2
Table of contents
Introduction p. 3
1. Theoretical framework and methodology p. 6
1.1 The Necessity for Cripistemology p. 6
1.2 Disability Studies and Crip Theory: The Social Model p. 6
1.3 Queer/Crip Theory p. 8
1.4 Cripistemology: Inhabiting the Clouds p. 9
1.5 Cripping Epistemology p. 10
1.6 Methodology p. 11
1.7 AI As Ideology: A Machine Learning World p. 12
1.8 Chapter overview p. 14
2. The Crip Cyborg p. 16
2.1 Crip Noise: Disabling the Cyborg p. 17
2.2 Blade Runner: Cyborg Anxiety and Supercrips p. 20
2.3 Medical Histories: The Datafied Body p. 23
2.4 The Crip Noise of Refusing Ideologies of Cure p. 25
3. Queer/Crip Temporalities: Narrative Structures of Illness p. 29 3.1 Temporal Structuring: Queer Theory Against Linearity p. 31 3.2 Queer/Crip Art of Failure: Failing Productivity, Failing Temporality p. 34 3.3 Algorithmic Temporality: Moving Forward p. 36
3.4 The Algorithmic Autistic p. 38
4. The Thinking Machine p. 41
4.1 Man as Machine p. 42
4.2 Machine as Man p. 44
4.3 The Autistic Thinking Machine p. 46
Conclusion p. 51
Bibliography p. 54
3 Introduction
I would rather be a cyborg than a goddes. Donna Haraway (2016).
Imagine a cold open space, where everything is made of steel and the only light source is a computer producing an endless stream of binary code, preferably in the familiar shade of toxic green on a black background. Submissive bodies are plugged in, flesh is exposed and body and machine have become fully entangled. We cannot see where one begins and the other ends. These and other ‘Wachoswki- isms’ no doubt come to mind when reading the title of this thesis, ‘AI Cripistemologies: Cripping The Thinking Machine’. The culture industry has been especially productive in producing images and stories that satisfy our desire for and interest in increasingly complex human-machine relations, bridging the gap between technology and body. The relation between technology, the body and embodiment has been much discussed in particular branches of feminist theory, beginning with Donna Haraway’s influential ‘A Cyborg Manifesto’ (2016).
Haraway’s cyborg is about ‘transgressed boundaries, potent fusions, and dangerous
possibilities’ (p. 14), but the disabled cyborg remains largely absent as the cyborg discussions tend to take a ‘fully functioning human and a fully functioning machine’ as their starting point (Quinlan and Bates, 2009, p. 51). This is ironic given that the particular hybridization of machine/human is often central to the lived experience of disabled individuals, especially those with physical and sensory impairments. The relation between disability and technology is one that is deeply nuanced, as technology simultaneously enables disabled individuals, while it also controls and surveils. I follow Donna Reeve’s (2012) call to look beyond technology ‘as ‘fixing’ impaired people in normative ways (and therefore to be rejected)’, but rather to approach it as a way to explore ‘the new ways of being in the world that emerge from living as cyborg’ (p. 107). In doing so, I use the term ‘crip bodyminds’, where crip refers to the reclamation of disability by disabled people. The use of the term crip is a political choice, one that I dedicate to my feminist crip co-conspirators:
But this is a word that has always been used to also connote the slowness of thoughts, as though the speed of thoughts could ever be clocked! The reclamation of the word crip, with its clipped sound, directly addresses the metaphor and the linguistic or rhetorical impact of the term. (Dolmage, 2014, as cited in Price, 2015)
With the use of the term bodyminds, rather than body, I refer to Margaret Price’s (2015) ‘crip politics of bodymind’ (p. 269). ‘Bodymind’ is a materialist feminist Disability Studies (DS) concept: ‘a socio- politically constituted and material entity that emerges through both structural (power- and violence- laden) contexts and also individual (specific) experience’ (p. 271). Its use refers to a lack in DS in engaging with questions of the mind, as the field tends to center around the physical and the sensory.
However, ‘bodymind’ means more than simply adding the mind to the existing framework; rather, it insists on ‘the inextricability of mind and body’ and takes seriously ‘how processes within our being
4
impact one another in such a way that the notion of a physical versus mental process is difficult’
(Schalk, 2018, p. 5).
Why then bring the question of crip embodiment to Artificial Intelligence? I argue that a crip bodymind approach to Artificial Intelligence (AI) is necessary as a way to unsettle and reopen our technological imagination, both specifically when it comes to AI and disability, but also more broadly regarding AI as an ideological and epistemological project. A cripistemology of AI and Machine Learning, which centers on crip perspectives, is a necessary intervention into the development of these technologies and their social impact. There has been an increased interest in the social and political implications of artificial intelligence, including issues such as bias and algorithmic fairness (Eubanks, 2019; Crawford, 2021; Kantayya, 2020; D’ignazio and Klein, 2020; Noble, 2018).
However, these important explorations have had little to say about the way that artificial intelligence is haunted by a particular ‘form of human’ or ‘humanness’; inspired by the epistemological work of Sylvia Wynter (2015) I approach this question from the perspective of ‘being human as praxis’, taking the human not as a stable category but as something that is constantly being made. To this I add that the human is also always created in relation to the technology with which we live and that which is currently being developed, and that this relation goes two ways.
A cripistemological approach to AI can be used in order to uncover AI’s cultural myths, by which I refer to the cultural discourse that surrounds the invention, development, uses, and
representations of these digital technologies. Myth here also refers to the work of French semiotician Roland Barthes and his writings on ‘modern mythologies’ (1957) as the dominant ideologies of his time. Following Simone Natale and Andrea Ballatore (2017), I use ‘myth’ as a way to discuss popular narratives and representations of technology, without necessarily focusing on whether a certain technology does or does not correspond to actual events. I do so because analyzing the kinds of narratives, and especially the kinds of narratives of our shared future and the future of what it means to be human, is already entangled in the development and goals of these technologies, regardless of whether or not they have already been made a functional reality.
Through a ‘machine cripistemology’, a term I use after Merri Lisa Johnson and Robert McRuer (2014), who link the term ‘cripistemology’ to Eve Sedgwick’s ‘epistemology of the closet’
(1990), I explore the intersections between crip theory, AI, and its cultural myths. I argue that this approach has the potential to expose the technoscientific production and naturalization of ableism and sanism, which I understand as the structural discrimination and social prejudice against disabled people and those with a mental disorder or cognitive impairment. Bringing in a host of other theories, including the politics of cure, queer theory and queer temporality, I write against existing research on Artificial Intelligence and disability that centers only on the potentials of technology for diagnosis and cure, and thus seeks to normalize, control and surveil the disabled bodymind. Such narratives
generalize the experience of disability with incompleteness and the desire of cure. I propose to think from the place of disability as an epistemic project, the relevance of which reaches beyond the
5
disabled subject alone into the broader field of media studies. Disability remains largely absent from media studies (Ellcessor and Krikpatrick 2019), while a disabled perspective provides a means to decenter the ‘normal’ user. Including disability has important consequences for the study of ‘media texts, industrial practices, social relations, media policies, modes of reception, and the design of technologies and spaces’ (Ellcessor and Krikpatrick, 2019, p. 140). It should therefore not only be of interest to those working with disability, but also to a broader audience interested in decentering hegemonic user positions within media studies.
6
1. Theoretical framework and methodology
1.1 The Necessity for Cripistemology
Reading back my own work, especially what I wrote early 2019 after the completion of my Honours research thesis titled ‘Doing disability studies embodied: Nothing about us without us’, I sense a deep yearning for other ways of (un)knowing my own disabilities and the feeling of not being included in existing models of disability. Writing at a period in my life where I was struggling with my disability identity and the reality of chronic pain, I wrote: ‘How do you criticize cure when cure is what you hold on to in the hope for an alternative? All the accessibility in the world will not make my pain go away. All the social awareness in the world will not get me out of bed.’ Knowing what I know now, I sense a dissatisfaction with the way that common critical models of disability have left behind the body, and the sometimes rather unpleasant ways in which that body makes itself known. The constant pain and exhaustion did not leave me the option to forget about the body.
The desire to (un)know disability and to center on disabled embodiment, as well as all the political implications of this, are precisely where the relevance of cripistemology lies. Its focus on epistemology becomes a way of attending to the materiality of the body, or as Bill Hughes and Kevin Paterson (1997) argue, a way of dealing with the ‘disappearing body’ in disability theory. This disappearance had been essential for disability theory and disability activism particular in its early stages, when disability rights where still non-existent. But the insistence on the social context of disability served a necessary function; it became a means to rebel against the medical frameworks that locate disability solely in the (individual) body in the context of medical institutions.
Understanding where cripistemology comes from then requires one to understand the inherent differences between Disability Studies (DS) and Crip Theory, two fields that have their own specific (intellectual) histories, as well as the role of the social model. While cripistemology may appear like a logical step in the process of understanding disability, from its first ‘official’ use (Johnson and McRuer, 2014) the term has existed ‘in response to’ a number of issues that have been raised for DS and as such the term does not come without some tensions. While the entirety of the institutional history of the term (which is mostly U.S. centric) is not relevant here, understanding some of these tensions is critical for understanding cripistemology.
1.2 Disability Studies and Crip Theory: The Social Model
Cripistemology has been a way to identify and address ‘the place of difficulty in disability studies’
(Johnson and McRuer, 2014, p. 251), though the plural ‘places of difficulty’ is more fitting here.
Among these places of difficulty is the issue of chronic pain, which I also raised, inspired by my own experience. Pain as a topic has long been avoided in disability studies, particularly in the social model
7
of disability. The social model was central to disability activism and disability rights movements during the 1970s and 1980s. For disability activists at the time, being able to establish themselves as a united group required them to distinguish themselves from non-disabled people. The social model, which says that people become disabled not so much by physical impairment but are rather disabled by society through inaccessible buildings and ableist attitudes etc., provided a basis ‘for emancipatory politics but not for an emancipatory politics of identity’ (Hughes & Paterson, 1997, p. 337). In essence, it was a deeply assimilationist policy intervention, with the goal being access to neoliberal institutions, such as schools and universities.
The way that pain as a lived reality of disability complicates the social model has been established by the work of crip theorist Allison Kafer in Feminist, Queer, Crip (2013). Arguing for the inclusion of pain and fatigue, Kafer writes, ‘Focusing exclusively on disabling barriers, as a strict social model seems to do, renders pain and fatigue irrelevant to the project of disability politics’ (p.
7). Kafer also problematizes the distinction that the social model creates between impairment as the physical aspect, and disability as the social aspect. Kafer argues that this creates a false dichotomy:
Although I agree that we need to attend to the social, asserting a sharp divide between impairment and disability fails to recognize that both impairment and disability are social;
simply trying to determine what constitutes impairment makes clear that impairment doesn’t exist apart from social meanings and understandings. (Kafer, 2013, p. 7)
Kafer’s more post-structuralist approach therefore rejects the notion that impairment and disability can be easily separated, or that the social can be contained within a single category. Similar critiques of the impairment/disability dichotomy come from post-colonial DS scholars, such as Helen
Meekosha (2011) and Jasbir Puar (2017), who have pointed out that part of this dichotomy stems from a deeply Western understanding of disability. Particularly the ‘problem’ of impairment has unraveled narratives of disability pride that are unable to account for the ‘production of impairment in the global South [emphasis in original]’ (Meekosha, 2011, p. 668). Meekosha and Puar (2017) discuss disability and debility in the context of colonial violence, (neo)colonialism, exploitation and how these produce disability and debility. These discussions reject a clear separation between the body, the social, and the political, as the disablement of Palestinians for example, as Puar discusses, cannot be separated from their political reality.
The post-social model of DS and crip theory have expanded and further theorized the notion of ‘social environment’ to broader questions of temporality and spatiality in an attempt to re-include the body. Including pain and fatigue, then, is a way to attend to the reality of (some) disabled bodies.
These interdisciplinary approaches provide a way to critique the ‘untenable separation between body and culture, impairment and disability’ (Hughes & Paterson, 1997, p. 326). Scholars like Kafer, Meekosha and Puar reject a clear separation between the body, the social, and the political as a means to center ‘becoming disabled’ over ‘being disabled’ as a process and assemblage, which is relevant also for cripistemology. However, it is important to remember that despite the number of criticisms
8
that can be raised with the social model and the central place it has taken in DS, it was also necessary in a world where a medical model of disability was practically the only lens through which disability was made sense of. The medical model located disability solely within the institutional context of the hospital, as an individual physical condition which reduced ‘the complex problems of disabled people to issues of medical prevention, cure, or rehabilitation’ (Shakespeare, 2010, p. 197).
1.3 Queer/Crip Theory
The theoretical tradition out of which cripistemology was born in part is related to the issues raised here with the social model of disability and with DS . During the 1980s and 1990s scholars started to integrate knowledge from the field of queer theory, starting with ‘queer disability studies’. Queer disability studies appeared at the intersection of disability, race, gender, and sexuality. Out of queer disability studies the term ‘crip theory’ was proposed. Originally it appeared in the work of Carrie Sandahl (2003) and was subsequently further developed by Robert McRuer (2006). Of course the usage of ‘crip’, in an attempt to reclaim this slur, mirrors what queer theory has done for ‘queer’.
An essential part of crip theory is to experiment with letting concepts travel and flow from queer to crip. Sandahl argues in ‘Queering the Crip Or Cripping the Queer?’ (2003) for the productive combination of crip and queer ‘to pinpoint where queerness and “cripdom” intersect, separate, and coincide’ (p. 25). For this reason, both Sandhal and McRuer disregard boundaries between crip and queer. One such concept is that of ‘compulsory-ablebodiedness’, which McRuer mirrors onto
‘compulsory heterosexuality’, arguing for a similarity between the two: ‘Like compulsory
heterosexuality, then, compulsory able-bodiedness functions by covering over, with the appearance of choice, a system in which there actually is no choice’ (McRuer, 2006, p. 8). It is within this system without choice that the cultural inability to imagine a world that willfully includes disability is reflected. Instead, disabled bodyminds exist as objects to be cured because they lack normalcy: ‘A system of compulsory able-bodiedness repeatedly demands that people with disabilities embody for others an affirmative answer to the unspoken question, “Yes, but in the end, wouldn’t you rather be more like me?”’ (McRuer, 2006, p. 9).
