Not robots; Cyborgs — Furthering anti-ableist research in human-computer interaction

This theoretical essay builds on existing literature to draw out the consequences of dehumanizing and disseminating autism discourses within the ﬁeld of human-computer interaction (HCI). Focusing mainly on narratives in HCI that frame autistic people as or like machines, we explore how dominant constructions of autism in HCI work to normalize the ﬁeld’s complicity in violent autism intervention paradigms, despite HCI researchers’ well-meaning intentions. We work towards developing crip-cyborgs as an alternative framework for understanding autistic people (as opposed to computers or robots) and suggest crip technoscience as a framework for research based on this alternative understanding. In doing so, we hope to enroll misguided but well-intentioned researchers in dismantling anti-autistic ableism, both in and beyond HCI

Since Weizenbaum's days, like with broader cultural robot/autism tropes, mechanistic characterizations of autism in HCI have become diffuse, lacking an identifiable origin. The prevalence and staying-power of this trope in HCI are evident in numerous repetitions of the claim that autistics "have a natural affinity for technology" (Boitnott, 2012;Caria, et al., 2018;Dees, 2021;Elliott, 2017;Frauenberger, et al., 2015Frauenberger, et al., , 2013Frauenberger, et al., , 2012Guldberg, et al., 2017;Lavian and Alshech, 2015;Majeed, 2017;Shamsudin, et al., 2021;Menzies, 2011;Navedo, et al., 2019;Newbutt and Bradley, 2022;Valencia, et al., 2019) [5]. Few citations for this phrase (if any citation is given) provide supporting evidence. More often, the citations provided point to similar works. Those similar papers either justify technological interventions on assumptions of affinity themselves or find that a small-sample or preliminary study showed promising, if inconclusive, results about a specific technological intervention (e.g., Parsons, 2015;Williams, et al., 2002;Wojciechowski and Al-Musawi, 2017). Paradoxically, some citations point to papers suggesting technologically-mediated autism interventions lack foundational evidence (e.g., Fletcher-Watson, 2014). These citational practices raise questions about standards within HCI for sourcing claims, particularly those purporting an entire category of people possess some essential characteristic.
Chasing these "natural affinity" claims in HCI through a deep citational rabbit hole [6], we eventually found some theoretical support for technological affinities as an essentialized characteristic of autism Lesser, 1998b, 1998a). This theory, monotropism, refers to the autistic tendency to focus incredibly intently on a single interest (Murray, 2018). While not commonly cited directly, most "natural affinity" claims we were able to trace to a primary source pointed to this theory. Murray (2011) positions monotropism in contrast to polytropism, the non-autistic tendency to divide energy across many interests (Murray, 2012). Lesser (1998b, 1998a) build their theory of monotropism on evolutionary theory and mathematical modeling of atypical biological survival mechanisms. They argue that the theoretical distinction between autistics and non-autistics is natural -something biological and essential to autism. They say that, like autistic people, computers (at least in 1998) are "naturally monotropic," singly-focused machines (Murray and Lesser, 1998a). Over two decades, Murray promoted monotropism as a mathematically and biologically grounded framework for autistic difference (Murray, 2018;Murray and Lesser, 1998b). In their work, Lesser (1998b, 1998a) clearly explain that, by positioning monotropism as biological, they intend to frame autism as a natural variation of humanity rather than something apart from it. We want to clarify that we support this intention of normalizing autistic existence as part of a broad spectrum of what it means to be human. Murray and Lesser (1998a) present monotropism as a way for non-autistics to better understand and build relationships with autistic people, which is highly laudable. The problem arises when this research is uncritically used to support claims about technological affinities as an essential characteristic inherent to all autistic people. While essentializing claims about autistics and technology remain popular, monotropism remains purely theoretical, having some explanatory benefits over alternative frameworks but lacking meaningful supporting evidence or empirical work (Milton, 2017;Wood, 2021). Through its commitment to this claim about autistic technological affinities, or techno-autism, HCI perpetuates cultural narratives that position autistics as things, not people.

