On January 26 we visited ATR (Advanced Telecomunication Research Institute International) a Japanese company of which some branches are specialised in the production of robots and whose headquarters are located close to Kyoto. Our interlocutor was Dr. Takayuki Kanda who gave us a a warm reception and detailed description of the research dedicated to in robots destined to social use. Within ATR this area of inquiry is the province of the ICR (Intelligent Robotics and Communications Laboratories) that conducts research on “communicative robots”. These artifacts are destined for use in the domain of services and while designing them much attention is given to the emotional dimension of social interaction. In relation with the question of interaction between human and artificial agents Dr. Kanda stressed the importance of “embodiment” in order to awake an emotional response in humans. That is to say, according to him, the physical structure of artificial agents, which plays a functional role similar to that of a human body, is fundamental to the emotional reaction produced by robots. This embodiment is the specific aspect that distinguishes robots from other types of artificial agents – i.e. virtual agents (see report 2). According to Dr. Kanda this corporal dimension plays a key role in emotional interactions between humans and machines. This thesis suggests that this line of research is, at least to some extent, in agreement with post-classical cognitive science, or embodied cognitive science,according to which access to others – the inter-subjective opening that permits emotional resonance – essentially depends on the body.
The embodied artificial agents of ATR that we encountered were of two different types. The first is ROBOVIE, which exists in many different versions, all of which are destined to be used as a tool of social information (publicity, education or in different service, i.e. as information interface in train stations) or as support for human agents (help to patients in hospitals or to residents of old age pensions, domestic help, or reception in public space, or even security). All versions have three essential aspects in common: (a) a small size, more or less that of a child; (b) a capacity of expression, especially through the movement of the eyes which are able to follow human gaze, and (c) a body made up of arms, a bust and moving parts which take the place of legs and that allows spatial coordination with humans being in physical space. Interaction with ROBOVIE is in no way threatening and generates a positive emotional reaction. ROBOVIE gives the impression of being able to recognize its interlocutors and to remember passed encounters, and is able to entertain simple conversations, similar to those of a child. However, as Dr. Kanda pointed out, all these abilities are to some extent a fiction, or at least, they are not what at first sight they seem to be. For example the capacity to recognize individuals with whom it has had past encounters is not really an ability which the robot has. What makes it possible is that, the first time a person interacts with ROBOVIE he or she is given a badge that sends out a radio signal. It is this signal that allows the robot to re-identify the person in the future. Rather than an intrinsic capacity of the robot it is a capacity of the system made up of the robot and of the (modified) social environment. Furthermore the capability of motor coordination with humans, as well as that that of having a pertinent conversation, that takes into account the history of past encounters, is not inherent to the robot but depends on a human agent who controls the artifacts response from a distance. The artificial agents of ATR – and many of those produced by contemporary sciences of the artificial -are “semi-autonomous”, that is to say they are in constant relation with a human operator who plays a fundamental role in the robot’s mobility and overall ability to interact. The most autonomous version of ROBOVIE actually available resembles a child robot. It has a sophisticated visual perceptive system, interestingly different from ours, that allows it to circulated, autonomously through the offices of ATR where it interacts with the persons it encounters using a limited repertoire of simple sentences.
The second type of artifact with which we interacted at the ATR laboratories was an android. Dr. Kanda, without any previous warning, introduced us into a room where was waiting, seated, an individual who had a rather dejected expression. This produced at first a somewhat intense emotional reaction, a mix of surprise and uncertainty. The size of the artifact (that of an average adult), as well as its human features, the details of expression and slight bodily movements, acted together to reproduce fundamental characteristics of a human subject, and it was not immediately clear that we were actually facing a machine rather than a person. The first impression was that of being in the presence of a human being who, for some difficult to defined reason, was not quite normal and a little threatening. It was not easy to identify the exact causes of this reaction. Perhaps is it because of the size of the machine and of its apparently autonomous character. However this artificial agent is also semi-autonomous and even unable to move around. It is attached to its chair that supplies it with the compressed air necessary for it to retain its human shape. The only mobility it actually possesses is that of its expressive facial and hand movements. Those are remarkably human like and are activated by an operator who controls the android from another room, playing the role of its sensor-motor center. This certainly restricts the possible uses of this type of androids in the near future and Dr. Kanda compared it to a kind of glorified “telephone” – a physical replacement of the person with whom one is taking at a distance. In fact this android reproduces the exact features of Dr. Hiroshi Ishiguro who is Visiting Group Leader of the Department of Communication Robots within ATR Intelligent Robotics and Communication Laboratories. This “double” allows him to ‘interact’ with members of ATR while being in his office at Osaka University.

Our visit to ATR laboratories suggests that, even in Japan where it is deemed to be most advanced, sciences of the artificial are still a long way from being able to produce autonomous agents. This visit allowed us to better perceive the possible social effects of social robots. It is clear that, at this point in the technological development of social robots, in order to give even partial autonomy to these artifacts, it is necessary to profoundly transform the environment within which they interacts with humans. It requires the transformation of the social environment into a technical system indispensable for the artifact to be able manifest, even the limited abilities which they do in fact exhibit. For example, it is necessary for human agents to have a radio emitting device that allows them to be recognized by artificial ‘agent’. The functioning of such robots also requires a simplified space, free of obstacles, in which the robot can move or recognized that it has been approached. This suggests that the use of such robots in hospitals, schools, or commercial centers, will require a significant modification of these environments in order to make them more “robot friendly”.