As the boundary between science/technology and society has gradually become permeable, and as science and technology have become subject to societal interrogation, more and more interdisciplinary collaboration has emerged as a way to not only encourage innovation but also engage the public. Under the guise of interdisciplinarity, however, heterogeneous modes of collaboration exist, in which the roles of the public are differently conceptualized. Using Andrew Barry and Georgina Born's 2013 framework for analyzing interdisciplinarity, including that of science and art, this article examines two instances of the interdisciplinary collaboration between robotics and theater in Japan and Taiwan. In particular, it analyzes how the perceived agency of robot actors is constructed onstage in these two cases. On the technological level, anthropomorphic, humanoid robots are used to construct the agency of robot actors in both cases. On the theatrical level, a comparison shows divergences in the role that theater plays and the potential role of the public. In contrast to the Taiwan case, for the Japan case a more active role of the public is conceptualized involving their use of social knowledge in judging the constructed agency of robot actors and their contribution to knowledge of human-robot interaction. Theater is used in both cases to demonstrate technology for the public. However, a comparison of the two cases demonstrates that, beyond that purpose, in which the public assumes a passive role, the interdisciplinary collaboration between two seemingly distant disciplines—robotics and theater, in this case—has the potential to foster knowledge production in which the public can assume an active role.
The symbolic richness of robots has been exploited by artists for decades. Because of their awkward or exaggerated gaits and movements, a staged performance of mechanical robots often hints at “the humanization of machines and the dehumanization of humans” (Dixon 2004: 15). The ever-present gap between the bodily movements of humans and robots, and the artificiality of the latter, is used by artists to express their views toward robotics in particular and technology in general. In this line of performance, the field of robotics is both the embodiment of technological progress and the object of artistic critique.
The audience's viewpoints toward robots are also influenced, probably more than by the stage performance itself, by movies and science fiction. The enormous influence of these popular works on the audience's perceptions of robots might also have negative effects, in a way different from artistic critique, on robotics as a field. As Japanese science fiction writer and robot culture researcher Hideaki Sena argues, the overly powerful robots depicted in popular works have resulted in audiences entertaining unrealistic expectations for robots. He specifically points out the impact of Astro Boy, arguably the most influential robot figure in postwar Japan. Although the cartoons of Astro Boy have encouraged some youngsters to become roboticists, it nevertheless creates a gap between audiences' expectations and real-world robotics, a gap Sena calls “Astro Boy's spell” (2003: 11). He urges that developers work to transcend this spell by encouraging exchanges between the science fiction about robots and current robotics and by creating new genres of robots (12).
Performing arts such as theater, which can incorporate elements of popular culture, have become a potential genre for robotics. As the trend of developing humanoid and social robots has flourished in recent years,1 the relationship between the performing arts and technology—robotics in this case—assumes more varied forms. Art does not necessarily, as it often did, assume the role of technology critic; instead, current robotics has begun interacting with art. Recently, international conferences on robotics and human-robot interaction have begun to hold workshops on art and robotics. For example, the call for papers for the International Conference on Robotics and Automation 2012 Workshop, titled Robotics and Performing Arts: Reciprocal Influences, concurs with Sena's advice, stating its general scope as follows: “The goal…is to celebrate the creative interaction between robotics technologies and art, and stimulate discussions on this topic. This full-day workshop will investigate the interactions between robotics engineering and performing arts, focusing on their reciprocal influences and how such influences can benefit both.” The overview of the workshop describes it as
highly interdisciplinary, bringing together prominent roboticists active in the artworld, and renowned scholars and artists using robotic technologies, including companies operating in the entertainment fields, in order to exchange ideas, practices and experiences on the key scientific, technological, artistic and commercial aspects emerging from the combination of art and robotics. Among the topics that will be explored during the workshop is how the interaction between robotics and performing arts can lead to the generation of new artforms, impact innovation and creativity in both the engineering and artistic field, and provide new possibilities for the emergence of new markets.2
The provision of new materials, by robotics research, for the performing arts to generate new art forms can be imagined without too much difficulty. However, what do the performing arts offer current robotics? Some possibilities emerge regarding this question. First, the performing arts provide roboticists with access to a public to whom they would otherwise have no access. Robot theater brings together those who are interested either in robots or in theater and provides a chance for the roboticists to address both—thus new markets possibly emerge. Through such interaction, the public may also come to be more than an immediate economic consideration. For roboticists working on robots aimed to be used in everyday life, the public may be seen as an invaluable resource to be mobilized, in the setting of performing arts, to test people's responses to robots.
This opportunity for testing is especially important for social robots, for they are designed to smoothly interact with people. Roboticists have been trying to figure out the right codes for the conduct of social robots, so that people will be more comfortable interacting with them. This search justifies the interdisciplinary collaboration with theater, because roboticists claim to find parallels between actors and robots. When put into specific situations, both actors and robots have to correctly play their roles. In other words, robots in human-robot interaction and actors in theater both aim “to get as close as possible to the unobtainable ideal (normal social player/actual character) by giving the correct outputs.…Robot/actor must appear to be something it is not, in order to provoke the desired response” (Lu and Smart 2011: 475). Because actors are thought to use their bodies to convey messages more clearly and efficiently than others, actors' behavior when acting can be used as a model for tuning the movements of robots in controlled situations onstage (Wright 2012: 13). In this way, theater can provide useful knowledge for human-robot interaction. An environment less constrained than a laboratory setting and more controlled than daily situations, theater thus serves as a test bed for tuning the behavior of robots by using audience as judges. In theater, robots, just like actors, aim to “pass” as their characters onstage. The key for the robots' ability to reach that aim is to make the intentions of the robots felt, as if they were autonomous agents. As argued by two roboticists, if the audience perceives the intentions of a staged robot, “then we can be sure that the robot is at least conveying some of the right social cues, which can then be transferred to more traditional human-robot interaction” (Lu and Smart 2011: 477).
Understood in this way, the interdisciplinary collaboration between robotics and theater seems to epitomize contextualization, a way in which some current knowledge is produced, which emphasizes the interaction between knowledge production and the public (Nowotny, Scott, and Gibbons 2001: 256). If the audience is given the role, in robot theater, of judging the “right social cues” for robots to convey, which transcend and can be transferred to “more traditional human-robot interaction,” social knowledge of the day-to-day interactions audience members have in their outside lives may help produce knowledge for human-robot interaction.3 However, as Nowotny, Scott, and Gibbons argue, although “contextualization means people,” these people can be legitimately conceptualized “either as passive objects or as active agents.…Giving a place to people in our knowledge refers to the ways that they are being conceptualized, which may—or may not—be supported by various forms of interaction, communication and participatory engagement with ‘real people’” (256–57). These different possible conceptualizations of the audience alert us to the variety of practices conducted under the banner of the interdisciplinary collaboration of robotics and theater. To what extent do roboticists open their restricted laboratories to artists and give a place to people? This question links the collaboration between robotics and theater to the relationship at issue between science and the public. How is the public addressed? Does the public really count in knowledge production? Are either robotics or theater, or both, transformed through their collaboration? These questions are important when considering the relationship between the public and knowledge production in the interdisciplinary practices of science, technology, and other fields, from an STS perspective.
