Lea Laura N. Michelsen reviews Machine Vision: How Algorithms are Changing the Way We See the World by Jill Walker Rettberg. Machine vision is all around us, for good and bad, but who has the power to influence how we use it?
A Perhaps Not So Posthuman Book with a Very Specific Point of View
Jill Walker Rettberg’s new book, Machine Vision: How Algorithms are Changing the Way We See the World, is a rich and well-communicated humanistic introduction to machine vision and the ways in which it affects human vision and life. Its examples are well-chosen and it gives rise to many reflections. The book develops a fruitful analytical approach to understanding the entanglements of humans and machines today. Its theoretical and analytical framework can be used in humanistic research and education to grasp and analyze complex technological developments and their sociocultural implications.
Machine vision is widely treated in humanistic literature, often with a critical focus on limitations, inherent technological biases and failures, and negative sociocultural effects. For instance, biometric facial recognition has been widely criticized for its tendency to privilege normative white, Western subjects and harm non-normative subjectivities (Gates, Our Biometric Future; Magnet, When Biometrics Fail). Object recognition has also faced scrutiny for their biases and racist categorizations (Crawford and Paglen, “Excavating AI”). Furthermore, machine vision in drones has been thoroughly criticized for reducing human life by creating a distance to the target (Maurer and Engberg-Pedersen, Visualizing War; Weizman, The Least of All Possible Evils).
Rettberg’s book presents a somewhat different attitude towards machine vision. While it contains critical perspectives, its approach is more playful and curious. Aiming to maintain hope and highlight the positive aspects of current technological developments, is an active choice. The book’s playfulness and predominantly positive attitude encourages researchers and students to engage with technology rather than fear it, thus providing hope for future generations.
Rettberg addresses feedback received during her writing process that notes how her book seems rather preoccupied with the human for a posthuman book on machine vision. To this she writes: “While I do think it’s important to de-centre the human and to acknowledge that we are just one of many species, I can’t help but see the world through human eyes, although I also see with technology.” (23). I would argue that we are not only presented with the world seen through human eyes, but through Rettberg’s eyes. This specific point of view, in many ways, works really well. It gives us relatable analyses, anchored in the embodied experience of a concrete person, and a more playful attitude towards technology. The downside is that it, at points in the book, seems blind to the differentiation in subject positions in the world. Rettberg is clearly well-informed about how machine vision can be harmful to different subject positions. Yet in the book’s focus on hope, it sometimes ends up rhetorically reducing that aspect of machine vision. This, I think, has to do with the ‘we’ that the book operates with. The book’s subtitle promises an investigation of “how algorithms are changing the way we see the world” (my italicization), but does it truly grapple with who this ‘we’ refers to? Rettberg states that the ‘we’ is meant as an inclusive ‘we’ – a ‘we’ that sometimes contains her and the reader, sometimes “a group of humans, sometimes humans and machines together” (23). She writes that it is not meant as a universal or oppressive ‘we’. Still, at points in the book, especially towards the end, there is a sense of ‘we’ that very much pertains to subject positions like Rettberg’s. I will unfold this in the remaining part of this review. For clarity, I will begin with a short outline of the book’s chapters.
Chapter Walk-Through
Rettberg is primarily interested in understanding the relationship between humans and technologies, as well as the context from which this relationship arises (24). She focuses on relationships rather than the technologies themselves to escape human-machine binaries and a “technology as tool”-thinking. Drawing on posthuman and technofeminist theory (especially assemblage theory), Rettberg seeks to move beyond a human-centric understanding of technologies as passive tools. Key philosophers in her book include Giles Deleuze and Félix Guattari, Donna Haraway, and N. Katherine Hayles. Rettberg analyzes a vast and diverse set of artistic and activist engagements, alongside her personal experiences with machine vision. She continually contextualizes her analyses historically and culturally.
This framework allows for an understanding of machine vision technologies as “companions” that share agency with humans in “more-than-human assemblages” (27). However, the book is perhaps not as posthuman as it seems at first sight. The title and table of contents clearly exemplify this. It is about how human perception is affected by machine vision. All the chapter titles emphasize human vision – what humans see with machines (whether they are seen, see more, see differently, see everything, or see less). Rettberg’s methodology centers around the human subject. “Machine vision situations” – concrete situations where humans and machine vision see and act together (12) – are analyzed from the specific position of Rettberg herself with a strong sensitivity to emotions and affect involved in personal experience.
In the introduction, Rettberg presents a rather broad definition of machine vision: “the registration, analysis and representation of visual information by machines and algorithms.” (3) This definition encompasses technologies like satellites, drones, facial recognition systems, surveillance cameras, smartphones with front-facing cameras, selfie filters, and technologies used for observing waves beyond human perception (night vision goggles, radar, ultrasound, LIDAR). It also includes what Rettberg calls “pre-algorithmic” forms of machine vision (26) such as water (reflection), fire (light), and stone (polished stone mirrors), glass lenses, microscopes, telescopes, painting, and photography. Crucially, Rettberg views machine vision not as these pre-algorithmic and algorithmic technologies in themselves, but as the relationship between these technologies and the humans who use them, as well as the context from which this relationship arises.
