How that robot made me feel, edited by Ericka Johnson

The landscape of studies addressing the issue of emotions and affects in their entanglement with technologies has grown rich in recent years. Attention to the subject was already present in the history of science and medical anthropology (Dror, 1999; Dror et al., 2016; Riskin, 2002, 2010); more recently, it has come from media studies (Blackman, 2019; Bollmer, 2023; McStay, 2018, 2024), science and technology studies (Atanasoski Ellis Clough Sedgwick, 2003).The essay collection How that robot made me feel fits into this landscape, with the specific aim of flipping an important question in the field of human–robot interaction (HRI): Instead of asking whether robots have emotions, the anthology starts with the realization that, in interactions with social robotics devices, it is our emotions that are at stake. The book consists of an introduction followed by nine chapters and is accompanied by a general bibliography listing all the references in the volume (bibliographical sources have been included at the end of each chapter as endnotes), a list of contributors with a brief biographical note, and an index.The contributions question how human emotions and affects are targeted and played on by social robotics systems. In the introduction, Ericka Johnson, the volume's editor, draws a clear line between how the humanities and social sciences study emotions (with phenomenological, cultural, and contextual approaches) and affective computing. The contributions in the volume are presented in a framework establishing a bridge between the fields of HRI and affect theory.1 As a corollary, the anthology presents the need to broaden the category of social robots to include software for emotion detection, robot animals, and assistive robots for work tasks, as well as humanoid or general-purpose robots.The contributions vary significantly in terms of approaches, fields of reference, methodologies, and insights into the subjects of study. This article is therefore a summary of the impressions from each chapter, highlighting the aspects that may be most relevant to the journal's readers.*The first contribution is the most interesting in the volume from the point of view of a social critique of affective computing. Written by Jennifer Rhee, it is specifically dedicated to emotional AI. Opening with a discussion of the narrative of affective computing with respect to autism, the chapter presents emotional AI as a technology of deauthorization. Deauthorization of whom? Primarily of autistic individuals and their power and credibility in speaking for themselves. Rhee shows how the heroics of the technology that characterize Rosalind Picard's narrative of affective computing (Picard, 1997) are built on the dehumanization inherent in the figuration of autism that is embedded, conveyed, and reproduced by emotion recognition (ER) systems. Rhee's critical focus is on the interweaving—both in the rhetoric of affective computing and the design of emotional AI—of two controversial theories: the theory of mind (ToM) or mind blindness (notoriously authored by the psychiatrist Simon Baron-Cohen) and Paul Ekman's theory of emotions. At the heart of the article is an ethical-political problematization, dismantling the oddly acquired authority of ER systems. This interrogation combines the elements of gender, race, and disability to ask: Who can talk about whose emotions? The stakes—at the political-epistemological level—are neatly identified by the author. The text concludes with the presentation of two artistic works from 2020: Vibe check by Lauren Lee McCarthy and Kyle McDonald and Elisa Giardina Papa's Cleaning emotional data, both of which invite the viewer to imagine emotional AI differently, troubling the conceptualizations of emotions at work in recognition systems and the colonial rationales they reproduce.Chapter 2, “The role of intentionality in human–robot interaction” by Tom Ziemke, explores HRI through one of the most established themes in the field: intentionality. The text distinguishes between two types of intentionality: The first is related to “biological autonomy,” or the autonomy that the construction of robotic and computational systems would help model scientifically; the second is related to the autonomy with which we seek to endow robotic systems, called “artificial autonomy.” The author acknowledges the tension between these two types of autonomy. In sum, the question is not so much whether intentionality operates objectively in devices endowed with artificial autonomy but rather how to deal with the awareness that this intentionality is assigned in interaction. In this regard, the author's proposal concerns the implementation of an “interactive autonomous systems practice” that would complement the systems autonomy proposed by Francisco J. Varela and Paul Bourgine (1992) to guide the expectations of humans. This is a matter of providing users with the ability to predict and interpret the behavior of robots and be aware of their limitations. Underlying the proposal is the belief that the main ethical and social problem concerning HRI is the deception of the human agent.Very different in content and tone is “The way things feel” (chapter 3, Ahmet Börütecene), which is a narrative fiction piece describing the feelings and perceptions of a group of