Andrew Barry and Georgina Born (2013) provide a framework useful for categorizing interdisciplinary practices. They identify three modes of interdisciplinarity that feature different interrelationships among component disciplines. One, the integrative-synthesis mode, proceeds by combining two or more different disciplines in a relatively symmetrical way. Subordination-service mode, on the other hand, features interdisciplinary practices in which one of the component disciplines takes a leading and one a subordinate role. The third, the agonistic-antagonistic mode, is characterized by “an agonistic or antagonistic relation to existing or prior forms of disciplinary knowledge and practice.…[It] commonly stems from a commitment or desire to contest or transcend the given epistemological and/or ontological assumptions of specific historical disciplines” (12). In addition to outlining the interrelationships among component disciplines, they also identify three “logics,” that is, rationales, behind interdisciplinarity. The first is the logic of accountability, which aims to bring science closer to the public. The logic of innovation, on the other hand, attends to arguments about the importance of research in providing innovative and economic outcomes. In contrast, the logic of ontology refers to interdisciplinary practices with the potential to change the very nature of their constituents, to foster “ontological transformation in both the object(s) of research and relations between the subjects and objects of research” (Born and Barry 2013: 249).
Specifically, Barry and Born explain these logics by focusing on the relationship between research and the public in the example of art-science, the interdisciplinary, collaborative projects combining science and art. Art-science in the United Kingdom usually abides by the logic of the public understanding of science, as conceived traditionally (Yearley 2005: 116–17). In this way, art “is understood as a means by which the (absent) public for science can be assembled or interpellated. Science is conceived as finished or complete, and as needing only to be communicated, understood or applied, while art provides the means through which the public is mobilized and stimulated in the quest for greater understanding of science” (Born and Barry 2013: 249). Another conception of art-science, the former projects and pedagogies of the Graduate Program in Arts, Computation and Engineering at the University of California, Irvine, was described as “evidenc[ing] a distinctive form of publicness in relation to art-science, one in which science is understood not as self-sufficient or complete but as transformed and enhanced through its engagement with art, just as art is transformed and enhanced through engagement with science” (Born and Barry 2013: 249–50).
As Barry and Born note, the three logics of interdisciplinarity, in this case relating to art-science, need not be exclusive. Projects featuring the participation of the public can be driven by the logics of both accountability and ontology (2013: 20). However, the three logics feature differing weight and importance for interdisciplinary practices and can therefore be examined distinctly (Born and Barry 2013: 249). The framework they offer is pertinent to this article: how the audience is connected to robot theater, and how the interrelationships between the component disciplines—robotics and theater—can be characterized by the modes and logics of interdisciplinarity.
Toward the end of 2008, two performances of robot theater were held in auditoriums on university campuses in Japan and Taiwan. Both tried to make the theatrical robots pass as autonomous agents in the eyes of the audience, but the modes and logics of interdisciplinarity underlying these performances differed. This article first examines how these robots were made to pass as autonomous agents—how well they achieved this depends not merely on knowledge produced in the laboratory—and then compares, in the two cases, how the public was conceptualized and its role in the production of knowledge for human-robot interaction. The comparison of the two cases demonstrates that the public can assume a positive role in knowledge production and that the interdisciplinary collaboration between robotics and theater has the potential to foster knowledge production and to arouse reflections on the potential problems that robotics might bring about, at the same time.
Materials that form the basis of this article include the publications of and semistructured interviews with relevant roboticists and theater professionals. I also collected the play scripts and watched recorded videos of the robot theater performances whenever they were available, and in the case of the Japanese performances, I personally attended two plays of robot theater. As a comparative study of a very limited number of cases, this article is presented with caveats. There are various forms of robot performance in Japan, so the case chosen cannot claim to be representative of all robot performances in Japan. On the other hand, the case in Taiwan, up to the present time, finds no equal and was performed only once. In fact, the details of its showing may well discourage other similar attempts. However, it is hoped that this comparative study will contribute evidence of how interdisciplinary collaboration can be productively conducted in different ways both in this emerging field and in a more general sense. Before going into the actual theatrical performances in the two countries, I first describe how more traditional human-robot interaction is conceptualized as a way to make robots pass as autonomous agents, so as to suggest the potential specificity of human-robot interaction in robot theater.
Animism, Anthropomorphism, and Human-Robot Interaction
To make people perceive the agency of robots, certain human inclinations can be utilized. Researchers have found that complicity arises between humans and their encounters with certain inanimate objects, including robots: people describe certain robots as if they have minds, and the interactions between humans and robots as those between two humans (Turkle 2010). Behind the construction of this complicity lies a more general tendency of anthropomorphism—“the ascription of human characteristics to nonhuman entities”—which makes people treat machines or computers as if they were human (Caporael 1986: 215; Turkle 2010: 325). A different but related concept is in effect when we see robots. Besides anthropomorphism, animism—attributing life to something lifeless—is usually also at work. Certainly there are moments when the two can be distinguished. For example, “we animate but not anthropomorphize…if we say, an automobile purrs like a kitten, and anthropomorphize but do not animate if we speak to our pet turtle. If we speak to the automobile, however, we both animate and anthropomorphize” (Guthrie 1993: 39–40). Nevertheless, the two often happen together, as “animism feeds into anthropomorphism: when we treat something as alive, we treat it as alive like us, with our full catalog of mental states” (Hutson 2012: 165).4 This human tendency has profound implications. As Linnda R. Caporael rhetorically asks, “If anthropomorphism is an inherent bias, how could we know if machines think? How could we distinguish its thought from our perhaps unwarranted ascriptions of thought?” (1986: 228).
The tendency of people to blur the boundary between the animate/inanimate and human/nonhuman is greatly used by roboticists. One of the reasons that humanoid robots are built is because they stimulate anthropomorphism, which in turn makes people more willing to engage with them. But how human-like must a humanoid robot be for this advantage to take effect? Mashiro Mori, a Japanese roboticist, argues for a limit on the human likeness of robots. With his experience of seeing things that look like humans or human parts, such as toy robots, Bunraku (Japanese puppet theater) puppets, and prosthetic hands, he surmises that as the appearance of things becomes more human, people's emotional attachment to them increases as well. On the other hand, movement is no less important. We may show more affinity toward a moving than a still robot arm. However, a mismatch between a thing's human-like appearance and its not-so-human-like movement creates revulsion instead. Mori names the phenomenon uncanny valley: the area where the mismatch is too great for people to identify with human-like things, including robots (2012 ). Therefore, although movement sometimes helps to foster people's attachment to robots, it can have a significant negative effect if not correctly enacted: a zombie is more eerie than a corpse. Many roboticists take the uncanny valley as Mori's warning not to build robots with a too human-like appearance. A certain degree of human likeness of the body and facial features and usually an unthreatening height, comparable to that of a human child, suffice to make people feel an affinity for many humanoid robots.
Japanese roboticist Hiroshi Ishiguro goes beyond a model that would suffice, instead taking literally Mori's advice that “we should begin to build an accurate map of the uncanny valley so that through robotics research we can begin to understand what makes us human” (2012 : 100). Beginning in 2001, Ishiguro built a series of robots with an extremely human look (he calls such a robot an android) to investigate the effect on people of the matches and mismatches between appearance and behavior in such robots (Ishiguro 2009: 38). Ishiguro's first android, Repliee R1, took his own daughter—then four years old—as a model. Because Repliee R1 is small in size and cannot accommodate many actuators, which makes its movements limited, Repliee R1 cannot help but invoke the feeling of the uncanny.