Robo-autism
Robot interventions for autism represent a flourishing niche within HCI (Spiel, et al., 2019), wherein research is often predicated upon claims that autistic people are more similar to and prefer machines over people. Suchman (2011) describes robots in human-robot interaction research as model (in)organisms. Model (in)organisms play on the concept of model organisms, which refers to using a particular creature as a proxy for scientific experiments trying to answer questions about other creatures or broader biological phenomena (see, e.g., Kohler, 1991). For Suchman (2011), robots provide a model that approximates humans -one that can pose as a substitute for people in experiments about human phenomena. Suchman (2011) explores how, through robot model (in)organism research, our understanding of human behavior and cognition is informed by tests on and with robots. As we gain a deeper understanding of humans, our understanding of robots (as human proxies) likewise adapts. Roboticists design robots to behave like humans. Humanoid robot design requires complex computer modeling intended to mimic human cognition. When robots respond to stimuli in expectedhumanlike -ways, researchers stake it to indicate their models' validity. Our understanding of human cognition, then, is reorganized around robotic algorithms. As robots behave more humanlike, our understanding of humans becomes more robot-like (Suchman, 2011). Understandings of humans and robots become recursive and co-constitutive (Suchman, 2011).
Robots and computers never quite match their human counterparts. There remain critical gaps between humans and robots (Mori, 1970), despite the ways our understanding of one is co-produced (see Jasanoff, 2004) with and by our understanding of the other (Suchman, 2011). Critical gaps between humans and computers reside in robots' peculiar ways of moving (Mori, 1970) and limited capacity for mutual intelligibility with human users (Suchman, 2006). The disjunctures between humans and robots are the same as those observed or alleged between humans and autistics (Baron-Cohen, 1995;Weizenbaum, 1983), leading to Weizenbaum's comparison of computers and autistics. More recently, Kaminka (2013) stated that nearly all robots are autistic. In understanding robots as autistic, it seems inevitable that notions of autistics (replete with natural affinities for technology) as robots follow. As we show next, such characterizations can harm autistic people in technologically mediated autism research.

Robot(ic) autism therapy
Socially assistive robot therapy represents an extension of the vast, pre-existing surveillance and control apparatuses (Roscigno, 2019; endemic to these interventional frameworks. Robots are deployed in robot-intervention research for autism to assist in or augment formalized autism intervention paradigms, such as applied behavioral analysis (ABA) (Spiel, et al., 2019;Williams and Gilbert, 2020). These interventions seek to rehabilitate autistic bodyminds, rendering them indistinguishable from their peers (Lovaas, 1987). As  explains, a key figure in ABA and autism, Ole Ivar Lovaas, is notorious for his involvement in the Feminine Boys Project. Through his involvement with the Feminine Boys Project, Lovaas created and refined gay conversion therapy using ABA. ABA for autism uses the same behaviorist interventions as conversion therapy.
More specifically, through continuous surveillance and highly regimented reward and punishment structures, ABA interventions for autism enforce cisgender, heterosexual, and ablenormative behavioral and communication norms on non-conforming bodyminds . ABA additionally targets verbalization over other forms of non-normative autistic communication (see, e.g., Alper, 2017;In my language, 2007;Zisk and Dalton, 2019), the reduction of autistic stereotypy (see, e.g., American Psychiatric Association, 2013), and, broadly, teaching the importance of compliance (see Sandoval-Norton, et al., 2019;Roscigno, 2019;. As Ne'eman (2021) explains, interventional paradigms that use "passing" as non-disabled as a primary outcome variable ignore the immediate stress that passing places on autistic people and the long-term health consequences of passing. Indeed, there appears to be a relationship between "camouflaging" [7] -the laborious process of hiding one's autism and behaving and communicating in non-autistic ways -and the disproportionately high rates of autistic suicide (Beck, et al., 2020;Cassidy, et al., 2020Cassidy, et al., , 2018Rose, 2018). Autistic survivors of behaviorist intervention therapies note that these therapies don't cure them. The therapies scare them into masking subconsciously to avoid the punishment they've been conditioned to expect if they behave and communicate like themselves [8]. In essence, prevailing behavioral autism intervention paradigms are profoundly harmful and predicated upon curative (Kim, 2017) and normative (Butler, 2004) violence.
As an overarching field, HCI often ostensibly values progress and social good (Bowers, 2018;Irani, et al., 2010;Lin and Lindtner, 2021;Pal, 2017). Yet, as Spiel, et al. (2019) and Williams and Gilbert (2020) show, a considerable proportion of autism research in HCI is complicit in behaviorist autism intervention paradigmsprojects of domination, control, and destruction (see Roscigno, 2019). Such research within a progress-driven field may appear paradoxical. However, several factors may help explain this vein of HCI research. First, discourses that objectify autistic people may obscure their needs and agency, allowing for interventional research framed in terms of progress for non-autistic research beneficiaries. Second, existing political economies within and outside of HCI may incentivize objectifying autism technology research and facilitating narratives thereof.

Robo-therapy, for the state
In a field where essentialist claims about the likeness of autistics to machines are naturalized, autistics are cast as inhuman. They are cast as unliving. They are framed as things. Raw materials to be acted upon (Broderick and Roscigno, 2021;Roscigno, 2019). If a typewriter breaks, you might repair it or throw it away. If a computer freezes or starts performing erratically, you might restart or ignore it and walk away. If autistics are essentially like computers and typewriters, if they are just things, then are their peculiar and problematized features similarly dismissible? Evidently so, at least some of the time, within HCI (see Williams and Gilbert, 2019;Ymous, et al., 2020). More often, though, as is evidenced by the proliferation of interventionist autism HCI research, autistics are subjects for repair.
As Lin and Lindtner (2021) explain, many HCI research activities are significantly motivated by ideals of progress, utility, and productivity. The field is driven by a sense of doing good -working towards (typically Western conceptions of) social progress (see Irani, et al., 2010;Lin and Lindtner, 2021;Pal, 2017). Yet, as Ymous and colleagues indicate, "what is allowed to be understood as 'doing good' is reliant upon entrenched sociocultural traditions of ableism" [9]. To render autistics less visibly autistic is an assumed good in HCI research supporting behavioral autism interventions. "Rehabilitation" of robo-autistics is an uninterrogated good, despite growing evidence of harm (Bascom, 2012;Beck, et al., 2020;Cassidy, et al., 2020Cassidy, et al., , 2018Rose, 2018;Unbound Books, n.d.). The unchallenged assumption about the goodness of ABA is not unique to HCI. Instead, it indicates an overlap between the ideals of HCI and ABA. There is a convenient fit between robo-autism discourses and ABA's positioning of pre-intervention autistics as not-quite-people, not yet deserving of rights (see Roscigno, 2019).
ABA shares HCI's commitments to the logics of progress and utility (Roscigno, 2019). As Roscigno (2019) explains, ABA and similar interventions can be thought of in terms of biophilanthropy. Biophilanthropy is a term coined by Schuller (2018) to explain the use of surveillance and control for the good of the state and economy but under the guise of charity. ABA interventions are nominally about helping autistic people move beyond the idiosyncrasies of autism and join the rest of society as productive and useful citizens (Roscigno, 2019). The notion that autistics must be rehabilitated -repaired -to have value within liberal capitalism perpetuates and justifies the violences of behavioral autism interventions (Roscigno, 2019). Further, various autism discourses (including those about technology) rhetorically distance autism from humanity, obfuscating the personhood of these biophilanthropic recipients while simultaneously normalizing the violences they endure as necessary to the well-being of the neoliberal state (McGuire, 2016;Puar, 2017;Roscigno, 2019;Rottier, et al., 2022). Within a neoliberal society, to be legible as human and deserving of citizenship and rights, autistic people must be made capable of productive economic participation (Mitchell and Snyder, 2015;Roscigno, 2019).