colleagues working in the baggage handling system of an airport. The team members are paired with a new colleague, Teller, a nonhuman-looking robot (the drawings depict a kind of thick mobile disk, similar to a robot vacuum cleaner, with icons that light up on the upper side) that assists them in their work tasks by predicting system activity flows and possible critical moments. The text follows the experiences of each of the group members, imagining how—through their interactions with Teller—the collection and transmission of information about the daily workflow are transformed into dynamic and embodied sensory experiences. The text's aim, as revealed in the end, is to employ narrative speculation to weave together concerns of design politics and critical viewpoints. However, it appears that the critical views were overlooked; instead, a picture is painted of affectionate collaboration with nonhuman colleagues that does not question the additional demand on the workers inherent in the interplay of the body and emotions as labor. To quote Donna Haraway's famous manifesto, “from all work to all play, a deadly game” (Haraway, 1991, p. 161).Chapter 4, “Robots playing with emotions,” by Roger Andre Søraa and Mark W. Kharas, inscribes itself in the field of science and technology studies and asks what it means to open the “black box” of robots “trafficking emotions.” The core argument lies in the assumption that robots need to be better accepted in society and thus that more research is required into the emotional relationships that “citizens” have with them. The idea of robots as a “technological imaginary” is employed to explain the qualitative data gathered through workshops, surveys, and interviews for a European research project called Robotics4EU, reported to be in the preliminary results stage.The fifth chapter (“Robot animals and the emotional labor of caregivers,” by Marcus Persson, Clara Iversen, and David Redmalm) takes the reader inside care homes for the elderly with dementia in Sweden. The study's perspective is thought-provoking, focused on the emotional labor (Hochschild, 1983/2022) of care workers. The animal robots that are introduced into the facilities (robotic cats that meow and purr) are presented as sociotechnical artifacts; they acquire meaning through context and social arrangement. We see in fact how, in different situations, these social robots acquire meaning and context for the residents through the hard and sometimes frustrating emotional labor of care workers. The quotes taken from the workers’ discussions are very vivid; their interpretation and classification do not adequately capture their poignant rawness and deep sensitivity.The volume editor also authored chapter six, titled “Robot meets pets.” The text has a rather personal tone and recounts her experience during the COVID period with two animal robots introduced in previous chapters: Paro the seal and a tabby cat robot. After a series of accounts of these animal robots’ encounters with neighborhood children, Johnson decides to attempt encounters with neighborhood pets, which she films and posts on the TikTok platform. These meetings do not seem to be a success in terms of interesting observations in “social dynamics” (the neighborhood pets do not show much interest in the robotic animals), but their “failure” is chronicled with humor. The author then ventures through the main themes and goals of the literature in the field of animal–robot interaction (ARI), drawing parallels with the field of HRI. The text concludes with a series of scattered questions, attempts at future analysis, and narrations of the author's own doubts, feelings, and ethical uneasiness.The theme of failure is also present in chapter seven, “The imperfectly relatable robot” (Katherine Harrison, Kavyaa Somasundaram, and Amy Loutfi), which asks: What happens when the human–robot relationship fails? With the new learning paradigm called reinforcement learning, robots learn through trial and error and rewards and punishments. How do these errors (occurring by design) impact HRI? The text is a collaboration between a social scientist and two roboticists, presenting how, inspired by reinforcement learning, the field of HRI might flip its attitude toward error—from attributing to it a generally negative value to seeing it as a resource from a managerial perspective. The article advances the ambition of integrating theories of affect and critical-political approaches concerning the sociocultural and historical determinants of “failure.” These theoretical elements, however, are only evoked; they do not acquire a transformative role in the analysis, which maintains a managerial lens.The penultimate chapter “The difference between you and me(chanical) robot” is authored by Karin Danielsson. Introspective in nature, the text attempts to convey an “existential inquiry” approach. The form is that of a diary: the narration of the author's inner monologue in connection to three dates in 2022. The situations that form the background and inspire the spontaneous reflections, sketched ideas, ruminations, and reported feelings are, in order: the setting of Danielsson's student's defense of a thesis titled “Being connected to the world through a robot”; the viewing, on social media, of a clip showing Disney's Spider-Man robot crashing during a stunt; and her