Cripistemology itself is also an example of this kind of conceptual or theoretical
interdisciplinary freedom. The term was inspired by the work of Eve Sedgwick, whose work was foundational for queer theory. Cripistemology follows Eve Sedgwick’s Epistemology of the Closet (1990) in which she writes, ‘many of the major nodes of thought and knowledge in the twentieth- century Western culture as a whole are structured – indeed, fractured – by a chronic, now endemic crisis of homo/heterosexual definition’ (p. 1). Cripistemology concerns crip ways of knowing, (un)knowing disability, and knowing from disability, while also refusing the idea that disability can ever be fully known. Johnson and McRuer update Sedgwick’s claim: ‘thought and knowledge in twenty-first-century Western culture as a whole is structured – indeed fractured – by an endemic crisis of ability and disability’ (p. 131).
9 1.4 Cripistemology: Inhabiting the Clouds
Cripistemology, as the crip approach to knowing, became a part of academic knowledge in its own crip way. After three years of ‘traveling through bar conversations, texts, and status updates’ (Johnson and McRuer, 2014, p. 130), residing in ‘the backwoods and branch campuses of disability and queer theory’ (Johnson and McRuer, 2014, p. 128), the term officially entered academia at
‘Cripistemologies: The Conference’, a conference held at the Center for the Study of Gender and Sexuality at New York University, as well as a double special issue of the Journal of Literary &
Cultural Disability, edited by Merri Lisa Johnson and Robert McRuer. As a relatively recent addition to Disability Studies and crip theory, cripistemology should be read as a response to what Johnson and McRuer (2014) call the ‘crip economy’: ‘Neoliberal disability epistemologies are highly lucrative – this much we know for sure. Disability identity is now part of capitalism’s array of target markets’ (p.
128).
Cripistemology sets out a complicated project. Attending to knowledge produced from the position of disability, while also questioning the validity of categorical distinctions and avoiding essentialist understandings of crip knowledge, requires an understanding that cripistemology, like all frameworks of disability, cannot and should not function as a catch-all term for the disabled
experience. Part of the project of cripistemology is acknowledging this, despite the fact that disability (rights) movements have ‘always desired better knowledge about disability’ in a world ‘incapable of valuing disability or recognizing the value generated by disability’ (Johnson and McRuer, 2014, p.
137).
The tension that results from this, between the risk of essentializing disability and discussing
‘crip ways of knowing’, is also present in the “Proliferating Cripistemologies” (2014) virtual roundtable held by Johnson and McRuer, in which Emma Kivisild asks, ‘Cripistemologies are epistemologies of slipperiness and clouds of meaning. So even though experience is not necessarily a way to authentic knowledge, is it a way to slide at the edges, inhabit the clouds?’ (p. 151).
Cripistemology is then a part of the process of recognizing that ways of knowing are inherently unstable, cannot be teased out and singularized, even if we desire to fully know disability. In fact, any theory that claims to provide a fully developed all-inclusive understanding of disability risks
becoming part of the crip economy.
Perhaps there are also things to gain from a certain mystified aura around disability, certainly when it comes to troubling the centrality of cure, which Eli Clare (2017) argues is essential to the politics of disability. For Clare, the politics of cure refers to a hyper-individual understanding of disability which locates ‘the harm entirely within individual human bodyminds, operating as if each person were their own ecosystem’ (p. 15). The centrality of cure to disability ultimately ‘seeks to return what is damaged to that former state of being’ (p. 15). Cure is of course almost always
preceded by diagnosis. To cure requires to know where to look, and it is the ‘knowing where to look’
10
that cripistemology problematizes in asking; can we ever fully know disability? Cripistemology recognizes that instability is not inherent only to the disabled body but to all bodies. Cripistemology is a way to ‘yearn’ (in bell hook’s terms) ‘for other ways of being and being in common, other ways of speaking and of shaping political, artistic, and scholarly solidarity’ (Johnson and McRuer, 2014, p.
248). Borrowing the words of Kivisild, a cripistemological way of knowing requires us to ‘inhabit the clouds’.
Turning to AI and ML technologies through a cripistemological lens, then, the type of
‘machine cripistemology’ that I envision is a necessary intervention in these technologies, and one that centers on crip perspectives. Whereas research that combines AI/ML and disability often centers on the potentials for cure, and therefore departs from the perspective of compulsory abled-bodiedness in which the crip body exists merely to be cured, a machine cripistemology has the potential to expose the technoscientific naturalization of ableist and sanist norms and cultural myths. In turn, uncovering human-machine relations from a crip perspective also works the other way around, in the sense that it also has implications for disability. Both disability and AI/ML share a complex temporality in the idea of ‘futurity’. In relation to disability ‘the future’ is often deployed normatively in the service of able- bodiedness and able-mindedness (Kafer, 2013). AI/ML often rely on grand narratives of tech-heavy futures. What then would the future of AI and ML look like, if we were able to imagine a world with disability in it?
1.5 Cripping Epistemology
The centrality of epistemology as the study of knowledge in this thesis is part of a larger and
longstanding tradition of questioning the boundaries of knowledge within feminism, queer theory and disability studies/crip theory and other radical epistemologies. This debate goes beyond feminist epistemology as a ‘distinctive female way of knowing’ (Longino, 1994) or a ‘crip way of knowing’, into what Haraway (1988) has famously called ‘situated knowledges’. Situated knowledge was born out of a dissatisfaction with the duality of objectivity-relativism as it existed in Western science, which claimed objectivity as impartiality and a ‘view from above, from nowhere’ (Haraway, 1988, p.
589). This knowledge model was attached to a very specific position (male, white, heterosexual, human), but transformed this specific position into one that is ‘universal’: what Haraway calls ‘the god trick’. Similarly, in the black feminist tradition scholars have been interested n standpoint epistemology as subjects speaking from a location and naming that location.
Situated knowledge serves not only as ‘a strong tool’ (p. 578) capable of preserving objectivity by being explicit about the context in which knowledge is created, but it also opens up further epistemological questions on the limits of knowing and objects of knowledge: ‘The codes of the world are not still, waiting only to be read. The world is not raw material for humanization; […]
the world encountered in knowledge projects is an active entity’ (p. 593). Such a worldview for
11
Haraway is one that colonizes, objectifies, and dominates, and that ultimately presumes that to know and to own are synonymous. Crip epistemology, then, does not mean arguing that disabled individuals
‘know’ disability (it rejects a binary notion of disability that separates disabled from non-disabled), or that the desire to fully ‘know’ disability is one that is productive. Instead, a crip epistemology
questions the presumption that cognition itself is the baseline for being human:
If conventional epistemologies always presume a subject who can know, a cripistemology will surely begin and end with a subject who knows merely that his or her ability is limited and that the body guarantees only the most fragile, temporary access to knowledge, to speech, to memory, and to connection. (Halberstam in Johnson and McRuer, 2014, 152)
This critique of epistemology is similar to that of Haraway, who points to the way that knowledge is produced from a specific standpoint or an embodied position. The type of epistemology that
Halberstam describes here requires a subject that is aware of its limitations, and embraces the ‘not knowing’ as much as ‘knowing’. Cripistemology thus not only refers to an epistemology of crip knowledge but also to the cripping of epistemology itself. Not only does a crip epistemology acknowledge the limits of knowing disability, as well as the problematics of desiring knowing disability, but it takes the disabled bodymind and the experience of disability and limitation to reflect on the production of knowledge more broadly. As Halberstam argues, the fragility of the body reminds one of the fragility and temporariness of knowledge itself.