Ishiguro then started to develop androids of adult size. In considering whether to build an android in female or male guise, Ishiguro decided that a female one would be preferable since it accommodates a wider range of test subjects—children are more hesitant to talk to adult men whom they do not know than to adult women (Ishiguro 2007: 226). The first full-size android, Repliee Q1, had a face created according to a computed combination of numerous Japanese female faces—an “average face.” Repliee Q1 looks relatively ambiguous regarding gender, so it may also look like a male when its dress and hair style change accordingly. According to Ishiguro, Repliee Q1 was seldom used in experiments, not because it looks both female and male but because its face was built from a two-dimensional average face that was not realistic enough (Ishiguro 2007: 228–9). A new version, Repliee Q1 Expo, was then created that took as a model a newscaster from the Japanese NHK broadcasting company and was shown in the Aichi Expo of 2005. After the expo, its eye contours and hairstyle were slightly changed to form Repliee Q2.5
Later on, Ishiguro built a new group of androids whose utterances are delivered in real time by operators located a certain distance away. He calls them geminoids, and they serve as three-dimensional avatars. Ishiguro built the first geminoid, Geminoid HI-1, in 2006, using himself as model. In 2010, another geminoid, Geminoid F, which was modeled on a female adult, was created. Only the upper parts of these geminoids can move. They cannot walk and are usually in a sitting position when on display. The facial expressions, utterances, and substantial bodily movements of geminoids can be controlled either by an operator or by computer programs. For them to imitate natural human movements, they have been programmed to display the unconscious movements, usually idiosyncratic from person to person, of a human's upper body. Various experiments have been conducted on androids and geminoids to investigate how people respond to them, and results are used to tune the movements of the robots so that they look more realistic. Results of these experiments also serve to chart the detailed contours of the uncanny valley.
For actual human-robot interaction, in addition to the appearances and unconscious movements of robots, how they respond to people greatly influences whether people anthropomorphize robots. Lucy Suchman investigates how particular cultural imaginaries—the ways human are imagined as a model of robots—inform how current roboticists build robots that interact with humans. She lists three important elements that roboticists prioritize: embodiment, emotion, and sociability; and she questions the actual materialization of these ideas (Suchman 2007: 229).
The basic tenet of embodiment is that humans learn, and thus intelligence is developed, by the interactions between their bodies and the environment. Therefore, an infant usually serves as a typical model of the construction of embodied robots. Provided with cameras, sensors, and actuators, humanoid robots may be programmed with cognitive abilities on a par with infants and may learn from interactions with people who serve as their “caregivers.” For instance, Cog, a humanoid robot developed at MIT in the early 1990s, can identify human faces and directs its eyes to moving objects. Moreover, Cog learns to manipulate the motion of objects with its arms (Keller 2007: 335). As for emotion, this is the purview of Cog's successor, Kismet, a facial robot. It has artificial eyeballs, eyelids, eyebrows, lips, and ears that make it look more human than Cog does (at least facially, since Kismet has no body, only a head and face). Kismet has several different facial expressions that can show its “emotions” and is installed with “motivations” for “social” interaction. It solicits human caretakers' responses to satisfy its motivations. Both Kismet and its caregiver are expected to learn from their interactions to elicit the “right” messages so that Kismet is satisfied, just like infants and their caregivers do (Keller 2007: 335; Suchman 2007: 237).
This embodied approach appears to promise to bring back the interactional context that was left out of the old, decontextualized approach in artificial intelligence (the so-called GOFAI, symbolic-processing-oriented “good old-fashioned artificial intelligence”). Critics, however, argue that in this new approach, the myriad interrelationships between the body and the world that are implicated in the rich meanings of the concept “embodiment” in social and cultural theories are reduced to simplified relationships between a preexisting world and a body. A dominant world that gives out stimuli and a body that receives these stimuli and responds—couched in manageable engineering terms—characterize this simplified relationship (Suchman 2007: 230–32). This can be seen from the brief description above of how Kismet works. A similar relationship can be found in roboticists' application of a cognitive model to the “social” interaction between people. In this model, the ability to infer people's intentions by observing their behavior is seen as critical to successful interactions. Clearly, such a conception of the social hinges on viewing “social interaction as the ability to predict the behavior of others and change one's own behavior in relation to these predictions,” which in turn is based on an individualized cost-benefit analysis of self-interests (Weber 2008: 861–62).
The cognitive model commonly used is the theory of mind (ToM) developed from investigating the “mind-reading” ability of chimpanzees in the late 1970s. It was then applied to human-human interaction by psychologist Simon Baron-Cohen in the study of autism, investigating the premise that a child who does not give attention to the object at which her caregiver looks (to share attention with the caregiver) may develop autism. The ability to share attention, scholars argue, is crucial for a child to develop a “theory of mind”—to recognize other people as “subjects with other minds” (Keller 2007: 335). Despite scholarly debates on ToM,6 just as in robotics the concept of embodiment is usually simplified as a relationship between a preexisting world and a body, the mechanisms critical to the development of the ability to read others' minds that advocates of ToM propose, such as intentionality detector, eye direction detector, shared attention mechanism, and theory of mind mechanism suggested by Baron-Cohen (1995: 32), seem—with their engineering tinge—readily applicable to be a working model for robot-human interaction. For example, ToM-inspired computer programs have been tried on Cog and a successor of Cog and Kismet, Leonardo, a robot with the appearance of a stuffed animal. Many robots with ToM-inspired computer programs, built in the United States, Europe, Japan, and elsewhere, are subjected to the so-called Sally-Anne test, or false-belief test—a psychological experiment to test whether children recognize that other people may hold different beliefs than their own. As can be seen easily, the interactions between these robots and people are usually modeled on the interactions between infants and their caregivers (Suchman 2007: 236; Weber 2008: 858).
As mentioned above, the cultural-specific image of the individualized, developmental model of humanness that materializes in robots is at work to facilitate human-robot interaction and to help robots pass as autonomous agents. The prevalent model of human-robot interaction therefore depends on an impoverished imitation of the human “social” world, which emphasizes cues, such as gazes and gestures, and certain relationships, such as a caregiver-infant relationship. Theater, on the other hand, although still limited, provides a much richer model of social interactions through its scenarios and settings. In theater, robot actors can be embedded in many-to-many social relationships, which not only move beyond the one-to-one relationship typical of human-robot interaction but also are closer to the interactions we experience in everyday life. As mentioned in the introduction, the audience can judge whether the playing of robot actors lives up to what their roles require of them in certain social relationships constructed onstage. Of course, whether the audience has a chance to make use of their social knowledge of everyday interactions—usually tacitly possessed—in judging the playing of robot actors depends on the degree to which the plots are socially rich, which is an empirical question. In addition to the one-to-one, individualized model of interaction that usually materializes in social robots, whether the theater provides any social meaning more than what is usually perceived in human-robot interaction is the question to be asked about the actual performances of robot theater.
Case 1: Japan
On 25 November 2008, the robot theater performance I, Worker premiered in an auditorium at Osaka University. It was directed by Japanese playwright and Seinendan Theater Company director Oriza Hirata and was the first in a series of robot theater plays that feature both robot and human actors (there have been six performances as of 2014). The plays are the result of a collaboration between Hirata and roboticist Hiroshi Ishiguro. Having been informed that robotics research at Osaka University had flourished after he was appointed as a member of the teaching staff at its Center for the Study of Communication-Design, Hirata proposed the idea of robot theater to the dean of the university and was introduced to Ishiguro, who is also faculty there.