Robot research extends beyond behavioral autism interventions in HCI, purporting to help the individual for the good of economic and state interests in many other contexts, too. DiSalvo (2018) notes that design research often seeks to support individual behavior change in the face of economic and social crises. For example, in response to the horrific realities of global climate change, an interaction designer may be likelier to make an app to help individuals track their carbon production rather than acknowledging or addressing governmental policy and industrial practices. DiSalvo describes this trend toward behaviorism and technosolutionism as "the abdication of the issue in favor of the (purportedly) tractable problem" [10]. We do not intend to suggest that HCI researchers harbor malicious intent towards autistics through their involvement in behaviorist autism interventions. Instead, a flawed sense of for whom our research should do good and by whom research is led, paired with the naturalization and adaptation of objectifying autism discourses within broader society, creates the perfect storm for fallaciously thinking supporting ABA constitutes progress and social good. Structural conditions may preclude researchers and designers from arriving at alternative ideas of "doing good," such as working towards dismantling the structures that naturalize maiming (see Puar, 2017) of autistic children in the name of progress. However, while HCI researchers in this space likely make well-meaning errors (see Mankoff, et al., 2010) [11] and without intending to cause harm, they benefit from the political economies of autism medicalization and intervention, which may disincentivize them from adopting alternative research programs. Mallett and Runswick-Cole (2016) argue that autism has become a commodity, bought and sold by myriad actors, from governments, to companies, to non-profit organizations, and to academic researchers. Broderick and Roscigno (2021) frame this commodification of autism as the autism-industrial complex (AIC). Through its involvement in autism intervention, HCI is entangled in the diverse political economy of those who trade in autism. This marketplace is built upon the construction of autism as a distinct ontological category -a deficient category positioned as demanding remediation (Broderick and Roscigno, 2021;Mallett and Runswick-Cole, 2016). A diversity of diagnostic, educational, behavioral, and therapeutic professions exist in response to biomedical and dis-animating constructions of autism and comprise the AIC. Autism is their capital. Yet, paradoxically, their work often tries to convert the things they trade into people, able to participate in civic and economic life. Thus, as Broderick and Roscigno (2021) explain, the various arms of the AIC have an imperative to both perpetuate the classification of autism and autistic people as either broken people or inhuman things and ensure a continual supply of autistic raw materials. We see this in massive lobbying efforts, which have succeeded in legislation deeming ABA as the only intervention for autism coverable by insurance within the United States (Roscigno, 2019). When a child is diagnosed with autism, their parents are given pamphlets about ABA and told it is the only available option for their child. Thus, parents become enrolled in the AIC, creating additional advocates for the continuance and advancement of autism services. And, to complete the cycle, service providers can claim they are simply fulfilling market demand driven by parents of autistic children. Even when evidence emerges that institutionalized conceptions of autism are incomplete or incorrect, these incidents, too, become opportunities for the AIC.

Robo-therapy, for HCI researchers
Classically, autism is constructed as an essentially male disorder. Baron-Cohen (1995) went so far as to theorize autism as arising from "extreme male brains"that is, autism represents extreme versions of stereotypically male characteristics like rationality, mathematical and technical aptitude, and lower (relative to females) empathy. Such a theory rests on gender essentialism and harmful stereotypes. We suspect the male-coded construction of autism is likely implicated in the attachment of technological affinities as an essentialized characteristic of autism and the readiness with which techno-autism discourses were adopted. In recent years, various facets of the AIC have had to reckon with the emergence of so-called "female-autism" as a clinical construct. The construction of "female-autism" exemplifies the AIC's co-optation of evidence incompatible with dominant autism narratives (Moore, 2021). As Moore (2021) explains, attempts to codify "femaleautism" as a distinct autism phenotype expand clinical definitions of autism to include gender-essentialist constructions of girls and women, alongside those of the "extreme male brain." Moore (2021) further articulates that by widening the clinical criteria for autism to include girls and women, including those who present as cisgender, rather than as particularly (extreme) "male" in their demeanor and interests, new markets are created. By potentially doubling the population from which clinicians can legitimately sample for autistic capital, the AIC grows.
Through its adoption of essentializing claims about autism and technology, many autism technology researchers are swept up by and complicit in the AIC. In adopting mechanical and robotic constructions of autism as a thing, HCI autism researchers provide the appearance of scientific legitimacy to objectifying claims that sustain the helping professions. Too, in developing technological interventions for techno-autistics, they create and expand new markets for autism technologies. And through their participation in the AIC, researchers stand to gain status, publications, grants, patents, and personal income. The field's entanglements with the AIC are not limited to a few autism researchers but to the ideologies of progress and utility underpinning much of the field itself. Because HCI is a field that has recognized curative violence as a social good (sometimes explicitly, through "social impact" awards), individual researchers have an incentive to conduct AIC-complicit research. While the AIC is pervasive, it is sustained by myriad, entangled political economies, such as that of HCI and its drives towards industrial production, innovation, usefulness, and progress (Dourish, 2018;Lin and Lindtner, 2021). Therefore, changing the nature of autism research in HCI must entail addressing not individual research programs but rather systemic and intersecting structures within and outside of the field, incentivizing certain kinds of research.

Not robots; Cyborgs
Creating enduring and systemic change is a monumental challenge and beyond the scope of what any single paper can hope to accomplish. Instead, we aim to challenge the robotic and computeristic autism discourses that underpin so much autism technology research. The neurodiversity paradigm (see Kapp, et al., 2013;Singer, 2017) frames autistic differences as variations in human experience and the presentations thereof. Within this framework, autism, disability, and impairment are normalized. Within this framework, it is not the humanity of autistic people that is suspect. Instead, what is suspect is the humanity of an institutionalized conception of personhood that casts autistic people as less human than non-autistics and machines alike. As a replacement for autistic-as-robot rhetorics, we invoke the figure of the (crip) cyborg. In providing alternative framings for understanding autistic relationships with technology, we hope to enroll additional researchers into the growing ranks of HCI scholars attempting to resituate autism research around the wants and needs of autistic people (e.g., Kender and Spiel, 2022;Keyes, 2020;Ringland, 2019;Spiel, et al., 2019;Spiel and Gerling, 2021;Williams and Gilbert, 2019;Ymous, et al., 2020).

Cyborg affinities along the cyborg/tryborg continuum
In classic (Haraway) cyborg theory, cyborgs are defined by what they are not [12] -by the multiplicity of categories in which they (refuse to) fit and the boundaries they transgress ). Haraway provides four definitions for a cyborg: (1) a "cybernetic organism"; (2) "a hybrid of machine and organism"; (3) "a creature of social reality"; (4) "a creature of fiction" [13]. As with Suchman's (2011) co-constructed human-esque robots,  cyborgs comprise an artifact/category that challenges existing dualisms between organism and machine and between subject and object. Cyborg bodies and existences are often imagined as those that, through highly integrative technology use, can transcend non-augmented bodily limitations. Such conceptions of the cyborg are critiqued by disabled writers, who dispute Haraway's notion of the cyborg as "a creature of fiction" [14].
Weise (2018) notes Haraway's cyborg is a metaphorical creature. Too, in common usage, Weise notes that the cyborg represents technological augmentation used to bring about superhuman capacities. Weise (2018) writes that these popular conceptions discount the embodied experiences of the disabled people who, for years, have assimilated variegated technologies to facilitate their continued existence. These disabled people, cyborgs by  second definition, predate Haraway's cyborg fictions, which imagine the cyborg as something not yet here or something one might choose to be (Weise, 2018). Technology use makes critical bodily functions possible for these disabled people, whom Weise (2018) refers to as common cyborgs. Disability studies scholars and common cyborgs point out the importance of recognizing that cyborg non-fictions are not always as shiny and exciting as cyborg-as-superhuman narratives would have us believe Shew, 2020Shew, , 2017Weise, 2018). Cyborg existence often entails various costs, limitations, and painful bodily sensations associated with one's cyborg appendages. Weise (2016) refers to people who expand their capabilities via technology for reasons other than necessity (e.g., leisure, aesthetics, productivity, or other luxuries) as tryborgs. These not-quite-cyborgs, cyborgs of metaphor and aspiration, differ from disabled cyborgs because of the ways privilege and power differentiate between their technology usage (see Nelson, 2020). Tryborgs imagine metaphorical cyborg futures without recognizing the historical and continuing existence of non-metaphorical disabled cyborgs (Weise, 2018(Weise, , 2016).
