The 1962 cartoon series The Jetsons featured a futuristic nuclear family: father George, mother Jane, and their offspring, Elroy and Judy. In the very first episode, we learn about the Jetson family's purchase of a housecleaning robot named Rosey. Rosey is, according to paleofuturist Matt Novak, "perhaps the most iconic futuristic character to ever grace the small screen." He explains that Rosey was high tech, but not perfect (perhaps the writers knew of the uncanny valley), and represented a certain vision of the future that involved robots designed to help humans in their day-to-day lives. No 1960s TV family is complete without a family dog, so in the fourth episode, Astro was introduced.
If Astro had been a real dog, and Rosey a real robot, they might have gotten along reasonably well. At least that's what is suggested by a new finding in the journal Animal Cognition.
Dogs are uniquely sensitive to human social communicative cues. Their transformation from wolf into dog did not just change their coloration and turn their tails curly and their ears floppy. By occupying a new social ecology within human culture, domestic dogs learned to interpret human communicative gestures in such a way that they even outperform chimpanzees when it comes to understanding cues like pointing or following the eye gaze of a human experimenter. Indeed, dogs are able to decipher the attentional state of a human and can capitalize on that information in order to steal unprotected food. Still, despite their impressive olfactory abilities - they're really good sniffers - dogs will choose the food indicated by a human, even if there's obviously another larger stash of food nearby.
As social animals, domestic dogs are also capable of eavesdropping. That is, they can gain important social information simply by observing an interaction between two others. For example, dogs rely on their gaze-following ability to understand whether a command issued by a human was directed to him- or herself or to another dog in the room. And when dogs watched two different experimenters play with another dog, they later preferred the human who was the "better" playmate.
Given their impressive suite of human-like social-cognitive abilities, researchers from the Comparative Ethology Research Group at Eotvos Lorand University in Hungary wondered how dogs would respond to a social robot. By using a robot, they reasoned, they could disentangle sociality from humanness. In other words they wanted to see whether dogs are specifically skilled at interacting socially with humans, or if they could engage socially with an individual who doesn't look at all like a human, but still acts like one.
They teamed up with a second group of researchers, from the Institute of Computer Engineering, Control, and Robotics at the Wroclaw University of Technology in Poland. They're the ones who provided the "People-Bot mobile platform," a robot that looked like a laptop computer on wheels with two arms. Each of the arms was capped with a four-fingered hand that could point and grasp objects.
Different groups of dogs participated in a widely used social experiment: given two buckets, either of which could contain food, would the dogs prefer to approach the bucket that the experimenter pointed at? Several studies have indicated that since dogs understand the pointing gesture, they approach the bowl to which an experimenter points. In this study, the researchers not only used human experimenters to point at food, but the robot as well. Each dog saw participated in the human experiment, and then in one version of the robot experiment: either with a "social robot," or an "asocial robot." The actual robot was the same but was programmed to act either socially or asocially.
To display evidence of its sociality, in the social condition a human and robot shook hands, talked with each other, and walked together around a room. Then, the robot called the dog's name, and dropped a piece of food in front of the dog. In the asocial condition, the human typed on the laptop that served as the robot's "head" for a few minutes. Then, he offered his hand to shake with the robot, but the robot didn't respond. The robot emitted a beep and dropped a piece of food in front of the dog.
Thus, there were two critical differences between the robots. First, the social robot responded contingently to the actions of the human, while the asocial robot, despite having the same behavioral repertoire, did not. Second, the social robot communicated with the human using speech while the asocial robot communicated using only non-linguistic beeps.
Then came the testing phase. If the dogs were able to distinguish between the two robots on the basis of their sociality, then they should follow the social robot's pointing gesture - not not the asocial robot's - to find food.
While the dogs were best at finding food when there was a human experimenter, they were still likely to find the hidden food when indicated by the social robot, while they performed worst when confronted with the asocial robot. It appears as if the dogs didn't perceive the actions of the asocial robot to be communicative. Despite displaying the some overt actions, the dogs behaved as if only the behaviors of the social robot contained useful information.
These findings were also borne out by the dogs' gaze. They gazed longest at the baited container when the pointing gesture was shown by a human experimenter, then next longest for the social robot, and least for the asocial robot.
The dogs' behavior show that while they did not treat the social robots exactly like human experimenters, they were more willing to treat the robots as social partners when they had observed that robot engaged in a prior social interaction. "Although the level of sociality shown by the robot was not enough to release the same set of social behaviours on the part of the dogs" as they show towards humans, the researchers write, "it had a positive effect on dog-robot interactions." When it comes to the design of robots, this suggests that their social abilities may matter more than whether or not they visually resemble humans.
Indeed, this study may tell us more about how to evaluate robotic design than about dogs' social-cognitive skills. Human experiment participants bring with them a set of assumptions about robots derived from the way they are portrayed in television and movies, from The Jetsons' Rosey the Robot to Star Trek's Data and Skynet. By using dogs, researchers can evaluate the social proficiency of their robots in a group of subjects who are unaffected by the trappings of human culture.
Lakatos, G. et al (2013). Sensing sociality in dogs: What may make an interactive robot social? Animal Cognition DOI 10.1007/s10071-013-0670-7
Photos by Eniko Kubinyi.
For more on dogs and domestication: