Artificial Sociality:Experimenting with Social Robots in Japanese Laboratories and Anthropological Analysis

In this dissertation, I report on anthropological research conducted on and with social robots and roboticists in Japan since 2017. I explore how social robots and their creators collectively experiment with, and possibly even reinvent, the phenomenon – hu-man sociality – they originally intended just to reproduce artificially. Empirically, the dissertation builds on a combination of ethnographic fieldwork in Japanese robotics laboratories, interviews with roboticists in Japan, and close readings of their published scientific work, as well as other forms of dissemination that introduce their research to the public. Conceptually, I draw on and contribute to discussions of more-than-human sociality and the performativity of experimentation in social anthropology and STS to investigate the processes of designing, developing, and evaluating robots that techno-logically simulate human sociality. As such, I offer opportunities to think along with me as I grapple with what it might mean to say that a robot is social. Robotic sociality is a contentious issue in the humanities and social sciences, where critics of social robotics have argued that the dynamics of human sociality do not lend themselves to technological or computational replication unless they are trans-formed beyond recognition. On this account, robotic sociality only exists in the imaginations of roboticists who cannot comprehend that their methods are essentially inadequate for replicating something so very central to being human. The problem is not only tricky replications but also the uses to which these reductive illusions are put. In this respect, roboticists often claim that social robots can provide novel advantages for investigating what it means to be human, to the extent that building social robots presents opportunities for determining and re-engineering fundamental mechanisms of social cognition, communication, and interaction both computationally and experimentally. The roboticists I worked with in Japan envision their emerging robotic creations as potentially offering new forms of social companionship in an increasingly disconnected world while simultaneously employing them as experimental tools for conducting innovative scientific research into the central mechanisms of human sociality and social interaction. Nevertheless, there still seems to be a tension between reproducing a form of robotic sociality that thrives in the ‘real world’ and the rigidity of computational and experimental methods required to produce both such replications and scientific knowledge about the mechanisms of sociality in the first place. Put differently, it becomes questionable whether the professed goals of social robotics – creating artificial entities that simulate sociality and using them as experimental apparatuses for building a new science of sociality – are realistically achievable given the constraints of computational modelling and scientific experiments. This dissertation explores these frictions between replicating sociality in robotic systems and the forms of experimentation such replications manifest. I suggest that such subtle tensions in the experimental practices of social robotics engender opportunities for thinking, designing, and enacting sociality differently. I trace how these tensions figure as central dynamics in the practices of social robotics but also how it pro-duces generative openings for thinking with alternative models of sociality in anthropo-logical analysis. As such, the dissertation does not offer an ‘ethnography’ of social robotics in contemporary Japan. Instead, I deploy the duplex of replication and experimentation to stage an experiment in anthropological analysis that works athwart theory via inventive and lateral forms of conceptualisation. This means that the tension be-tween replication and experimentation not only characterises the parallel goals of social robotics – replicating sociality artificially and designing new experimental apparatuses – but also figures as a crucial component of the model of artificial sociality I offer as a different way of appreciating the transformative potential of social robotics. Artificial sociality is concerned with a relational dynamic in which modelling sociality becomes a defining feature of sociality itself. It tries to move our imagination from questions of intelligence and thinking towards sociality and relations. It focuses on how models of sociality are necessarily transformed in practice when social robots interact with dynamic relational worlds, in turn looping back on the models they were initially designed to simulate, like human sociality, so that these models are similarly transformed in the process. This also means that replication and experimentation, analytically routed through artificial sociality, effectuate an understanding of anthropological analyses as a form of lateral redescription that works athwart central problematics in social robotics and anthropology by questioning conventional distinctions between the empirical and the conceptual, as well as the relations of correspondence between them. The dissertation thus produces analytical momentum by bringing two different experimental forms of inquiry together, anthropology and social robotics, and exploring the space of equivocal comparisons and models that emerge as a result. Throughout the chapters in the dissertation, I unravel the relational and performative interfaces be-tween replicative ambitions and experimental practices in the enactment of artificial forms of sociality in Japanese robotics laboratories and beyond. In the first half, I lay the ethnographic and conceptual foundation for the dissertation at large. In chapter 2, I review recent work on robotics in Japan and so-called Japanese ‘robot culture’. I show that, for the most part, the existing research gloss over the scientific aspects of social robotics in Japan, focusing instead on critiquing the cultural, ideological, and political imaginaries enacted in the visions of the Japanese as uniquely accepting of robots. In chapter 3, I survey and discuss several ways of model-ling human and nonhuman sociality in social anthropology, STS, social robotics, and beyond. Here, I am primarily concerned with models of sociality that position humans as uniquely social animals, as well as recent work by anthropologists, STS scholars, and others who argue for extending and expanding models of sociality beyond individuals and humans. Based on classical and contemporary work in STS and anthropology, chapter 4 develops an analytical perspective on the performativity of experiments, focusing on the interfaces between scientific experiments and stage theatre and how per-formative and inventive experimentation can be used anthropologically. Each of the three chapters in the dissertation’s second half explores problems central to experimental research in social robotics and which are similarly evocative figures in anthropological theory. In working with these problems, I deploy analytical devices that intersect with, expand on, and occasionally trouble the duplex of replication and experimentation. Chapter 5 explores the analytical generativity of scripts in the ‘Curiosity Experiment’, which I was introduced to at a seminar organised by Hiroshi Ishiguro’s laboratories at Osaka University and ATR. I trace how the Curiosity Experiment was staged in three settings (a seminar, an interview, and an article) to show that while scripts are associated with replicability and uniformity, they can also be vehicles for the proliferation of differences and multiplicity. Chapter 6 expands these discussions based on my experiences as a participant in the Curiosity Experiment and observations of subsequent iterations. Here, I show how an algorithmic form of artificial curiosity reconfigures interactions between humans and robots via continuous expansions and contractions of interactive complexity. I describe such pendular motions be-tween simplicity and complexity as a process of complication and discuss how this troubles the obsession with complexity in anthropological theory. Chapter 7 uses prototyping to explore how different experimental renditions of embodiment in social robotics simulate, extend, or delete the physical and social dynamics of human bodies. I show how the recursive characteristics of prototypes un-fold the mutual transformation of both robotic and human bodies, revealing their intrinsic indeterminacy. Finally, the epilogue in chapter 8 reconnects the problems (difference, interaction, and embodiment) and devices (scripts, complications, and proto-types) to rethink the critical potential of artificial sociality as a model of sociality that models socialities. I then think through issues I have left mostly untouched, like culture, affect, ethics, and gender, which are critical for ensuring that artificial sociality becomes a valuable contribution to anthropology, STS, and social robotics. Lastly, I consider how to move artificial sociality outside laboratories without losing the tensive creativity between replication and experimentation.