Weise has used prosthetic legs for decades (Weise, 2016), but identifies a specific moment in 2010 as marking when cy [15] became a cyborg -when cy transitioned from a purely mechanical to a prosthetic knee (Wong, 2019). Although Weise's identification as a cyborg relates to integrating computerized parts with cy's body, cy is open to all disabled people claiming cyborg (Wong, 2019). For example, Shew considers herself a cyborg, despite not having any computerized parts (Wong, 2019). Scholars have interpreted Weise and Shew to construct different definitions of what it means to be a cyborg. Earle (2019) delimits cyborgs to those disabled people with necessary (for facilitating activities and functions of daily living) technological artifacts that are "within the body, strapped to it, or carrying the body in some way" [16]. Yet, Weise's cyborg construct includes those whose continued existences are maintained by less mechanical technologies, like anti-depressants. Williams and Gilbert (2019) define cyborgs as "those of us whose life and cognition are reliant and co-constituted by the existence of an interface and technology" [17]. This definition is helpful within the context of autism, as it explicitly includes people with learning, intellectual, and cognitive disabilities whose cognitive functioning is technologically aided.
There are myriad instances of autistic technology use in non-interventional contexts to facilitate daily functioning. In the case of mobile computers as alternative and assistive communication (AAC; see Alper, 2017;Zisk and Dalton, 2019) devices, essential communicative capacities are moderated by whether an autistic person has access to appropriate technologies. Harrison and colleagues (2019) explore ways autistic people appropriate off-the-shelf mobile computing devices to mediate undesirable / curate desirable sensory experiences for stress management. Zisk and Dalton (2019) address how autistics effectively use these same devices to communicate non-verbally (see, also, Alper, 2017). In considering William and Gilbert's (2019) cyborg alongside that of Weise (2018), there is an argument that these autistic technology use cases are cyborgian.
Importantly, disabled people are not alone in cyborgian necessary technology use. Transgender people represent another example of cyborg becoming. Transness and disability, and autism, in particular, share some parallels. First, trans is not a category completely separate from disability: many trans people are also disabled, and many disabled people are also trans (Bumiller, 2008;Warrier, et al., 2020). Sex and gender are sometimes used as lenses to understand autism, such as via the common rhetoric (conceived by Baron-Cohen) that autism is a case of "extreme male brain" (Jack, 2011). Further, dominant medical establishments often consider both autism and transness as traits to extinguish, and ABA shares some common history with trans "conversion" therapy efforts (via Lovaas, who was involved in both) (Gibson and Douglas, 2018). In their resistance to these conversion/extinction efforts, trans rights and disability rights also share parallels (Bumiller, 2008).
Trans people often become cyborgian by incorporating technology to fundamentally change their bodies (O'Shea, 2020). Some trans people incorporate technology into their bodies as part of medical gender transition processes, like surgeries and hormones (Gill-Peterson, 2014). Many trans surgeries involve implants incorporated into a person's body to physically manifest gender change. Hormones used for gender change are a technology that fundamentally changes a person's biology. In other cases, technologies can augment or expand trans people's experiences, bodies and identities (Haimson, et al., 2020). As a few examples, in Cárdenas' projects Becoming Dragon (2010) and Becoming Transreal (2012), virtual reality systems enable gender exploration, and for Nelson (2020), hand chip implants are queer cyborgian approaches. Stone (1995) described trans people as cyborgs because of how they sometimes incorporate technology into communication practices (in what she called "prosthetic communication"). Too, Stone (1995) suggests trans people are rendered cyborgs by their status as "boundary creatures," occupying liminal spaces not only related to gender but also at the edges of the human/machine boundary. Intersecting trans and disabled identities mean that some people are cyborgs across multiple dimensions.
While non-disabled and cisgender people also sometimes use technologies that augment their experiences, these augmentations do not raise to the level of cyborg status (Weise, 2016). As Nelson (2020) articulates, two unique elements -which trans and disabled people share -are necessary for claiming cyborg: (1) marginality (i.e., augmentations to one's body are intertwined with one's marginalized identity/ies); and (2) change (augmenting ones body changes its fundamental makeup, whether physically, biologically, or in terms of how one relates to the world).
