A political analysis and critique of social robots could pursue many factors, for instance, form, function, interaction, and experience are the commonplaces of design and design criticism. These factors, however, are surface expressions of the principal quality that makes social robots distinctive as a category of computational objects. That is, form, function, interaction, and experience are outcomes or effects of the design of a robot 's embodiment. To understand how embodiment is treated in the design of agonistic encounters with robots—how it evokes political issues and can be interpreted politically—requires first clarifying how embodiment can be understood as a designed quality of an object.
Unlike procedurality, a quality that is easily attributable to the medium of computation, claiming embodiment as a defining quality of robots is difficult because it is often construed as a quality that is distinct to living entities. Theories of embodiment from phenomenology and embodied cognition have woven their way into robotics research through the fields of artificial intelligence and the associated questions about what is required for knowing and acting in the world.9 Embodiment in robotics is similar to notions of embodiment in philosophy and cognitive science, where embodiment is concerned with the body of an entity and its capabilities in defining being in the world. But within robotics discourse, embodiment is not limited to living entities: it can just as easily be a quality of a nonliving entity.
The particular characteristics and effects of embodiment are a significant topic of research in the field of human-robot interaction. Roboticist Kerstin Dautenhahn is one the scientists at the forefront of this research, and her work is useful for framing an understanding of embodiment beyond living entities. As Dautenhahn and her coauthors state, in terms as equally applicable to a robot or human, embodiment can be characterized as "that which establishes a basis for structural coupling by creating the potential for mutual perturbation between system and environment" (Dautenhahn, Ogden, and Quick 2002, 400). In her research, Dautenhahn has been concerned in particular with embodiment from an experimental science perspective. She has developed a framework for embodiment that can be operationalized and quantified, such that degrees of embodiment might be empirically compared among different entities in different environments, with different capabilities and affordances. From this effort, she has developed a definition to describe the state of embodiment within any system: (Dautenhahn, Ogden, and Quick 2002, 400):
A system S is embodied in an environment E if perturbatory channels exist between the two. That is, S is embodied in E if for every time t at which both S and E exist, some subset of E's possible states with respect to S have the capacity to perturb S's state, and some subset of S's possible states with respect to E have the capacity to perturb E's state.
Even when the goal is not the quantitative measurement and comparison of embodiment, such a definition is useful for reframing common notions about embodiment. Such a definition enables moving beyond the notion that embodiment is a quality limited to living beings. As the variables in the definition denote, embodiment is particular-to-particular structures. Each configuration of a robot establishes a different set of possible couplings and mutual perturbations with the environment. Each environment or entity is likewise characterized by certain qualities and affordances that engage these configurations differently, resulting in a diversity of kinds of embodiment. Extending embodiment beyond living entities alone to include the artificial is significant because embodiment can therefore be treated as a quality that can be shaped and manipulated by design—by the choice and arrangement of particular aspects of the robot, including its hardware; software; capacities for sensing, processing, and actuation; form; and behavioral qualities.
But the form and structure of embodiment alone do not shape encounters with robots. Context must be considered with embodiment. The meaning of different kinds and experiences of embodiment will be construed based on the broader sociocultural contexts in which the robot is set. The home is not the battlefield; service animals are different from pets; seals are not familiar domestic creatures. So even with Dautenhahn's operational definition, embodiment is far from a reductive mechanistic quality. It is highly contingent. Different configurations or reconfigurations of embodiment produce different human-robot encounters in different contexts depending on expectations and desires in these situations. Recall the design of PARO 's embodiment—soft fur, purring sounds, and wriggling motions in response to being stroked. Such a design may be appropriate for the contexts of the home, retirement center, or hospital but not to comfort wounded soldiers on the battlefield.
Since embodiment makes robots distinctive as a category of computational objects, it is a promising site of agonism. Designed embodiments and the subsequent forms, functions, and interactions that run counter to our expectations produce deviations from the norm in social robot design. These deviations result in experiences that challenge the familiar and proverbial in social robot design and expose topics of debate concerning our future relationships with robots. Examples of such designs and encounters often lurk just outside the established fields of interaction design and human-robot interaction. Suchman and others have drawn attention to such examples as sites of inquiry.10 My intention is to build on those analyses through the frame of agonism to draw out design exemplars of social robots that evoke political issues and relations.