1.6 Methodology
This thesis is conceptual rather than analytical and as such takes a methodological approach that allows categories like ‘disability’ and ‘AI’ to be treated as concepts that are constantly becoming and shifting. This is a type of methodology that has been explored by feminist scholars such as Rosi Braidotti and Griet Roets (2012), who, inspired by Gilles Deleuze and Felix Guattari (1980), argue that ‘bodies and subjects are socially created in the affirmative actualization of the encounter between subjects, entities and forces which mutually affect and exchange parts of each other’ (p. 166).
Braidotti and Roets refer to this type of methodology as one that is nomadic: ‘retelling, reconfiguring, and revisiting a concept, phenomenon, event, or location from different angles’ (p. 168). Revisiting AI through the lens of disability and disability through the lens of AI opens up the meaningful potential for analysis of both these concepts. In ‘Nomadism: Against Methodological Nationalism’, Braidotti (2010) establishes a nomadic philosophy of the subject as the basis for a methodology: ‘a collectively assembled, externally-related and multilayered subject that acts in a time-continuum clashes frontally with the established view of the European subject of knowledge’ (p. 409). Such a framework of subjectivity, that both ‘aims at achieving epistemological and political accountability’ while
‘unveiling the power locations which one inevitably inhabits as the site of one’s subject-position’ (p.
409), is fitting given the radical epistemological project of cripistemology. This is precisely why a
12
cripistemological approach to AI, a field that tends to claim objectivity, is one that is productive and necessary.
Rejecting a universalist perspective on disability/AI and fixed identarian categories, while taking inspiration from radical feminist epistemologies and situated knowledge, brings its own set of unique challenges. As (un)knowing disability implies that the categorical borders between disability and ability are not to be trusted, and are situated in time and place, and that we must reject a singular narrative of disability and the disabled experience: How to justify the call for crip knowledge while also arguing that the category of ‘crip’ or ‘disabled’ is inherently unstable? The ongoing project of questioning the boundaries of disability and ability indeed appears to be in direct conflict with the appeal of situated crip knowledge. At the same time, I argue that this question does not necessarily need a clear universal answer because it is precisely within this friction between the universal and the personal that new knowledge is produced.
As Braidotti (2010) argues, rejecting universalism while stressing situatedness has both ethical and methodological consequences requiring ‘a specific form of accountability’: ‘to rethink the interconnection between the self and society in an accountable manner’ (p. 410). The answer certainly is not the be found in ‘methodological nationalism’ but rather in replacing ‘linearity with a more rhizomatic and dynamic style of thinking’, the method of ‘creative repetitions’ (p. 412). Being ‘in opposition to’ should then not be read as a state of negativity, but rather as part of a creative process:
‘a shift of paradigm towards a positive appraisal of differences, multiplicity and complexity not as an end in themselves but as steps in the process of recomposition of the coordinates of subjectivity’ (p.
414). Opposing the ideology of cure in AI then does not end in simply rejecting it; it is part of the creative process of cripistemology, or returning to Haraway (2016): ‘staying with the trouble’ as a means to explore complexity.
1.7 AI as Ideology: A Machine Learning World
AI research is a hot topic, as proven by the €8 million grant that was awarded in March 2021 to Dutch universities (including the UvA) and venture capitalists to ‘accelerate the application and marketing of innovations in the field of Artificial Intelligence (AI)’1. AI, or the promise of AI, has become a fully- fledged economy within what Nick Srnicek (2018) calls a ‘data-centric economy’ or ‘platform
economy’ (p. 152). Or in other words: AI is big business. The influence that AI has managed to obtain not only has implications for the future of technology and governance, but the promise of AI also has profound epistemological consequences for how we consider knowledge, data, and the way these are processed. The epistemology that AI proposes relies on three assumptions, as Paola Ricaurte (2019) argues: ‘(1) data reflects reality, (2) data analysis generates the most valuable and accurate
1 https://www.uva.nl/en/content/news/news/2021/03/millions-for-ai-knowledge-and-investment- consortium.html?origin=WjFV1JMkTIeZv24DcLlDsg
13
knowledge, and (3) the results of data processing can be used to make better decisions about the world’ (p. 351). These assumptions have a large impact on the technological imagination of AI for disability, including in the centering of data over lived experience.
AI, or more specifically Machine Learning (ML), is essentially the science of statistical mapping, and as such it relies on the presumption that it possible to translate reality into data points that can be placed on a map. AI is then also the science of categorization and classification, the main
‘activities’ of an algorithm. While AI and ML are terms that are often used interchangeably, AI is the broadest way to think about advanced computer intelligence while ML refers more specifically to a particular process in which machines take data and ‘learn’ from this data (Reese, 2017). ‘Learning’ is followed by either pattern recognition, otherwise referred to as classification, or pattern generation, also known as prediction. More recently breakthroughs in ML have combined classification and generation into Generative Adversarial Networks (GANs), which allows machines to generate entirely new content or ‘synthetic media’, including ‘deep fakes’ (Pasquinelli and Joler, 2020, p. 12).
As a term itself ML makes an interesting promise regarding the nature of ‘learning’. Matteo Pasquinelli and Vladan Joler argue that:
Machine learning learns nothing in the proper sense of the word, as a human does; machine learning simply maps a statistical distribution on numerical values and draws a mathematical function that hopefully approximates human comprehension. (Pasquinelli and Joler, 2020, p.
9; italics in original)
The promise of AI lies in the ability to reimagine collective knowledge as patterns. Anxieties over the normative power of AI go further back to the statistical power of the state, such as it appears in the work of Michel Foucault (2020), whose work has been deeply influential in establishing how such power has been essential in establishing social norms and the creation of the ‘normal’ and the
‘abnormal’: ‘AI easily extends the ‘power of normalisation’ of modern institutions, among others bureaucracy, medicine and statistics (…) that passes now into the hands of AI corporations’
(Pasquinelli and Joler, 2020, p. 13). What I will be looking at in the context of this thesis is not the technology behind AI/ML, but rather what Roberge et. al. (2021) have called the ‘cultural life of Machine Learning,’ which encompasses epistemological questions of knowledge production together with AI’s tendency to normalize, and how these take shape in AI’s imagined future(s). What happens before the algorithm is made? How is the proliferation of AI changing the way that knowledge is produced?
These speculative questions are not only common in critical AI studies; speculation itself is also a key element to the development of AI. The process of investment in the digital economy, or the
‘invention economy’ of Silicon Valley, is guided not so much by rational expectations as by
‘fictions’: ‘which can be defined as characterizations of the economic future which have no basis in empirical fact and seek to make the future the source for profit-making decisions’ (Appundurai and
14
Alexander, 2019, pp. 5-6). The focus on downstream value (speculative) rather than upstream value (technological inventions) has a large impact on the way that AI is being developed. This important speculative element leads me to a very broad definition of AI. Following critical AI scholar Kate Crawford (2021), I use AI to refer to ‘the massive industrial formation that includes politics, labor, culture, and capital’ (p. 9).
These questions suppose that the impact of AI on the way that we think about technology, knowledge production, but also embodiment, which can already be traced even before the machine has been made, the data has been gathered and the algorithm been written. I take inspiration from queer theory and futurity in arguing that whether the future will be what we predict it will be does not really matter, we are already living in it. As such my approach problematizes linear temporalities, as any good nomadic subject should do. In Braidotti’s nomadic methodology, time appears non-linear:
‘not Chronos, but Aion, the dynamic and internally contradictory or circular time of becoming’ (p.