In the first chapter, “Seeing More,” the main argument is that although machine vision has been anthropocentrically framed as a passive tool for “seeing more,” it should be understood as a “more-than-human” assemblage. Rettberg discusses central examples, problematics, and debates in the history of vision, representation, and post-representation – touching on the history of linear perspective and photography, issues related to objectivity and truth, and the shift from representational to operational images – to demonstrate how machine vision technologies have always affected and altered their users. This argument takes as its starting point the extreme poles of techno-determinism (represented by Vilém Flusser’s apparatus theory) and the more techno-optimist social constructivism (represented by Judy Wajcman) (28-33). Of course, one could challenge Rettberg’s characterization of Flusser’s theory as techno-determinist, but I will set that aside for now.
These poles are explored in chapter two, “Seeing Differently,” through case studies of Flusser’s theory of the camera and an art film, titled kino-eye, created by the Soviet Kinoks collective. This chapter also analyzes a concrete and very personal assemblage: the co-vision of Rettberg and her family with the robot vacuum cleaner, Alfred. The Alfred analysis sidesteps the extremes of techno-optimism and determinism by showing how being part of the Alfred assemblage allows for “more-than-human” ways of seeing and, through that, accessing “more-than-human” truths about the family’s private home and life. Drawing on philosophers like Jane Bennett, Rosi Braidotti, and Anna Tsing, Rettberg employs assemblage as a broad posthuman concept (33). An assemblage can be anything, really. It can include human agents, nonhuman agents (like technologies), tech companies, cultural contexts, emotions, geological materials, etc. To understand machine vision and how humans see with it, Rettberg argues, “we need to understand the whole assemblage” (28), meaning all its agents and their interrelations. Therefore, she proposes using the concept of cyborg vision to analyze how humans perceive the world together with machines (81). Machine vision can help situate humans in new ways, expanding perception beyond bodily limitations (52) and facilitating access to “more-than-human” truths about the world (160).
While Rettberg mostly views the “more-than-human” truths gleaned from the Alfred assemblage as lighthearted, chapter three, “Seeing everything,” unfolds the negative consequences of a far less humorous assemblage: the Oak Park assemblage – an American machine vision surveillance assemblage. She analyzes automated license plate readers (ALPRs) in Oak Park, where she lived with her family while serving as a visiting researcher at the University of Chicago. Starting from her husband’s assault at an Oak Park train station, Rettberg conducts a complex and impressive analysis of machine vision, examining Flock Safety surveillance cameras, Ring doorbells, the Neighbors app, emotions of fear, anxiety, and trust, local histories of racism, discrimination, and inequality, and activist practices as part of the assemblage. This analysis shows the strength of Rettberg’s use of assemblage theory.
Chapter four, “Being Seen,” analyzes three diverse case studies: selfie filters and biometric facial recognition, automated grocery shopping, and a fictional AI dictator named Thunderland in a young adult novel. It focuses on how machine vision “sees” humans with a normalizing algorithmic gaze. The analysis of an Amazon grocery store is particularly interesting. It portrays Amazon as an overarching assemblage that can see with its various machine vision “eyes” – including Amazon grocery stores, online shopping trails on Amazon.com, Amazon-owned Ring doorbells, etc. This analysis illustrates how implementing machine vision can eliminate points of social interaction, thereby amplifying emotional agents like fear in a society.
While chapter four ends by speculating on the risk of an oppressive, totalizing, and normative algorithmic gaze, chapter five, “Seeing Less,” cultivates more hope. It explores art-activist practices and movie heroes that exploit machine vision. Through Haraway’s figure of “the trickster”, Rettberg analyzes how machine vision technologies are not omnivoyant after all and how they can be evaded and fooled. Tricking is conceptualized as a way of reclaiming human agency within “more-than-human” assemblages (145).
In the conclusion, “Hope,” Rettberg ends on a rather optimistic note. Hope, she writes, lies exactly in the trickster – the playful rebel who explores new ways of seeing the world together with technologies (22). The book concludes with a list of the things Rettberg loves about machine vision. Although she also mentions reservations and concerns, she chooses to remain hopeful despite these. Finally, she directly addresses the reader, asking, “What kinds of assemblages do you want to participate in” (161, my italicization). This question leads me to my ambivalence towards the book’s overall argument, particularly regarding its conceptualization of agency.
The Issue of Agency
My ambivalence towards Rettberg’s overall argument centers on its conceptualization of agency. I am both sympathetic to and skeptical about how the book envisions agency.
On one hand, by maintaining a human-centric focus, the book effectively holds humans accountable. It calls humans into account. It views humans as agents within “more-than-human” assemblages who need to take responsibility for how their machines are developed and used. A posthuman approach that did not emphasize the human could risk under-theorizing human responsibility and agency. Rettberg’s book, however, leaves room for thinking agency. It gives way to the idea that we, as humans, not only need to take responsibility but also can take responsibility; that we have agency to change how technologies are developed, thereby, fostering hope for alternative, more sustainable, and ethical assemblages. One should not, I think, underestimate the value of this.