In some cases, other marginalized groups, such as women and racial minorities, may also be considered cyborgs. For instance, Nakamura (2014) drew attention to the gendered labor of technology work by discussing how a community of Navajo women may be regarded as "natural" cyborgs in relation to their technologically mediated factory work. And of course,  famously described how women could be cyborgs, particularly women of color. Yet some critique Haraway's cyborg theory as ignoring power differentials between white women and women of color (DeCook, 2021;Puar, 2020;Sandoval, 2000;Schueller, 2005;Wilkerson, 1997). Thus, applying Haraway's theory in disability and trans contexts, which must meaningfully involve intersectionality, is fraught. Cyborg status, then, may involve a third element: (3) multiple layers of marginality.
Within disability studies, some, like , express concerns about the degree to which Harawayan cyborg theory risks portraying technological relationships as shared, yet depoliticized, phenomena devoid of references to power differentials. Still,  finds utility in cripping the cyborg. Like queer(ing), crip(ping) is both a noun and a verb (see Sandahl, 2003). Hamraie and Fritsch write that to crip is to assert "the noncompliant, anti-assimilationist position that disability is a desirable part of the world" [18]. Kafer writes, "Cripping the cyborg, developing a non-ableist cyborg politics, requires understanding disabled people are not cyborgs because of our bodies (e.g., our use of prosthetics, ventilators, or attendants), but because of our political practices ... Cripping the cyborg, in other words, means recognizing that our bodies are not separate from our political practices; neither assistive technologies nor our uses of them are ahistorical and apolitical" [19].  reconceptualization of the crip cyborg (or "cripborg," to borrow a term coined by Stevens in Nelson, et al., 2019) requires that we look beyond the interfaces between disabled bodyminds and machines (e.g., autistic children and humanoid robots) to the effects and assumptions of and within these technologies and their disabled usage.

Autistic cripborgs (and what their existence means for HCI): Centering disabled people's needs and knowledges
A critical distinction between things and people is that we attribute agency and self-determination to one but not the other. Within robot-assisted autism interventions, Spiel and colleagues (2019) note that autistic children have little or no agency in determining how they interact with the intervening machines, nor are they permitted to engage in sense-making of or with the robots. Autismrobot research appears unconcerned with achieving mutual intelligibility (see Suchman, 2006) between autistic children, robots, and/or non-autistic clinicians and caretakers. Instead, a primary purpose of robot studies can be understood as an effort to render the autistic child intelligible to the non-autistic other. The socialized robot now plays a role in the ablenormative socialization of the (robot-like) autistic child. The significance of imbalances in the order and direction of these relationships go unquestioned in this literature. A hierarchy seems to exist in which humans and robots can act upon one another, but wherein autistics can only be acted upon. Autistics are configured as so "epistemically absented" [20], so distant from intelligibility, that they have less agency and capacity for sense-making than the robots to which they are so often compared.
In contrast to (dominant constructions of) autistic robots, or robotic autistics, cripborgs have agency. Cripborgs have politics . Autistic technology use is inherently political, if only because systemic power structures and the AIC construct autistics-as-robots discourses as apolitical. Claims of "natural affinities" may obfuscate the politics of autistic embodiment and practice through convoluted citational practices and inadvertent influence of broader cultural tropes and market forces. However, making essentialized claims about minoritized people will always be political. Importantly, autistics are not cripborgs solely because of the technological practices imposed upon them by behaviorists and technology researchers. Autistics are cripborgs because of the ways they appropriate existing technologies to make space for and community with themselves and to advocate for their recognition as agentic.
The Internet has been an integral technological component of autistic communities and the autistic self-advocate movement for decades. Sinclair (2012a) writes about how, as the autistic self-advocacy movement was first coalescing, autistic community members shared online spaces with parents of autistic people. Parents led the prominent autism organizations, organized the autism conferences, and set the agendas for policy-and research-oriented advocacy (Sinclair, 2012a). Sinclair (2012a) recalls themself and fellow autistics posting on a parent-dominated forum in the early 1990s in the days following a large autism conference. The parents rebuked the autistic forum users. The parents said the autistic posts were of no interest to them. They had trouble grasping the notion that autistics might conceivably be interested in ongoing discussions about themselves and that the topics of importance to autistics might be relevant to organizations that ostensibly exist to support them.