As a playwright and director, Hirata has become famous in Japan for the compelling but also controversial ideas about theater that he has promoted since the 1990s. Traditional acting emphasizes the immersion of actors into their characters so that they express their characters' emotions onstage. This is achieved by training actors to put themselves in their character's place so that when an actor expresses the emotions of the character, these emotions are genuinely experienced by the actor, as if the actor and the character become one onstage. To the contrary, in his “contemporary colloquial theater theory,” Hirata deemphasizes actors' emotional performance. He does not think much of actors' interpretations. In his view, modern drama aims to depict the state of mind and emotion underlying certain behaviors exhibited by the characters and thus stresses motivation and mentality and deems actors' interpretation important for expressing such an emphasis. In contrast, Hirata's aim is not to depict motivation and mentality by way of actors' interpretation (Hirata 1995: 35) but to let the audience imagine what happens on the stage happening in real life by plainly presenting daily conversation. Hirata is therefore known for such famous yet notorious mottos as “Actors do not need to have minds” and “Actors need only act like robots” (Sasaki 2011: 202).
Hirata has provided two reasons that he was attracted to robot theater. One is its innovative nature: “When put into an environment in which one can realize what has not been realized, an artist will definitely try.”7 The other, however, is more of an afterthought. The idea of a robot actor aligns with his notorious mottos: if the audience is moved by theater performed by robots, which definitely do not have emotions, this will testify to the applicability of Hirata's ideas about theater. According to Hirata, the audience wept at the final scene of I, Worker, performed only by two robot actors, Takeo (Robot A), presumably male, and Momoko (Robot B), presumably female (Sasaki 2011: 202):8
Robot A: Ikue was crying a little.
Robot B: Oh?
Robot A: The sunset…tomorrow the weather will be fair as well.
Robot B: Yes.
Robot B: (Moves to another position, extending its arms.)
Robot B: Did you know? You should never tell a human to buck up when they are depressed.
Robot A: (Moves only its head to look back.) Yes, I had heard that.
Robot B: Humans are difficult.
Robot A:…Because we do not know what it is like to buck up.
Robot B: Yes.
Robot A: (Moves to another position.)
Robot B: When you were watching the sunset…
Robot A: (Looks at Robot B.) Hm?
Robot B: It was pretty, wasn't it?
Robot A: (Lowers its head.)…I suppose, but…
Robot B: (Glances sidelong across the audience.) What?
Robot A: (Looks toward stage right.)…
Robot B: But, what?
Robot A: It looks pretty because you're watching it with someone else.
Robot B: Is that so.
Robot A: So that even when you're watching it alone, you remember when you watched it with them, and it looks just as beautiful.
Robot B: (Looks at Robot A.) You think so?
Robot A: I can't think of it any other way.
Robot A: I have no other way of understanding what is beautiful.
Robot B: No?
Robot A: But we are not yet that advanced.
Robot B: (Slowly turns to look stage right.)
Robot A: (Gazes in the same direction.)
(The lights are gradually dimmed.) (Hirata 2010)
I, Worker sets its scene in a near future when humans and robots live together, against the backdrop of the problems of NEET (not in employment, education, or training) and hikikomori (social withdrawal or shut-in) among young people that plague some sections of contemporary Japanese society. The story is about a young couple and their two service robots, Takeo and Momoko. While the wife, Ikue, is supposed to be a homemaker, the husband, Yuuji, is currently unemployed and has no desire to find a job. At the same time, while Momoko prepares food for the family and is seen as indispensable for the couple, Takeo, oddly enough, is losing his motivation to work, just like Yuuji. The theme of the play, as the conversation between the four main characters unfolds, is an exploration of what it means for humans/robots to work. The robot failing to do work—which defies the definition of the word robot—brings the theme of the play into bold relief.
The two robots used in I, Worker were Wakamaru, humanoid, “robot-like” robots developed by Mitsubishi Heavy Industries based on robots built by Ishiguro. A Wakamaru moves on wheels, is one meter tall, and has a chrome yellow body and hands and silver arms.
After each robot theater performance, questionnaires designed by Ishiguro's laboratory were distributed among the audience. They include such questions as “Do you agree that the robot in the play is similar to a human?” and “Do you feel that what the robot says reflects its thought and feelings?” When the play I, Worker was performed on 25 November 2008, 70 percent of the audience reported either feeling or moderately feeling empathy for the robots (Kuroki 2010: 65). One of the reasons that these robots solicit positive responses from the audience, according to linguist Satoshi Kinsui, lies in the locution robots in these plays use. Kinsui specifically points out interjections, such as the word well, that were given by these robot actors. These interjections are highly uncharacteristic of the monotonous, mechanical expressions that people often think of as the speech of robots, or, in the words of Kinsui, they are the “least robot-like” elements of language (2010: 66). As these vague expressions reflect an exceptional ability of interpretation in reading the situation, it is not unreasonable for people to think that these utterances can be given only by humans, or those comparable to humans. Although the lines of the robot actors are preprogrammed, by showing a command of the nuances of Japanese language these actor robots seem, in the eyes of the audience, to have thoughts and feelings of their own.
Case 2: Taiwan
On 27 December 2008, a robot theater performance was staged in the auditorium of the National Taiwan University of Science and Technology (NTUST). It was the first robot theater performance in Taiwan and was initiated by roboticist Chyi-Yeu Lin (Jerry Lin) of the Department of Mechanical Engineering at NTUST. The performance primarily featured four humanoids, two wheeled and two bipedal.
The idea to stage robot theater in this instance came from Lin's experience of seeing a music show performed by robots in the Aichi Expo of 2005. There, the Toyota Company presented several partner robots, either wheeled or bipedal, that played trumpet, trombone, horn, violin, and drums. What Lin had in mind was robots, with facial expressions that the Toyota partner robots had lacked, that performed.9
Considering that all humanoid robots with a size comparable to humans developed before that point in Taiwan were wheeled, the addition of two bipedal ones was especially ambitious. Each of them is 160 cm tall and has an artificially skinned face, one (named Thomas) modeled on Lin's son and the other (named Janet) on a daughter of a friend of Lin's, that displays several facial expressions. They have two arms that can move and legs that can walk. Installed with a music notation recognition system and voice synthesizer, not only do they autonomously sing songs written with simplified musical notations, but their lips move to match the sounds they sing (Lin et al. 2011: 949). In the course of building their faces and walking capacity, Lin received assistance and advice from roboticists at Tokyo University of Science and Waseda University. The two wheeled robots, on the other hand, are 124 cm tall and have large, silver-colored trunks. Compared to their wide shoulders, their heads look small. Their arms are similar in design to those of the bipedal robots (Lin et al. 2009: 375), and they can also be installed with the hardware and software needed for singing by recognizing musical notation.
After the main technology for the four robots had been tested, Lin sought help from a team of theater professionals affiliated or linked with the Taipei National University of the Arts (TNUA). Michael Lee-Zen Chien, a lighting designer and professor at TNUA, organized a team consisting of theater designer, stage designer, and two media artists for the work. The collaboration between the engineering team and the theater team began about six months before the actual performance.10 After considering what the four robots could do onstage, the theater team and Lin decided that the performance would consist of five programs in theatrical form: singing by reading musical notation, puppet theater, an opera scene, portrait sketching, and a dance featuring human dancers and a drum-playing robot. Although the robot theater contained technical “shows” that made it less artful, it nevertheless combined many elements of theater. Not only were the robots given names, but they were put into story lines within which their characters were constructed. In order to accommodate the time needed for the robots to go on and off stage, interludes were prepared between the programs. The actual performance took about fifty minutes in total.
