September 20, 2008
I Robot, I Human
Two interesting articles on robotics describe and study the issue of "how human-like should robots be?" We have anthropomorphized many objects but robots, since most are interactive, easily lend themselves to being endowed with human qualities and personalities. We have already seen that our floor-cleaning robots get named, as do the bomb-defusal robots that have been deployed in Afghanistan and Iraq. When they break down or get blown up, their users have even been known to mourn.
Robots are becoming more capable. Some now have visual systems that combine digital cameras with image-processing software to allow them to figure out what they see and make decisions based on that information. Recent robotic rovers sent to Mars are able to plot their own courses around rocks in order to reach a specified destination. Robotic manipulators are also becoming more dexterous—precise enough, indeed, with the use of new touch-feedback systems, to perform brain surgery.
The next step suggests that general-purpose service robots could soon be stepping out of development laboratories and into people’s homes. What will they look like? According to an article in the Economist, current robot developers fall into two camps. There are those trying to make their machines as anthropomorphic as possible, even though that tends to complicate things. Honda’s ASIMO, for example, walks, talks and looks like a child in a space suit. At the other extreme is Care-O-bot, developed by the Fraunhofer Institute for Manufacturing Engineering in Stuttgart which is functional and easy to make but not visually attractive.
How do we respond to robots?
A research group from the Clinic for Psychiatry and Psychotherapy at Aachen University, Germany, published a study in the Public Library of Science Journal(PLOS) in which they used functional magnetic resonance imaging to find out how people’s brains respond to various types of robots.
The researchers gave a group of volunteers notebook computers and asked them to play a simple “prisoner’s dilemma” game against four different opponents, who were also armed with laptops. The first opponent was a laptop on its own; the second a laptop in which the keys were pressed by a pair of robotic hands without a body; the third opponent was a humanoid robot (known as BARTHOC Jr and developed at Bielefeld University) and the fourth was a real human.
The participants were introduced to their opponents in a trial game, and then went into the fMRI scanner believing they were still connected to the same opponent via a pair of video goggles and a control box. They then played a series of games with each opponent as their brains were monitored. But they were not actually playing the opponents they had just seen. Instead, they were fed a random series of responses to their moves.
The “prisoner’s dilemma” game was chosen because it involves a difficult choice: whether to co-operate with the other player or betray him. Co-operation brings the best outcome, but trying to co-operate when the other player betrays you brings the worst. The tendency is for both sides to choose betrayal (thus obtaining an intermediate result) unless a high level of trust exists between them. The game thus requires each player to try to get into the mind of the other, in order to predict what he might do. This sort of thinking tends to increase activity in the medial prefrontal cortex and the right temporo-parietal junction.
The scanner showed that the more human-like the supposed opponent, the more such neural activity increased. A questionnaire also revealed that the volunteers enjoyed the games most when they played human-like opponents, whom they perceived to be more intelligent. The researchers believe that this demonstrates that the less human-like a robot is in its appearance, the less it will be treated as if it were human. That may mean it will be trusted less—and might therefore not sell as well as a humanoid design.
The authors of the study point out that "we expect the results of our study to have an impact on long-lasting psychological and philosophical debates regarding human-machine interactions. Moreover, the findings of the present study will affect the designing of robots that are utilized for direct human-robot interactions. As human-robot communication will play a key role in future, the present study may help to integrate theory and research from neuroscience and social robotics in order to place this work in a broader conceptual framework and promote synergy across fields."
Economist Technology Quarterly I, Human
Posted by rsk at September 20, 2008 10:16 PM