412). The rhizomatic nomadic approach of AI is one that is essential in analyzing the kind of future AI imagines, and therefore the kind of present that we inhabit.
1.8 Chapter overview
The following chapters of this thesis will all expand on the theoretical framework that I have thus far established, together with an assemblage of theories I treat as primary sources. Each of these theories allow me to focus on different theoretical aspects of the disability/AI paradigm. The second
chapter focuses on the figure of what I call the ‘crip cyborg’, in which I contextualize Haraway’s cyborg figure and analyze the way that disability as a critical category has been left out of this conversation. Bringing in the work of authors invested in crip technoscience, a small subfield of disability theorists inspired by cyberfeminism, as well as an example of the crip cyborg in sports, I propose the figure of the crip cyborg as ‘a way out of the maze of dualism in which we have explained our bodies and our tools to ourselves’ (Haraway, 2016, p. 67). If indeed the dream of the cyborg is ‘not of a common language’ but instead ‘a powerful infidel heteroglossia’ (Haraway, 2016, p. 68), then it is about time that disability is included.
The third chapter focuses on queer/crip temporalities and the narrative structuring of illness and disability as an enforcement of the linear temporality that the phrase ‘get well soon’ implies. I develop this argument in relation to algorithmic prediction, specifically in the case of Autism
Spectrum Disorder (ASD), which I argue diagnoses autism as a condition of ‘unrelationality’. Instead, inspired by the work of Erin Manning I propose an autistic form of relationality. I argue that there exists a similarity between the algorithm and autism as what I call the ‘algorithmic autistic’, a unique kind of temporality of the feedback loop. In this chapter I bring in the work of disability writer Eli Clare (2017), who establishes a politics of ‘anti-cure’, as well as work of queer theorists Lee Edelman, Jack Halberstam, and Elizabeth Freeman. In doing so, I question the linear narrative of illness as being one from diagnosis to cure. I argue for the similarities between crip and queer in this regard,
15
because it is particularly within queer theory that authors have problematized linearity from the perspective of the heterosexual family and the child (Edelman, 2004), and have begun to establish a notion of queer temporality (Freeman, 2011). Especially the work of Halberstam (2011) on The Queer Art of Failure as a productive force of refusal provides a much-needed critique of the rush to produce AI for diagnosis, with no regard for the desirability and the social, political, and personal
consequences of diagnosis. Taking my cue from the queer art of failing, I propose a crip art of failing in discussing the possibility of crip futurism(s).
The fourth and final chapter explores the various metaphors in the space of computation and the body: man as machine, machine as man, and what I call the autistic ‘thinking machine’. I place the discussion of these metaphors in the context of Artificial General Intelligence (AGI), a specific subfield of AI preoccupied with the notion of machinic intelligence and human-machine relations. I argue that AGI fetishizes intelligence, a concept that is deeply problematic from a socio-historical perspective. In this context the ideal of the thinking machine, then, or the human without all the
‘useless’ aspects of being human, returns throughout history as the cultural obsession with machines passing for humans, a common trope in the science-fiction genre. However, at the same time this view of machines also takes the form of the human as machine, boiling down all bodily processes to pure scientific, schematized fact. Nonetheless, the thinking machine also functions as a particularly poignant autism trope, as both scientific and non-scientific discourse tends to imagine autistics as mechanistic beings devoid of empathy and feeling. My approach in this chapter is inspired by Melanie Yergeau’s work in Authoring Autism (2018) and what she calls ‘autistic rhetorics’. Writing this chapter as an autistic person, I follow Yergeau’s call to ‘queer the motifs, structures, modes, and commonplaces’, whereby ‘to author autistically is to author queerly and contrarily’ (p. 6).
16
2. The Crip Cyborg
Donna Haraway’s whirlwind “A Cyborg Manifesto” (2016), originally published in the Socialist Review, has made a lasting impact on the development of feminist post humanist theory. Her rejection of rigid boundaries, especially between human/animal and
human/machine, and her call for a politics of affinity later developed in
‘The Companion Species Manifesto’
(2016), function as poignant critiques of Western traditions of patriarchy,
colonialism, and essentialism, among others. In response to these traditions, Haraway proposes the cyborg myth, embracing the claim that ‘the cyborg is our ontology; it gives us our politics’ (2016, p.
7). The cyborg reworks nature and culture: ‘the one can no longer be the resource for appropriation or incorporation by the other’ (p. 9). To be cyborg is to be an assemblage, subverting ‘myriad organic wholes’ (p. 12). While a very influential theory, Haraway’s manifesto has been criticized by Disability Studies scholars for her lack of engagement with disability, which is notable given that disabled bodyminds have often encountered this nature/culture divide in a variety of complex ways.
While disability is shortly mentioned by Haraway in the Cyborg Manifesto, a real engagement with disability as a critical category is entirely missing.
Haraway’s Cyborg Manifesto was a foundational influence for cyberfeminists in the 1980s.
Running with Haraway’s ideas on assemblages, boundaries and organic wholes, Linda Dement’s CD- Rom artwork Cyberflesh Girlmonster (1995) explores the female grotesque and monstrous, where boundaries between body parts and our individual flesh have completely dissolved. For Cyberflesh Girlmonster Dement collected scans of
body parts and digital recordings of sounds of about thirty women during the Artists’ Week of the Adelaide Festival 1994. From this material, Dement created macabre monstrous bodies that were animated and made interactive.
Cyberflesh Girlmonster rejects the fleshy boundaries that separate us, and as such the work is a kind of cyborg politics that
Figure 1 Dement, L. (1995). Cyberflesh Girlmonster [CD-ROM].
Figure 2 Dement, L. (1995). Cyberflesh Girlmonster [CD-ROM].
17
reimagines female collectivity in Western patriarchy. I use the artwork as an example here of the boundary crossings and assemblages that a cyborg politics inspires. Dement’s entangled human bodies defy categorizations in their unrecognizability as she creates new bodyparts that refuse to be named yet still recognizably human in their fleshy tones. At the same time, Dement’s cyberflesh girlmonster could also be read as disabled, as I argue in this chapter it is through the rejection of the
‘fleshy boundaries’ that the disabled bodymind also defies categorization.
It is the concept of boundary crossing in relation to categorization and classification that is central to this chapter. Looking at the centrality of categorization and classification for the project of ML, I look at medical applications of AI and deep learning more specifically (what has been called
‘precision medicine’ in particular), and how the crip cyborg problematizes categorical boundaries. In order to do so, I first explore disability criticisms of Haraway’s work. Then, using Paralympian Oscar Pistorius as an example of what could be called a disabled cyborg, I look at the way that boundaries are reinforced once crip bodies risk dissolving the boundary between ability/disability. Pistorius and his sports career reveal a variety of cultural and social anxieties over bodies, technology, and the capacities of crip bodyminds.