reading, during a train ride, of Kazuo Ishiguro's Klara and the Sun (2021). This highly personal and self-centered narrative may puzzle the reader seeking a systematic argument. The text betrays some inclination to see empathy as a social “skill” or “ability” that safeguards the differentiation between humans and machines and that the author seems anxious to protect, for instance, wanting to teach her children “what it means to be human.” In this sense, there is a tension between the professed—alas not pursued—necessity to discuss the work of scholars critical of essentialist theories, to open up to the conjunctural and historical (genealogical) dimensions of the categories structuring the field of human sciences, and to examine the author's own existentialist motivations of self-discovery as a search for the human from the human, with the promise of learning the intrinsic qualities and “maybe make me become a better human” (p. 181).The volume concludes with a chapter by Daniel White and Hirofumi Katsuno titled “Mindful machines.” This well-structured, informative, and elegantly written essay invites the reader to explore the Buddhist “model” of emotions to challenge hegemonic ER models. Opening with a story from the Pāli Canon called the Sallasutta—a short Buddhist allegory of emotion—the two social anthropologists ask how today's technologies equipped with artificial emotional intelligence provide emotional care. The hegemonic model of “artificial emotional intelligence” (as shown in the first chapter) is that of “basic emotions,” specifically the Facial Action Coding System (FACS) by Paul Ekman and Wallace V. Friesen.2 The study arranges the argument in three parts. The first part discusses proposals in the field of AI and robotics ethics advanced by researchers who draw on Buddhist philosophy, such as Soraj Hongladarom and Peter Hershock. Such discussions invite the reader to radically rethink the “hyperidividualistic” conceptions on which the Western idea of autonomy that models the automation of AI systems is formed, to lead toward a relational and mutually constitutive conception of reality. The second part explores how Buddhist-derived concepts of emotion, particularly “compassion,” were influential in the field of robotics in Japan before the emergence of North Atlantic–rooted affective computing. This refers in particular to the work of Mori Masahiro (the roboticist who popularized the term “uncanny valley”). The conclusion exposes the study's proposal: the notion of “mindful machines,” exemplifying and discussing it through two collaborative ventures between Buddhist practitioners and engineers, one in Japan and one in Cambridge. This compelling chapter leaves two open questions. The first question concerns the impact that the Buddhist holistic conceptions of reality would have had—for example, on cybernetics (Hayles, 1999; Pickering, 2010)—if not for Gestalt theory. One may wonder, in this respect, how this—undoubtedly seductive—conceptual opposition would play out from the standpoint of an analysis that takes into account the historical dimension of AI on which current ER technologies rest (i.e., connectionist AI and pattern recognition). For the second question, the conclusion seems to leave in suspense the issue of whether and in which respects the “mindful machines” diverge from the “conventional” ER technology described at the beginning (and eloquently rebuked in the first chapter of the volume) in terms of software.*To a readership interested in sociopolitical critique and cultural analysis of robotic technologies targeting emotions, I recommend paying particular attention to the first, fifth, and last chapters of the collection. In different ways, the first and the last chapters effectively critique and highlight aspects related to ER systems, exposing the normativity of their classifications and problematizing their underlying conceptions and current stakes. I suggest the fifth chapter because it sheds some light on reproductive and affective labor that is still inconceivably invisibilized by the glittering imagery that informs the computer science literature on assistive social robotics. It therefore represents a necessary gesture in the field of HRI.My advice to readers of the journal is to supplement the introduction with literature discussing the affective turn and the current debates around emotion theory scholarship. The introduction to the volume tends to promote a simplistic, bipartisan approach to the various positions, aiming for self-validation based on disciplinary affiliation; however, the reality is far more intricate.To the literature cited in the apex of this review, it is good to add, in this regard, readings that are already a few years old, particularly Kyla Schuller's fine historical work The biopolitics of feeling (2018), the collection of essays Timing of affect (Angerer, Bösel, & Ott, 2014), as well as a cautious3 reading of Ruth Leys's work, The ascent of affect (Leys, 2017).The main challenge of this volume is to analyze how specific social and cultural contexts relate to the epistemic structures behind emotional AI and social robotics, including their history, imaginaries, and politics, and how these relate to theories of affect and emotions. Not all contributions fully succeed in meeting this challenge, but it remains an important and valuable aim.