A robot developed by Computer Science Ph.D. candidate Justin Hart GRD ’13 at the Social Robotics Lab may pass a landmark test by recognizing itself changing in a mirror.
Self-awareness, the ability to recognize oneself as distinct from one’s surroundings, is a mark of higher-level cognitive skills. This test was first developed to test the presence of self-awareness in animals, and requires the subject to recognize a change in its appearance by looking at its reflection.
In the mirror test, developed by Gordon Gallup in 1970, a mirror is placed in an animal’s enclosure, allowing the animal to acclimatize to it. At first, the animal will behave socially with the mirror, assuming its reflection to be another animal, but eventually most animals recognize the image to be their own reflections. After this, researchers remove the mirror, sedate the animal and place an ink dot on its frontal region, and then replace the mirror. If the animal inspects the ink dot on itself, it is said to have self-awareness, because it recognized the change in its physical appearance.
Only a few species of animals, including chimpanzees, bottlenose dolphins, magpies and elephants, have passed the test.
Ambarish Goswami, a principal scientist at Honda Research Institute in California, said that a robot could never be self-aware in the same way an animal can be. Instead, the kind of limited self-awareness for which the researchers plan to test is “purely an image-processing program.”
To adapt the traditional mirror test to a robot subject, computer science Ph.D. candidate Justin Hart said he would run a program that would have Nico, a robot that looks less like R2D2 and more like a jumble of wires with eyes and a smile, learn a three-dimensional model of its body and coloring. He would then change an aspect of the robot’s physical appearance and have Nico, by looking at a reflective surface, “identify where [his body] is different.”
Brian Scassellati, associate professor in the Computer Science Department and Hart’s Ph.D. advisor, said that Hart’s research postulates that robots can demonstrate some of the characteristics that qualify as self-awareness — in this case, recognizing the spatial relationship between mirrors and real life.
If Nico passes the self-awareness test, the technology could have important implications for the field of robotics, Scassellati said. For example, a robot that has a working model of itself can self-calibrate.
“In terms of building more robust, more stable systems, every robot, even from an assembly line, is slightly different,” he said. “So as they wear and tear they will change differently. If you can adapt to that you can build systems that last longer and are more complex in initial construction.”
A self-aware robot could also repair itself if damaged, or at least compensate for damage sustained.
“You could picture a car that has a popped tire, notices it has a popped tire, and self-adjusts steering to adjust for a popped tire,” Hart said.
Goswami said that this kind of technology exists to a certain degree in the form of sensors, like when the flat tire light turns on in a car, but that for a robot with a working model of itself, this type of self-adjusting technology is feasible in the near future.
Hart said that because the field of artificial intelligence is a fairly new one, this kind of development is still far away. He added that the robotics community would still benefit from the technology required to create self-aware robots.
“In current engineering practice, there’s a lack of updating,” Hart said. “We should be able to change models if a robot damages or degrades.”
With a self-aware robot, keeping robot models updated will be significantly easier.
Hart has used the robot Nico for purposes beyond self-awareness models. In early 2010, Hart ran a study in which Nico played rock-paper-scissors with participants, occasionally cheating. When Nico cheated, participants grew angry and blamed Nico.
“They [participants] would attribute agency to the robot,” he said. “It implied that the robot was thinking in the mind of the subjects.”
A self-aware robot, Hart said, is not a thinking or feeling robot — instead, it is a robot programmed to know itself.
Goswami gave the example of a robot programmed to come to a screeching halt once it reaches a staircase, or a similar drop in height. The robot stops moving, but because sensory input relayed to the program commands it to, not because the robot has a fear of falling. The humanoid robots of film are beyond the realm of current robotics, but Hart said that if Nico passes the test, he thinks the robotics community “can still get a lot of mileage out of the low-hanging fruits of robots which ponder about their bodies.”
Hart said he plans to conduct the test within the next few months.