I use cyborg theory in the context of AI and medicine because of Haraway’s visionary
description of communications and biology, one that aptly applies to AI and ML: as ‘the translation of the world into a problem of coding’, a search for a common language in which ‘all resistance to instrumental control disappears and all heterogeneity can be submitted to disassembly, reassembly, investment, and exchange’ (2016, p. 34). While the Cyborg Manifesto was written in the 1980s, many of the technological advancements (and their social implications) that it describes have only further intensified in what has been called ‘datafication’ or the ‘the datafied society’ (van Es and Schäfer, 2017). This process of datafication has had a large impact on the way that knowledge is produced and objectivity is constructed. Haraway’s work already provides a way of subverting this process: ‘The biggest threat to such power is interruption of communication’ (p. 34).
As a rejection of universalism, the cyborg rejects what Haraway calls the one code that translates all meaning perfectly’, which for her is ‘the central dogma of phallogocentrism’ (p. 57).
Cyborg politics then is one of disturbance, as it insists ‘on noise and advocates pollution, rejoicing in the illegitimate fusions of animal and machine’ (p. 57). Given the absence of disability in the Cyborg Manifesto, despite the fact that disability has much to teach us when it comes to dismantling
nature/culture and human/machine binaries, it is clear that for Haraway, the noise of cyborg politics is not a crip noise.
2.1 Crip Noise: Disabling the Cyborg
The potential for a crip noise within the politics of the cyborg, as well as a critique of Haraway’s inability to engage with disability, has been taken up recently in the field of feminist science and technology studies (STS) in what Aime Hamraie and Kelly Fritsch have called ‘crip technoscience’
18
(2019). As an amalgamation of crip theory and feminist STS, which both embrace situated
epistemologies, crip technoscience explicitly departs from disabled users’ and makers’ experience of technology, as a way to ‘slide at the edges’ (Johnson and McRuer, 2014, p. 151). Therefore, crip technoscience is a response to technology created for disabled people from an abled perspective; as such it follows the disability activist’s motto: ‘Nothing about us without us’. Crip Technoscience aims to reveal the way that ableism limits our technological imagination. The ‘technoscience’ in crip technoscience is to be understood critically, as it has been in feminist STS, to refer to ‘the productive and non-innocent entanglement of scientific knowing and technological making’ (Hamraie and Fritsch, 2019, p. 3). Crip technoscience can be seen as a way of speaking back to Haraway’s cyborg or to disturb the politics of the cyborg itself.
The cyborg then emerges in crip technoscience as a complicated figure, given that the imaginative potential of the cyborg has a complicated relationship with disability. The cyborg supposedly transgresses boundaries as a ‘cybernetic organism, a hybrid of machine and organism, a creature of social reality as well as a creature of fiction’ (Haraway, 1991, p. 149). Like Haraway’s politics of affinity, crip technoscience stresses the importance of ‘interdependence’ as political technology: ‘the weaving of relational circuits between bodies, environments, and tools’ (Haimraie and Fritsch, 2019, p. 12). Similar to the cripistemological critiques of the impairment/disability separation of the social model, crip technoscience refuses mainstream disability technoscience that views disability as an individual experience. This mainstream view has two consequences: firstly, the goal of technoscience (such as assistive technology) becomes only to encourage independence, and secondly it treats disability and technology as apolitical phenomena ‘rather than material-discursive entanglements that take shape through struggle, negotiation, and creativity’ (p. 12). This view clashes with that of cripistemology and other assemblage-based analyses of disability, such as the work of Puar (2017), who argues against an understanding of disability as a fixed state: ‘Disability […] exists in relation to assemblages of capacity and debility, modulated across historical time, geopolitical space, institutional mandates, and discursive regimes’ (p. xiv).
Disability and crip theorists outside of crip technoscience have also been rightly pointing out Haraway’s lack of engagement with disability, despite the obvious connections that could have been made between impairment and technology (such as in mobility aids). Among these critiques is that of Donna Reeve, who argues that cyborg theory is often limited ‘to consideration of how technology either restores functionality or normalizes the person with little discussion of the cultural/social implications of prosthetics, or of the lived experience of body and prosthetic’ (Reeve, 2012, p. 94).
Elizabeth Ellcessor (2017) reflects on the absence of disability from cyborg theory and cyberculture studies. She notes that disability is often only used as a metaphor ‘to illustrate abstract theories’ (p.
1763) rather than a critical category, functioning as what disability scholars David Mitchell and Sharon Snyder (2001) call a ‘narrative prothesis’.
19
Including disability as a critical category is then part of both crip technoscience and cripistemology, as both center the (technoscientific) production of disability, impairment, madness, illness, disease, debilitation, etc. As an epistemic project, the machine cripistemology that this thesis proposes is centered on the often intimate human-machine relations that disabled individuals
experience. As Laura Forlano (2017) puts it:
… my body is networked and dependent on a system of technologies that is fragile, vulnerable, and prone to breaking down … the disabled cyborg exists within, between, and out of sync with intimate infrastructures in which the world collapses onto the body and, at the same time, the body expands out into the world. (p. 3; italics in original)
As Forlano’s description of her personal experience shows, the cyborg that Haraway speaks of has already existed as a lived experience for crip bodyminds. However, Haraway’s Cyborg Manifesto is unable to imagine the critical potential of disability due to the fact that she (uncritically) departs from the place of a ‘fully functioning human and a fully functioning machine’ (Quinlan and Bates, 2009, p.
51). The potentials for a crip noise of cyborg politics lie in the way that disabled bodyminds dismantle precisely this human/machine binary, such as the work of Forlano does. ‘Cripping’ the cyborg means to take seriously the critical potential of disability, not as a narrative prothesis, but rather as a noise artifact. If using disability as a narrative prothesis means to use it only as metaphor rather than include disability’s material and political reality, embracing disability becomes a radical act in a system that aims to exclude the disabled bodymind through the ideology of cure, which I will return to later in this chapter. Instead of departing from machine + body, as Haraway does, crip noise notices that one does not exist without the other. Both the machine and the body are part of interpretative processes and assemblages. Haraway’s failure to note this lies at the basis of the exclusion of disability in her work, as there is no space for the disabled cyborg if we expect the human in the human-machine relation to be ‘fully functioning’ according to abled terms.
It is interesting that the critical potential of disability and disease emerges in another text by Haraway, ‘The biopolitics of postmodern bodies: constitutions of self in immune system discourse’, first published in 1989 and reprinted in Simians, Cyborgs, and Women (1991). In this essay she discusses the social power of biomedicine and biotechnology in the context of immune disorders. She looks at the language that is used in these fields and how these have been impacted by the ‘discovery’
language of popular science-fiction such as Star Trek: ‘Science as heroic request and as erotic technique that goes ever deeper’ (Haraway, 1991, p. 205). Within this system, disease has become ‘a subspecies of information malfunction or communications pathology; disease is a process of
misrecognition or transgression of the boundaries of a strategic assemblage called self’ (p. 212). As a point of failure within an otherwise functioning system of communication flow (the body), it is disease itself that becomes the external noise disrupting the internal information flows. Haraway’s analysis of immune disease research uncovers the narrative tropes that surround disease and disability, understanding disease as not simply as a given biological fact but rather as a socially situated
20
assemblage. Such narrative tropes include the use of war-like and colonial language, as well as science-fiction tropes. Disease ‘penetrates’ and ‘invades’ and we must ‘battle’ it, language that is also not uncommon during the current COVID-19 pandemic. In this framework boundaries between the healthy inner body and the diseased outside world must be protected at all cost. Such a framework fails to note the interconnectedness of all life forms.
