn and perform behaviors. For the experiments, the robot was located on a children’s play mat that was scattered with some children’s toys and some colorful objects.

The authors of this 24-page paper first provide ten pages of general introduction about their experimental objectives and approach, and a short discussion of selected background literature. They present their arousal-driven architecture and the algorithms the robot uses when selecting behaviors based on patterns it detects in the input data from its sensors. The paper includes the uniform resource locator (URL) for a video about the robot’s behaviors (http://www.youtube.com/watch?v=tndSnyUWqBI).

In the next ten pages, the authors describe their three groups of experiments with the robot. The first experiment group trained the robot to adapt to and carry out consistent patterns of behavior while operating in varied environments. The second experiment group had the robot take two different behaviors in each ten-minute experiment with a laboratory staff person. The main reported finding from this second group was that “caring” responses by the person toward the robot contributed to faster learning by the robot. In the third experiment group, instead of using laboratory personnel in the role of potential caregiver, the experiment used a sequence of random volunteer visitors at the London Science Museum. The volunteers who had experienced the robot’s “needy” behavior rated interacting with the robot as more enjoyable than the volunteers who had experienced the robot’s “independent” behavior.

In the three-page conclusions section of this paper, the authors point to the professional communities that are most likely to be interested in their findings: emotional and conceptual development, human-robot interaction, developmental robotics, robot design, and developmental psychology. The paper’s two tables provide the best terse summary of this lengthy human-robot interaction paper.