The programs and the interludes between them were combined in the form of a play, Robot Fantasia, which “is about a child's imagination for robots” (Lin et al. 2009: 375). The main theme of the robot theater agreed with the composition of the audience: along with guests from NTUST and roboticists, a hundred children from children's homes were invited. The performance began with contrived interactions between a small robot, which was not one of the four main robots, and a boy. The boy asked the small robot to play games such as hide-and-seek with him, but it kept on moving and turning around. The boy then got annoyed and shouted at it: “Don't keep on just turning around and moving! How stupid you are! You're not able to do anything!” A female character called Modern, played by a seasoned theater actress, then turned up and tried to inculcate in the boy a sense of respect for those behind the building of robots: “Building robots is no easy task.… It takes step-by-step work.… Even simple movements such as turning around take effort to design the circuit.… Robot builders are pioneers who spare no difficulty.”
The first of the five programs featured the wheeled humanoid Pica, named after Picasso. Pica reads romanized Chinese characters and sings. Before Pica sang children's songs, Modern appealed for and found a volunteer from the audience to write new lyrics to the music of those songs for Pica to sing. Pica's first few notes of the songs revealed its monotonic and mechanical tone and caused laughter from the audience. Pica was later assigned the job of creating a portrait sketch of the boy onstage. It took Pica about five minutes to complete the job. Several simple strokes led to the final portrait, below which was Pica's signature. A tinge of abstractness might make the audience think of Picasso, the origin of the robot's name. However, the background narration in the process of Pica's sketching—an enumerating of the technologies involved—totally unveiled the artistic abstractness of the sketch, tearing all of its mystery down to a combination of image and signal processing technologies.11
Before Pica's sketching, the two bipedal robots, Janet and Thomas, “sang” songs from the musical The Phantom of the Opera by lip-synch after manipulating a pair of marionettes. These small marionettes were modeled after and dressed exactly like the main female and main characters of The Phantom of the Opera. In the last program Ringo, who was similar in appearance to Pica, played drums. Ringo's drumming accompanied a dance by human dancers. After the dance, Janet and Thomas showed up again and kissed each other.
After the scene, the boy turned up again, repeatedly kissing the small robot, which elicited the response from Modern: “I know you will treat robots kindly. What a good boy!” When Modern and the boy were about to step down from the stage and conclude the performance, Modern suddenly stopped abruptly and was still for a moment. An investigation by the boy revealed that the sudden stop was caused by a device failure, and Modern was exposed as a robot from the future. After the problem was fixed, Modern asked the boy to keep the secret from others. The last few words Modern said to the audience concluded the performance of Robot Fantasia: “Be sure to examine devices inside your body so that you don't crash!” These last few words revealed the rationale according to which the theater team had arranged the performance: “With guidance of a female robot which comes from the distant future to the Earth, modern robots' development and functions are introduced” (Lin et al. 2009: 375).
The Construction of Seemingly Autonomous Agents by Theatrical Techniques
This section contrasts the ways in which robot actors are constructed as seemingly autonomous agents onstage. They fall into two main categories: technology and theatrical techniques. As shown in section 1, roboticists continue to use a variety of technologies to improve the quality of the appearance and unconscious movements of robots, as well as of human-robot interaction, so that robots appear more like agents. On the other hand, by way of the verbal as well as nonverbal interactions that take place onstage, theatrical techniques help to transform robot actors into their characters, filling the gap between robot actors (as objects) and the roles they play (as autonomous agents). Scenarios and stage settings contribute to make robot actors into their characters as well and should be categorized broadly as theatrical techniques. Scenarios and settings are not limited to one-to-one interaction, so theatrical techniques have the potential to go beyond the “more traditional,” individualized, one-to-one human-robot interaction and contribute to knowledge production of human-robot interaction. I will examine these ideas at work in the Japan case first and then the Taiwan case. I have mentioned the technology at work in these robots in some detail already, so this section focuses mainly on the theatrical part of their image creation.
In the Japan case, the status of robot actors as objects is transformed in theater in several ways. Take the Wakamaru, for example. First, unconscious bodily movements are emphasized. Second, because the Wakamaru shows no noticeable gendered traits, in I, Worker an apron is put on the “female” robot to differentiate it from the “male” robot. In addition, to make the gestures and movements of the Wakamaru more real, during the production of I, Worker a Bunraku master was consulted. Hirata explains that a primary reason for this consultation was to address the problem of how to impose gendered movements onto robots, something that can be learned from the long history of Bunraku performance (Hirata and Ishiguro 2010: 20). Gendered traits in bodily movements were thus emphasized. For example, the positions of the two hands when the robots stand still are different: the female robot put them before its abdomen; the male one, beside its body (Kuroki 2010: 62–63).
More important than bodily movements was the language that these robots used. As the above analysis by linguist Kinsui shows, the nuances in utterances such as the use of the interjections “hm” and “well” give the robot actors a command of language comparable to that of Japanese people. These nuances reflect the subtle context-reading ability that people have. Also, the robot actors had lines that were tailored for their characters. These all helped to transform the robot actors into their characters. The demonstrated competence of language and carefully crafted scripts put the robot actors into their characters by creating the characters' personalities, which were formed during conversations between robots and human actors. These characters were thus created for the most part through utterances, and to a lesser degree through the gestures of the robots. In I, Worker, the female robot Momoko is presented as a considerate character, with its femininity expressed through gestures and a voice that are stereotypically female. Here we can also see how, in addition to movements and language, the play's scenario works on a large scale to shape characters that reflect stereotypically gendered roles. In the final scene of I, Worker, as mentioned in section 2, in contrast to the female robot, which usually hesitates, the male one not only tends to show its inclination toward inferential reasoning but speaks decisively: “I can't think of it any other way” and “I have no other way of understanding what is beautiful.”
Hirata's theatrical techniques, such as scenario setting, clearly use social stereotypes to make robots more like autonomous agents. On the other hand, in contrast to roboticists' attempts at technological solutions, scenario setting is used to respond to and reframe the uncanny valley phenomenon in a way that moves beyond the individualistic, more traditional human-robot interaction. A case in point is Hirata's fourth robot production, Three Sisters, Android Version, which premiered in October 2012. Here, Hirata's notorious motto “Actors need only act like robots” finds another exemplification in a more complicated form. The structure of the play is roughly based on Three Sisters by Anton Chekhov, but the time and space are set in near-future Japan in a seaside city where factories for electronics and robots had moved overseas due to the appreciation of the Japanese yen, and only a robotics research institute remains. Two kinds of robots are used in this play. One is Geminoid F, and the other is Robovie R3. The Robovie R3 has the appearance of a service robot, and it plays the role of a service robot in the play.12
The story goes as follows. The father of the Fukazawa family, who died several years ago, was a famous roboticist. He was survived by three daughters and a son. The youngest “daughter,” named IKUMI, is a geminoid. “She” was built by the father to preserve the thoughts and appearance of the family's human youngest daughter, Ikumi, who was believed to have died of an illness before her father died. One evening, family members and friends gather at a farewell party at the Fukazawas' home for Fukazawa's former student, Nakano, who is about to leave for work in Singapore. Toward the end of the play, the geminoid unexpectedly reveals that Maruyama, a male guest and former colleague of Fukazawa, sexually harassed her in her childhood. Although the geminoid was programmed to keep all the former memories of the human youngest daughter, Ikumi, no one is sure whether the geminoid tells the truth. What is worse, the sudden appearance of Ikumi throws the party into further turmoil. It turns out that the father actually built the geminoid IKUMI because Ikumi became a shut-in. After IKUMI was built, the family claimed that Ikumi had died and did not disclose the secret to outsiders until the farewell party.