2.2 Blade Runner: Cyborg Anxiety and Supercrips
Figure 3 Eisele, Johannes, AFP/Getty Images. (2012) ‘Oscar Pistorius competes in the London 2012 Olympic Games’
The boundaries between human and machine are perhaps most obvious when it comes to assistive technology. Especially in combination with sports, the cultural space where we measure all the things our bodies are capable of doing, where the abled ‘fit’ body and ‘health’ is fetishized, the disabled cyborg threatens the hierarchical ordering of disability and ability as we know it. Sport as the ‘central discursive site for producing and reproducing the modern body’ (Norman and Moola, 2011, p. 1269) is the space where human excellence is expected of the natural athletic human body and cheating by using ‘unnatural’ substances such as performance-enhancing drugs is ruled out by drug testing. The ideal of physical human perfection indeed has proven to be so powerful that it has been used in all kinds of nationalistic and fascist discourses. Consider Leni Riefenstahl’s Olympia (1938), which uses a fascist aesthetic that relies on the visual connection between bodies and political ideologies, in which very specific bodies are coded as a reflection of the ‘social characteristics of the society in which they are embedded’ (Weber and Black, 1999, p. 62)
What happens, then, once this ideal strong healthy body is challenged by a disabled cyborg that far surpasses it in terms of quantifiable capabilities? These kinds of questions are exactly what arose when South African runner Oscar Pistorius, using high-tech protheses called ‘Cheetahs’ (fig. 3), applied to run in the non-disabled Olympics. Pistorius’s carbon fiber blades resulted in his being referred to as the Blade Runner, a reference to the popular science-fiction film about untraceable
21
androids, which directly relates Pistorius to cyborgs. Pistorius became a Paralympic champion and was the tenth athlete that competed at both the Paralympic Games and the Olympic Games. Despite objections by the International Association of Athletics Federations (IAAF) Pistorius competed in the 2011 World Championships in Athletics, winning a world track medal. He also competed at the 2012 Summer Olympic Games and was the first double-leg amputee to do so. Throughout all this, Pistorius appeared as the inspirational disabled athlete, able to overcome the ‘tragedy’ of his ‘broken body’.
The controversy that arose from the IAAF’s initial plan to ban Pistorius’s participation sparked the interest of disability theorists, who, rather than looking at questions of ‘level playing fields’ or ‘technological advantage’, are interested in the way that the case of Pistorius shores up
‘boundaries of the humanist subject and the natural body’ (Norman and Moola, 2011, p. 1273). As Pistorius transgressed the boundaries between human/machine and ability/disability he revealed the porousness of these boundaries as unstable leaky categories. As such he is precisely an example of Haraway’s ‘dirty ontologisms’ or what I have called crip noise. Threatening to collapse the ‘very ontological and corporeal security of the humanist subject’ (p. 1274), the cultural response can only be one that reveals a deep-rooted anxiety over what disabled bodyminds are and are not supposed to be capable of. At the same time, we must also recognize that what allowed Pistorius to do so is extremely expensive assistive technology, one that is deeply exclusionary and is available only to a select group.
What produces excess must be contained, and so the transgressions of Pistorius’s cyborg body must be reframed according to scientific evidence of ‘unfair advantages’. For Pistorius this meant that his track performances were monitored and analyzed, and that he underwent a series of scientific tests which were meant to determine how much energy his protheses saved him (therefore giving him an advantage) compared to abled runners. In this regard scientific inquiry itself becomes a method of controlling crip noise. It should be noted here that the reliance on science as the thing that preserves the boundaries between human/machine and ability/disability functions precisely as Haraway’s ‘one code’, because its ontological status relies on the presumption that such boundaries cannot and should not be broken. On the same spectrum of what I call ‘disabled cyborg anxiety’ also lies the story of the
‘supercrip’, who functions as an inspirational narrative of the disabled individual who manages to defy all odds, proving that anything is possible if you are dedicated enough (Berger, 2008). The story of the supercrip reifies the conception of the ‘normal’ abled body, functioning as what disability activist Stella Young (2014) has called ‘inspiration porn’ aimed at motivating abled people.
Inspiration porn then limits disabled individuals in terms of what can be recognized as an
achievement, whereby achievements become only those things that compare to an abled norm. For Pistorius as supercrip this meant being able to scientifically proof the amount of exertion that it cost him to run compared to abled runners, whereby the scientific evidence both is aimed at normalizing his achievement in terms of effort and energy (similar to the abled body). At the same time the
22
supercrip status inherently places Pistorius in contrast to other disabled individuals unable to match his achievements which confirms to an abled norm.
What happened once Pistorius shot and killed his girlfriend, Reeva Steenkamp, is also relevant in the context of the supercrip image that he had built for himself. The complete shift in his public persona, from one of the supercrip to the man crying in the courtroom trying to appear sympathetic (an act that many felt was highly performative and fake), leads to a whole new set of disability narratives being placed on Pistorius. These are most notable in satirical cartoons that newspapers began to publish, which included one that refers to the argument made by Pistorius’s defense team on the ‘two Oscars’ (fig. 4). One the one hand, Pistorius was portrayed as a scared and vulnerable disabled person, whereas on the other hand his public image painted him as a strong person achieving beyond all expectations. These ‘two Oscars’ were said to be the result of the amputations of Pistorius’s limbs before his first birthday, a theory itself based on the notion of disability as trauma.
Figure 4 Zapiro. (2016) 'There are two Oscars'. The Times.
The example of Oscar Pistorius shows how assistive technology and prosthetics tend to focus on independence, individualism and performance according to abled standards. As such, it exists in contrast with understandings of disability that place the bodymind within a much larger assemblage and the commitment to interdependence that the ‘Crip Technoscience Manifesto’ (2019) by Aimi Hamraie and Kelly Fritsch proposes. Refusing a mainstream disability technoscience that essentializes disability to an individual experience, rather than a collective political one, Hamraie and Fritsch stress interdependence as a means to ‘understand how technoscience can simultaneously be entangled with global networks of domination and also provide opportunities for kinship and connection’ (p. 12).
This echoes a Deleuzian approach that stresses interconnectivity, symbioses and ‘becoming-bodies’
over solid bodies, especially if we are to understand assistive technology as a ‘tool’: ‘tools are inseparable form symbioses or amalgations defining a Nature-Society machinic assemblage … a society is defined by its amalgamations, not by its tools’ (Deleuze and Guattari, 2013, p. 90). Crip technoscience offers a Deleuzian politics of crip alliance and solidarity, in which tools exist within a
23
larger amalgamation of crip community in a system of compulsory able-bodiedness. This is decisively different from the politics of the supercrip, which celebrates individual achievement rather than collective achievements and solidarity.