On the other hand, I find it problematic that the human often occurs in the singular – “our human point of view” (10), “the way we humans see” (15), “how we humans are affected” (15). In reality, this ‘we’ is more differentiated. That the analyses are seen through Rettberg’s specific eyes becomes problematic exactly when those eyes presume that they see for a common ‘we’. The analyses suggest that ‘we’ can use technologies in “good” or “bad” ways, controlling or liberating, hurtful or caring; that we can change our technologies and how we use them. For example, Rettberg writes:
“Using machine vision to see more allows us to observe the world from a different viewpoint than our own. Used well, this can help us become more aware of the universe around us, and more aware of each other’s different experiences of it. But these technologies can also be used to distance us more from other humans, whether by caricaturing them, as we saw with photography, or by creating a distance from them that makes it easier to kill or disregard them, as we saw with the thermal cameras.” (57)
Statements like this imply a unified ‘we’ that can choose whether or not to use technologies and whether to employ them in good or bad ways. But this might be a simplification that overlooks the fact that we do not always have a choice in participation, and that, importantly, we do not all possess the same possibilities of choosing if and how we participate in assemblages. This is a question of power and right. The right to look (Mirzoeff).
When Rettberg poses the question towards the end of the book about what assemblages ‘we’ want to participate in, it reduces the sensitivity towards the differentiation in subject positions otherwise present in the quote above and in analyses like that of the Oak Park assemblage. The sensitivity to the victims of caricature, of discrimination in facial recognition, the victims of drone killings. Do we truly have a choice? Do they? Can everyone choose what assemblages to engage with and how to use them? Who has the right to see with machines? Who has the option to participate, and who does not? The hopeful conclusion, which portrays us all as participants who can choose how to engage, lacks sensitivity towards the differentiation of subject positions in the world. One may enjoy playing with selfie filters, using a robot vacuum cleaner, unlocking a phone with FaceID (160), and role-playing a hacker in Watch Dogs (156) — as Rettberg writes she does — and these experiences may feel like agency, but they are not necessarily accessible to everyone. Not everyone has the option to adopt a trickster position.
Moreover, if one accepts this conceptualization of agency, what is the actual scope of it? I sympathize with Rettberg’s insistence on hope and alternatives, especially in contrast to succumbing to fears of a dystopian technological future. I, too, believe it is crucial to maintain a sense of human agency amid the pervasive growth of machine vision environments. However, I also think that one should be careful not to reduce agency to mere entertainment. While anxiety and fear can indeed be depoliticizing, as Rettberg suggests with Bennett in mind (159), I wonder if entertainment and fun—playing with selfie filters or engaging in other playful activities—may have a similar effect. Might these experiences create an illusion of agency that distracts from the need for real change? As I wrote in the beginning, I do not really think that Rettberg means that our agency is reduced to playing with selfie filters or playing games. I just think that, rhetorically, stating in the conclusion that she loves these things, reduces the book’s sensitivity towards the differentiation in subject positions. Put otherwise, I think that other analyses she conducts of agency in the context of machine vision – for example her analysis of the artist Richard Mosse’s reappropriation of infrared photography (55-56) or her analysis of the histories of racism at play in the Oak Park assemblage (chapter 3) – are more persuasive and much more nuanced.
Haraway’s concept of situatedness carries with it an imperative: to seek out other perspectives. For Haraway, the aim is never merely to account for one’s own situatedness. Instead, the aim is to do so in order to shift that very situatedness to allow for “at least” a double vision; preferably a multiplicity of visions (Haraway, “Situated Knowledges,” 589). Rettberg’s book conceptualizes a double vision from the beginning: a cyborg vision or co-vision of humans and machines. The book also has double vision in the sense that it does provide complex analyses of agency in the context of machine vision that operates with more complex uses of the ‘we’. But at points, and especially towards the end of the book, it seems to somehow leave the double vision a little bit. At these points, I would have wished it had ventured to stay in the complexity of its double vision; in a more complex understanding of agency and of the ‘we’.
Works Cited
Crawford, Kate and Trevor Paglen, “Excavating AI: The Politics of Images in Machine Learning Training Sets,” AI & Society Online (2021/06/08).
Gates, Kelly, Our Biometric Future: Facial Recognition Technology and the Culture of Surveillance, Critical Cultural Communication, New York, NY: New York University Press, 2011.
Haraway, Donna, “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective,” Feminist Studies 14, no. 3 (1988): 575-99.
Magnet, Shoshana Amielle, When Biometrics Fail: Gender, Race, and the Technology of Identity, Durham: Duke University Press, 2011.
Maurer, Kathrin and Anders Engberg-Pedersen (ed.): Visualizing War, Emotions, Technologies, Communities, New York and London: Routledge, Taylor & Francis Group, 2018.
Mirzoeff, Nicholas, The Right to Look: A Counterhistory of Visuality, Durham and London: Duke University Press, 2011.
Weizman, Eyal, The Least of All Possible Evils: Humanitarian Violence from Arendt to Gaza, London; New York: Verso, 2011.