Interestingly enough, Ikumi claims that she is not sure whether there was a sexual harassment or not, explaining that humans forget. By juxtaposing the words of the youngster daughter, Ikumi, and her geminoid, IKUMI, the play underscores the issues of memory, trauma, and death. In the play, except for the scene in which Ikumi and IKUMI both are present and argue about whether there was sexual harassment, they do not appear onstage at the same time, and much more time is awarded to the robot IKUMI because it acts in place of the socially disabled Ikumi. The two dress exactly the same. Moreover, as the human Ikumi is a shut-in, she is pale, and enters and leaves the living room rather abruptly—quietly but rudely—just like IKUMI.
In the play, the geminoid IKUMI shows up first, and the audience at the beginning thinks (wrongly) that the human Ikumi is dead. When Ikumi first appears suddenly with an expressionless face onstage, a feeling of uncanniness—toward the human Ikumi, not the geminoid IKUMI—arises. The similarity of movements and sound of the two turns the idea of the uncanny valley upside down. In such a setting, a human—alive—can also make people feel uncanny. Moreover, the audience exclaims in consternation at the scene in which the two are both present—Ikumi shows up suddenly—when the quarrel between IKUMI and Maruyama is brought to a climax. According to an anthropologist who attended as a member of the audience, as the question of who is the real daughter puzzles the audience, an eerie feeling permeates the theater (Feng 2014). As the audience has become familiar with IKUMI as the story has unfolded, not only has the social role of the corporeal Ikumi been replaced by the mechanical IKUMI in the story, but the human Ikumi has been defamiliarized, corporeally, to the audience. The audience begins to wonder whether the one with the precise memory (IKUMI), or the one who wants to forget a traumatic past (Ikumi), is the real daughter. Actually, in the story, IKUMI often reveals uneasy facts about interpersonal relationships that others want to conceal. And this somewhat malicious personality turns IKUMI into a seemingly autonomous agent, one with an animate existence yet different from humans, just like the two robots in I, Worker.
For the Taiwan case, let us look at the appearance of the robots first. Pica and Ringo are more robot-like. The two bipedal robots, Janet and Thomas, have faces made with artificial skin that look human-like. However, below their necks there is not much covering of their mechanical, silver-colored bodies, especially their arms and legs. Lin asked about two hundred Internet users to see the videotaped performance and evaluate. Regarding the likeness to humans of the facial and limb movement of Janet and Thomas, “more than half of the audience were not satisfied” (Lin et al. 2013: 8). Lin claims that the uncanny valley effect is held responsible, and instead of being interested in various autonomous technologies installed in the robots, “audiences care a great deal about details other than the abilities that the robots possess, such as resemblance of appearances, motions and sounds.” Lin indicates that the problem lies at the technology level: “The research team needs to improve mouth shapes for the robots when they speak or sing onstage and the motion smoothness and speed of the theatrical humanoid robot's body movements” (Lin et al. 2013: 8).
In addition to work on the appearance of the robots, naming is another strategy Lin utilized to anthropomorphize the robots. Before they were put onstage, all four robots were given names and exposure in the news. While Janet and Thomas were named on a more casual basis, Pica and Ringo were named after famous persons according to the specific technology installed in them. This strategy was tried as a way of putting Pica and Ringo into their characters. The artificial intelligence that was used did give the audience a feeling that the robots were autonomous when performing certain technologies. For example, in Pica's singing, the lyrics of the last song it sang were written by a volunteer from the audience. Modern even showed the music score to the audience, assuring the audience of Pica's autonomy. In accordance with the composition of the audience, songs chosen for Pica's singing were children's songs, and the audience clapped to the rhythm and turned the auditorium into a class party. This shows how human-robot interaction was conceptualized in the Taiwanese robot theater. With the help of Modern, there was assisted interaction between the robots onstage and the audience. Human-robot interaction was thus not to be read between the script's lines, as in the Japanese case, but was facilitated by Modern instead. Later on, Pica was assigned to perform its “real” work, work that lives up to its name: sketching.
Compared to singing and sketching, puppeteering is special. The former two demand artificial intelligence technology, but, as Jerry Lin admits, to make a robot simply move its hands does not involve sophisticated technology.13 However, from the perspective of theater, puppeteering is a deliberate device used to constitute a seemingly autonomous agent. If well manipulated, puppets can be animated, which makes a lively performance while alluding to a hidden agent inside them. In the robot theater here, the scene was designed exactly this way: Janet and Thomas were unveiled after their puppeteering. They presented themselves as full-scale counterparts of the two marionettes. With Thomas wearing the same white mask as the puppet “he” had been controlling, Thomas and Janet “sang” songs from the musical—in lip-synch. Janet and Thomas were in this way assumed to play the two main characters of the musical, and their arms moved to match with the lyrics. However, in addition to the mechanical appearance of their bodies, as mentioned above, the range of their movements was also restricted.
After Janet and Thomas finished their singing, Modern praised the “love” between the two robots as admirable, saying that she was moved by their love, full of twists and turns. Modern thus tried to transform the two robot actors into their characters by emphasizing the opera's emotional side—a characteristic of humans. However, because the mechanical movements of the two robots did not coordinate well with their vivid singing (prerecorded human voices), the gap between the characters and robot actors was not fully compensated for by the narrative of Modern. A discrepancy existed among the humanizing narrative of Modern, the mechanical movements of the robots, and their human-like voices.
The comparison of the theatrical techniques of the two cases points to the extent to which the audience has a chance to make use of their social knowledge, and to the level of complexity to which this social knowledge reaches, in judging the playing of the robot actors. The robot actors in both cases try to stimulate the anthropomorphic tendency through their human-like appearances. Both attend to the uncanny valley effect, and gendered appearances and voices are emphasized in particular. This emphasis resorts to the gender stereotypes the audience might owe to their experience of everyday social interaction. Beyond appearance, however, the theatrical techniques used to transform the robot actors diverge in the two cases. In the Taiwan case, as the scene featuring robot actors Janet and Thomas shows, the actress Modern tries to invite the audience's empathy toward the robot actors through her narrative. The narrative content of Modern's appeal is stereotypical of an individualized relationship between a human female and a human male. In the Japan case, on the other hand, the scene in which Ikumi and IKUMI are present at the same time, arguing with Maruyama and with each other in front of other family members and guests, for instance, is a situation more complicated than can be connected with gender stereotypes.
For the audience, the Ikumi/IKUMI juxtaposition is a moral dilemma, the judgments of which need a deeper social knowledge about the desire of Ikumi to escape from traumatic memories, the insistence of IKUMI to expose the fact of Maruyama's wrongdoing, the eagerness of the family members to know the truth while protecting Ikumi, and the need of Maruyama to keep up appearances in front of acquaintances. IKUMI's seemingly autonomous agency is constructed in a series of such situations, and if the audience is to judge whether IKUMI lives up to her role onstage, they have to mobilize social knowledge distilled from past experiences of complicated many-to-many social relationships. This is not the individualized, one-to-one model of human interaction as usually envisaged in human-robot interaction.