2.3 Medical Histories: The Datafied Body
Moving beyond the figure of the (disabled) cyborg, in this section I will explore the broader topic of what Deborah Lupton (2018) calls the ‘datafied body’ in light of Machine Learning developments in medicine, including ‘precision medicine’ (Ferryman and Pitcan, 2018). In many ways these Machine Learning developments are a continuation of the datafied body, a term that Lupton uses to denote a particular understanding of disease as a form of information malfunction in the context of digital health technologies. According to Lupton, these technologies reduce the body to patterns, signs and signals to be predicted: ‘the body [is] a data repository’ (p. 2). Archived in the cloud, our bodies become mere transmitters of data. In ‘Biopolitics of Postmodern Bodies’ (1991) Haraway also refers to a similar phenomenon, one that she refers to as ‘technological holism’. Writing in the context of early cybernetic communications, she argues, ‘In the 1950s, biological bodies became technological communications systems’ (p. 219). Both the datafied body and technological holism then uncover a certain desire in medicine, whereby the body and its signals emerge as objects to be read by machine, the dataset/the computer becoming the ‘key’ to crack the body’s code. Not only have computation and communications served as important metaphors in medical literature; simultaneously the biological body has also served as a metaphor for computation and AI. Computers can be infected with viruses and we use diagnostic tools to figure out what is wrong. More broadly, Machine Learning has relied on a notion of ‘intelligence’ in comparison to the human mind.
This desire to ‘read’ the body in terms of a set of codes and data repositories has been at the basis of the development of precision medicine, defined by Kadije Ferryman and Mikaela Pitcan (2018) in a report for Data&Society as ‘the effort to collect, integrate, and analyze multiple sources of data in order to develop individualized insights about health and disease’ (p. 7). According to them precision medicine should be seen as an evolved form of personalized medicine, which is what Lupton’s work also focuses on. The goal in precision medicine then is to utilize the data generated by the datafied body in order to provide highly tailored healthcare. In the American context, Ferryman and Pitcan’s report shows how precision medicine is an emerging science linked to the Human Genome Project (HGP). HGP was an effort to map the entire sequence of human DNA started in 1990. Researchers believed that mapping human DNA would give new insights into health and disease, a project that is directly in conflict with disability politics. Disability scholars have been critical of the ‘technological progress’ myth that HGP propagates and the eugenic implications of a drive towards what Priscilla Wald (2000) calls ‘future perfect’.
Both HGP and precision medicine are examples of the ‘big data revolution’ in healthcare (Kayyali et.al., 2020). As such they fit in a larger focus in Western medicine on raw data, patterns,
24
signs and signals, providing an ideal space for Machine Learning systems that are by nature machines of classification. The process of classification and the relation between classifiers and the construction of social categories has been much studied by critical AI scholars, noting especially that categories are never neutral, and that the process of categorization is always deeply political. Classification has also been a much-studied topic within Disability Studies, where disability scholars such as Rosemarie Garland-Thomson (2008) shave analyzed the construction of disability in opposition to the ‘normal’
body, and how such difference becomes stigmatized. There is no category without there also being that which falls outside of its categorical boundaries; normalcy only exists by the grace of abnormality and vice versa. The datafied body then becomes a source for extractivism (of data) and as such relies on a framework of disability that reduces the disabled experience to individual impairment, which for disability scholars is a medical framework of disability rather than a social framework of disability.
However, developments such as HGP, genome editing and precision medicine go beyond simply extracting data from the body. Instead, their focus is on prediction and (early) intervention, working towards a particular future in which there seems to be less and less space for disability. The desire to develop new ‘ways of knowing’ the body then makes sense given AI’s status as an ‘atlas’ project, as Kate Crawford (2021) calls it. Similar to Haraway’s ‘one code’ Crawford dissects the ‘colonizing impulse’ in the AI field: ‘a desire not to create an atlas of the world but to be the atlas – the dominant way of seeing’ (p. 11; italics in original). A strong connection between knowing and power emerges, one that is only further enforced through the sheer speed by which classification has become possible.
Especially given the recent breakthroughs in independent Machine Learning that Luciana Parisi (2019) refers to as ‘the automation of automation’, as the ability of algorithms to create new
algorithms independently, even further transforms the speed by which machines are able to classify.
The exclusion of disability in the framework that I have thus far described is perhaps nowhere as obvious as in genetic engineering or ‘genome editing’. Genome editing intersects with AI, for example in what is called ‘CRISPR gene editing’, in which a technology is used that finds RNA- guided proteins in bacteria to edit an organism’s DNA, where an algorithm is used to predict the gene editing outcomes. The goal of AI in this case is then to aid genome editing through its ability to predict certain outcomes. AI then only serves to further develop what was already started in the field of genome editing. Given the impact of such developments on the survival of disability, it makes sense that the emergence of the disability rights movement (at least in the U.S.) coincided with the development of routine genetic intervention in reproduction (regardless of the intervention of AI).
Such interventions have prevented fetuses with certain genetic conditions from being born (Shakespeare, 1999). Leaving aside the question of whether this is a good or bad intervention, the development of such technologies and the possibilities for intervention themselves create a division between what we regard as ‘good’ or ‘bad’ genes. An example of these technologies is the quest to find the genetic ‘origins’ and neurological indicators of diagnoses such as Autism Spectrum Disorder (Keyes, 2020), a goal that autistic disability scholars such as Yergeau (2017) have contested. Much
25
remains unclear about the development or origins of autism. The desire to find a ‘single origin’ then overlooks the social and political context in which autism becomes a diagnosis in the first place.
Keyes et. al. (2021) refer to this as the idea of the ‘cerebral subject’, where ‘true’ identity is located within the brain:
AI is simply to the twenty-first century what ‘genetics’ was to the twentieth, or anthropometrics to the nineteenth – a tool of inquiry that, buoyed by both popular and academic understandings of it as an unprecedented and unimpeachable source of truth, is deployed to legitimize (rhetorically or methodologically) the same old schemes of division and disparity’ (p. 170).
For Keyes et. al. AI is merely a system of control, providing scientists with the one code to divide and conquer. The quest to isolate and remove autism’s ‘origins’ fails to note the real nuance of autistic experience, narrowing the diagnosis down to a source of all sorts of ‘problems’, rather than a rich and complicated experience like all human existence. It also takes away attention from issues of
accessibility, and measures that could be taken in order to make society more accessible for autistic individuals.
This drive towards getting ‘rid of’ disability has been criticized by disability scholars like Rosemarie Garland-Thomson (2012), who makes a case for the conservation of disability. Garland- Thomson reveals the contradictions between disability as universal to the human experience and the collective urge to refuse it, arguing that disability is wrongfully characterized as being only about disqualification: ‘disability is understood to disqualify us from access to the benefits and status of the properly human’ (p. 340). For her, the necessity of preserving them lies in ‘the prevalence,
persistence, and enduring sturdiness of disability rather than its fragility or vulnerability’ (p. 341). It is clear that currently Machine Learning interventions in healthcare cannot escape their historical ties to racism and eugenics in AI. Tracing the deeply problematic histories of early AI research, Sarah Myers West (2020) writes about Silicon Valley founder William Shockley, a physicist who spent decades promoting racist eugenic ideas about IQ differences and race, forced sterilization, and the ‘decline of civilization’. Shockley unsuccessfully attempted to persuade John McCarthy, who coined the term
‘Artificial Intelligence’, to join his cause. More recently, Jeffrey Epstein, who also promoted eugenics research, was a key funder of MIT’s Media Lab, providing money to support AI researcher Marvin Minsky and his work. In the next section I will look more closely at the way that eugenic tendencies in Machine Learning manifest itself in the form of a looming threat of cure in which disability remains at its core an absent presence.
2.4 The Crip Noise of Refusing Ideologies of Cure
As ‘engines of order’, a term used by Bernhard Rieder (2020) to refer to a wide variety of ordering techniques, algorithms and Machine Learning have been instrumental in transforming ‘static