Modes and Logics of the Interdisciplinary Collaboration between Robotics and Theater
Section 4 shows the knowledge needed to judge the performance of the robot actors in the two cases and suggested that knowledge of many-to-many social relationships is needed for the Japan case. In this section, by using the framework identified by Barry and Born (2013) in the introduction, I categorize the interdisciplinary collaboration between robotics and theater in these two cases and explore the implications of that comparison with regard to the relationship between the public and knowledge production.
I first examine the mode of interdisciplinarity, that is, the interrelationship between the component disciplines at work in these productions. The Japanese case can be said to belong to the integrative-synthesis mode, in which robotics and theater are in a relatively symmetrical relationship. In fact, before Hirata proposed the robot theater, Ishiguro had begun doing it. Suggested by the president of a robotics software development company, theater was seen by Ishiguro as a promising software application for robotics (Hirata and Ishiguro 2010: 16). However, with only the help of students, Ishiguro's robot theater had remained amateur, and it would not have gone far without Hirata's participation.14 As we have seen, Hirata not only provided scenario setting and scripts for the robot theater but also fine-tuned the gestures and movements of the robot actors in order for these robots to look more human-like—goals with which Ishiguro had long been concerned and that had motivated his building of various robots and experimentations toward achieving this aim by technological means.
The theatrical team in the Taiwan case likewise helped to improve the human-likeness of the theatrical robots. In rehearsals, the team asked opera singers to perform what the bipedal robots would for the opera scene, and they recorded the performances. The movements and tempo of the recorded rehearsals then became the basis of the computer program controlling the bipedal robots. The facial expressions and gestures of the opera singers in the recorded rehearsals also served as models for the robots to perform, and it was hoped that the timing of facial expressions and gestures matched the lyrics in a way that would be convincing to the audience.15 However, most of the staged programs were conceived by Lin before the theater team was formed. The theater team gave advice on the order of the staged programs so that it accorded with common principles of theater, which mix crescendos with diminuendos, and on the size and gestures of the puppets Janet and Thomas manipulated. Team members also provided a scenario for human actors to fill the time needed for the robots to enter and leave the stage—to smooth the way for an exposure of the audience to robots. Overall, the theater team discussed the staged programs with Lin and his lab members and tried to put these programs into a common form of theater after evaluating and testing their practicality.16 The Taiwan case therefore can be characterized, in terms of Barry and Born (2013), as an example of the subordination-service mode in which one component discipline—robotics—took a leading role. The robot theater was well known to many Taiwanese roboticists, as many of them were invited to the play. In contrast, few in the theater profession in Taiwan knew of the event.17
Viewed from the three logics of interdisciplinarity suggested by Barry and Born (2013), the rationales of the two cases overlap to some extent. Both are governed by the logic of innovation, searching for a new market of which the composition was not yet known; those interested in robots, as well as [theatergoers], would become attracted to robot theater, but the groups would not have come together if the collaboration had not occurred. Entertainment is an important element to foster a potential new market for both Japan and Taiwan teams.
The next question to address is how the two cases are governed by the logic of accountability, that is, the logic of “breaking down the barriers between science and society and leading to greater interaction, for instance, between scientists and various publics” (Barry and Born 2013: 14). As Barry and Born argue, although this logic brings science closer to the public, the logic comes in various guises in which the public is differently conceptualized. If art-science is practiced in the way of the traditional public understanding of science, it equates only to the popularization of science or an “aesthetic legitimation,” in which the public is summoned to witness preformed science, and art serves only “to render science communicable, comprehensible or non-alienating” (Barry and Born 2013: 15). But art-science can also be practiced in a way that emphasizes participatory public engagement. In this way, it is possible that the knowledge and experience of lay publics and nonexperts will be considered, and such interdisciplinary research is likely to create experiments in which not only the relationship between the public and knowledge is different, but both art and science are transformed (Barry and Born 2013: 20; Born and Barry 2013: 250); thus, this new relationship is characteristic of the logic of ontology.
In the Taiwan case, the public is conceptualized in a way more like the traditional public understanding of science. To launch robot theater—a combination between robotics and theater—is to find more chances to expose the public to robots, products based on an ever-developing robotics technology:
By uniting more powerful artificial intelligence, many faster and smaller multi-functional sensors, more efficient communication technology, and faster, economical computer calculation, intelligent robots in [the] new age can provide numerous new potential utilities beyond our imagination.… [Robots'] high price and costly maintenance prevents common people to own [sic] a robot. The concept of “Robot Theater” is to offer an easy opportunity for people to watch the entertaining performance of theatric robots and become more attached to intelligent robots. (Lin et al. 2009: 374)
Lin's response to the results of the questionnaire—which indicate that fewer than half of the respondents were satisfied with his robots' facial expressions and marionette manipulation—was to further technological development. Technology was then held accountable for all the problems the staged interaction revealed: “The important purpose of developing theatrical humanoid robots is for them to integrate into our society and even please human beings. Thus, their interaction with human beings should be enhanced in future applications and program designs.… In terms of appearance, facial expressions, motions and sounds, the resemblance of robots to human beings still requires further development to truly meet customer expectations” (Lin et al. 2013: 10). In this view, not only does theater play a minor role in producing knowledge for human-robot interaction, but the audience is seen as passive customers who “easily become bored” and can be pleased only by further technological development (Lin et al. 2013: 1, 10).
The time frames set in the scenarios perhaps coincidentally, yet nicely, captured the passive relationship between the audience and robotic technology. The robot character from the future, Modern, in human disguise, introduced the current level of robot technology to the audience. The staged programs show that robots “have started to learn human skills” and “may express love and with [sic] emotions” (Lin et al. 2009: 375). The scene in which Modern points out to the boy the hardships of building intelligent robots exemplifies the assumed relationship between technology and the public. Robots are still in development, and as they become more advanced in the hands of roboticists, people can anticipate more intelligent and versatile robots that will better serve human needs, including the need for entertainment. All we have to do is recognize the efforts of roboticists and wait for them to bring us a future in which functionally competent robots help us. Audiences are conceived of as witnesses of the current state of robot technology. Although they may be somewhat detached onlookers, they are sure to be entertained by the more and more intelligent and human-like robots of the future. This time frame establishes a future-oriented development and places the audience in the here and now, looking forward to the ever-developing future given by roboticists. This characterizes the kind of robotics that finds unconditional support from the audience in the form of the interdisciplinary collaboration of robotics and theater, in which the latter component discipline plays a minor role.
The Japan case renders the audience a more active role, compared with the Taiwan case. This is shown in two ways. First is the relationship between robot technology and the audience. Precisely speaking, the audience's attention is not directed predominantly to current robot technology. In this sense, the audience is not a mere witness to current technology. Instead, as the capabilities of robot actors in the story setting far exceed what robots can actually do now, the audience is invited to imagine what robot technology in the future might bring them. All of the scenes of Hirata's robot theater are set in a near future in which robots serve as servants, companions, or surrogates to humans. But no matter how closely humans and robots live with each other, they remain psychologically distant (Hibino 2012: 40). Although the time frame of both the Japan and Taiwan scenarios is future-oriented, in the Japan case robot technology is presented as having a limit in creating emotions, as shown in the final scene of I, Worker. The scenarios therefore invite the audience to reflect on various situations they might encounter in the future.
Second, the role of the audience in the knowledge production of human-robot interaction is worth examining. Every time Hirata's robot theater is put on, the audience is given questionnaires designed by Ishiguro's laboratory to single out precisely the moments in which, in their eyes, the theatrical robots are the most and the least human-like. Also, following a convention of Japanese theater, Hirata often offers “aftertalk,” a discussion session of some thirty minutes after each performance of robot theater. In these sessions, Hirata (and sometimes Ishiguro, or both) explains why he got involved in using robot actors, and the audience is given a chance to exchange ideas about robot theater with Hirata.18 Taken together, these opportunities give the audience more chances to participate in transforming robot actors into their characters. As we have seen, by incorporating social phenomena such as the emergence of NEET (not in employment, education, or training) and shut-ins into the story line, Hirata actively uses social knowledge that the audience already has and tries to elicit their responses to a familiar social setting in which robots represent a new variable. In this way, Hirata is arguably using the audience's existing social knowledge of the human world to experiment with the ongoing construction of their new social knowledge of a world in which humans and robots live together. In this robot theater, human-robot interaction therefore goes beyond an individualistic, one-to-one relationship and is embedded in a collective social setting in which, if robots are seen as autonomous agents, they are seen as such because of their relationships to other humans and robots.
Hirta's theater therefore provides a quasi-daily setting for the robotics field to experiment on the human-robot relationship. As we have seen in Three Sisters, Android Version, the uncanny valley phenomenon is not always limited to a one-to-one relationship; here it is recast as embedded in a social relationship comprising more actors. In this way, this robot production has at least broadened the problem space of robotics, if not brought about fundamental changes in the field itself. For human-robot interaction in the Japan case, art has given robotics some new ideas, as Hirata's directions for the movements of robot actors are being generalized and weighted with context-dependent rules. If these rules are successfully constructed, a “motion editor” can be used to adjust a robot's movements according to its interaction context for it to present more natural movements (Bono et al. 2014: 332). On the other hand, if this motion editor is realized, this may mean that theater is also transformed: the artistic intuition of a director, to a certain extent, can be replaced by the processing of a computer and its application of some mechanical rules. As this is still ongoing, it is not certain yet whether theater and robotics have transformed each other, and thus the rationales behind the interdisciplinary collaboration, in addition to the logics of accountability and innovation, can also be said to be driven by the logic of ontology, as Born and Barry argue (2013: 249–50). Suffice it to say that the component disciplines have the potential, in this collaboration, to realize a possible transformation by engaging with each other and with an audience given a more active role in knowledge production of human-robot interaction.
This comparative study therefore shows the differential roles offered to the public for their social knowledge to play in the collaboration of two disciplines previously thought to be distant. In broader terms, however, the collaboration of theater and robotics not only serves as a case of interdisciplinarity for investigating public engagement in the process of knowledge production. It also provides a chance through which the relationship between technology and society in these two East Asian countries can be glimpsed. The different future-oriented scenarios in the two cases constitute an allegory of technology (as represented by robotics technology) in relation to society. In the Taiwan case, the audience is conceptualized mainly as witness and recipient of current robotics technology, as if technology were the main character, calling for support from a complaisant society. This passive relationship accommodates to the view of ever-lasting technology development, as roboticists and related professionals there try to catch up with more advanced robotics technology. In the Japan case, however, although the robotics technology is more advanced, Hirata's scenarios always depict the decline of Japan (Bono et al. 2014: 333), which makes robotics technology, as represented by robot actors, stand out in stark relief against a general societal anxiety about the future. And this anxiety cannot be alleviated by more technology, as the problems IKUMI causes in Three Sisters, Android Version show us. This depiction therefore invites a more active participation in reflections on not only benefits but also potential problems that technology would bring about.
I would like to thank Dr. Ruey-Lin Chen, two anonymous reviewers, and editors of EASTS for their constructive comments on previous versions of this article. My thanks also go to Interchange Association, Japan, and the Sumitomo Foundation for funding part of this research.
Social robots are those robots claimed to convey and “read” social cues according to which they interact with humans.
International Conference on Robotics and Automation 2012 Workshop, “Call for Papers,” http://www.robotics-and-performing-arts.sssup.it/cfp.html (accessed 9 July 2014), and “Overview,” http://www.robotics-and-performing-arts.sssup.it (accessed 9 July 2014), emphasis added.
Section 1 explains how roboticists conceptualize and practice individualistic, “more traditional,” human-robot interaction.
Animism has different meanings and connotations in anthropology of religion and in psychology. In the former, it is often associated with “primitive” thinking, and this association has been criticized as Eurocentric; in the latter, it is seen as an inclination of children, which will diminish when they grow up. However, scholars also argue that it is relatively universal in adult humans because animism has an evolutionary advantage (Guthrie 1993). As for anthropomorphism, it is also held to be relatively universal in humans, but variations among people do exist (Waytz, Cacioppo, and Epley 2010). Further debates of the two concepts are beyond the scope of this article.
From the perspective of gender, it is clear that Japanese roboticists, Ishiguro in particular, take gender stereotypes as useful in facilitating human-robot interaction (Robertson 2010), as can be shown from the fact that Ishiguro's first android built in adult size was female in appearance. However, as his team realized, the impression androids give to people can easily be changed by changing their eye contours, hairstyles, and dress, so Ishiguro and colleagues have conducted various experiments using androids with different looks. In these experiments, gender is operationalized as a variable. For example, they use an android or a geminoid (see text) as a dummy nurse sitting alongside a medical doctor to test whether it eases patients’ tension. Sometimes the android/geminoid is male in disguise. Considering that male nurses are rare in Japan (about 7 percent of all the nurses), although these roboticists may take a stereotypical view of gendered appearance, it can be argued that they take a relatively agnostic view toward gendered roles in these experiments.
For example, from social interaction and phenomenological perspectives, critics argue that ToM assumes a gulf between an individualized self and others that can be bridged only by theoretical means—inference, for example (Leudar and Costall 2009; Zlatev et al. 2008).
Oriza Hirata, interview by the author, 22 August 2012, Tokyo.
Hirata's belief that he might justify his “contemporary colloquial theater theory” through the use of robot actors is contested by theater commentator Hiroko Yamaguchi. She argues that it is possible for the audience to be moved by anything onstage, be it dogs or inanimate objects like leaves or puppets, and the audience's attachment to these objects does not relate to whether there are minds within them (Kawai and Yamaguchi 2010: 99). While I concur with her on this point, this article is concerned with, and the problem of robot theater lies exactly in, how such an effect is produced onstage by using robots comparable to humans in size.
Jerry Lin, interview by the author, 14 May 2013, Taipei.
Michael Chien, interview by the author, 10 December 2013, Taipei.
This can be compared to a similar performance of “autoportrait” drawn by a robot arm designed by several German artists who formed the group Robolab and developed several robot performances by using industrial robots. What is special about autoportrait is that, every time a portrait of a member from the audience is done, the performing robot arm then rubs the portrait off, without exception. To some audience members, this steadfast act of erasing one's own work seems to have philosophical overtones and gives the impression that the robot arm has a certain “personality” trait that serves to construct its perceived agency.
With a size a little bigger, and a ratio of head to body size larger than those of Wakamaru, Robovie R3 also moves on wheels. Besides acting in Three Sisters, Android Version, beginning in 2013, Robovie R3, with its longer battery life, has replaced Wakamaru for the play I, Worker.
Hiroshi Ishiguro, interview by the author, 5 July 2012, Osaka.
Aftertalk also serves a marketing function, as directors in these discussion sessions often appeal to the audience to help promote the performance to others. Sometimes in Hirata's aftertalks, the audience airs views on theater/robots very different from, or even oppositional to, those of Hirata or/and Ishiguro.