Before long, an autistic person configured a forum by and for autistic people. Parents were allowed to join, but this forum was autistic space, and parents were subject to community norms (Sinclair, 2012a). To our knowledge, this event -along with the year-long autistic vs. parent flame-war that followed -is the first example in a long history of autistics leveraging digital spaces to reclaim narratives about themselves. We can see current-day examples of this kind of online activity on many forums, personal blogs, and popular social media sites (see, e.g., Osorio, 2020;Seidmann, 2020). Before Lesser (1998b, 1998a) ever wrote about autistic biology sharing similarities with computer architecture, autistics were using computers not because of an affinity for computers but out of an affinity for one another. They used computers and communication technologies to organize. To plot out a political agenda (Sinclair, 2012a). Maybe autistics do have a "natural affinity for technology;" perhaps confirmatory evidence for this claim will emerge from a groundbreaking study tomorrow. But this autistic computer use would still be cripborgian, agentic, and political, rather than robotic.
Were autism researchers in HCI to understand autistics as cripborgs rather than computers, it would not spell the end for the AIC. Researchers could continue reaping the benefits of autism research (publications, grants, income and accolades) but could also, at least potentially, provide direct benefits to the autistics within their research studies. The frameworks of crip  and neuroqueer (Rauchberg, 2022) technoscience can help explain this shift from research about autistics in service of non-disabled needs to autism research in favor of autistic needs.  articulate crip technoscience as a means for designers to align themselves and their work with disabled people's activist-oriented needs and knowledges. They place crip technoscience in opposition to disability technoscience, which refers to more traditional forms of individualizing, isolationistic, biophilanthropic and otherwise (techno)ableist (see Shew, 2017) approaches to disability and technology. We need a shift within autism technology research in HCI that moves autistic people from passive recipients of violent "charity" to leaders in designing their technological futures. Researchers more familiar with roboautism than cripborgian autism might look to crip technoscience and its four central commitments for inspiration.
The first commitment of crip technoscience is to center the work of disabled people as knowers and makers . Williams and Gilbert (2019) review disability technologies in HCI and find that HCI researchers struggle to recognize ways their work might adversely impact their users. They suggest this difficulty stems from researchers' lack of familiarity with the embodied and lived experiences of disabled people. Clinical autism discourses paint autistic people as devoid of self and self-knowledge (e.g., Baron-Cohen, 1995;see, also, Yergeau, 2013). Cripborg autistic existence challenges this notion by highlighting the politics and intentionality of autistic technology use. This commitment to disabled knowledges and to autistics-as-cripborgs suggests HCI researchers should reevaluate their methodological approaches to autism technology research, favoring participatory methods in which autistic people (as opposed to their non-autistic parents, teachers and caregivers) are enrolled as crucial stakeholders and subject-matter experts. Given issues around working with people with intellectual disabilities and the need for parental consent when working with minors, building and navigating these methodologies will require care and nuance. Spiel and colleagues (2018) model the necessary reflexivity and nuance for HCI researchers interested in ensuring the wellbeing of disabled research participants.
The second commitment of crip technoscience is to "access as friction" [21]. This commitment rejects disability design and intervention forms wherein inclusion and accessibility are synonymous with assimilation. Ymous and colleagues (2020), in a paper some HCI researchers felt was contentious, described ways that expectations of non-disablement within HCI are at odds with the realities existing as a disabled researcher within the field. They call out the violence inherent to sweeping ableism within HCI. It would be entirely understandable for HCI researchers involved in the types of practices Ymous and colleagues (2020) point out to feel uncomfortable. Yet, a commitment to "access as friction" asks these researchers not to look the other way -to stay with the trouble, so to speak (see Haraway, 2016). Places where the demands of disabled people diverge from dominant research priorities and perspectives in HCI can be generative. Williams and Boyd (2019) demonstrate that, by not shying away from these frictions, HCI researchers can build coalitions and work towards defining and practicing alternative types of research mutually beneficial to all parties. Such papers, like the present papers, can be treated as invitations to participate in new research (autistic-led) programs rather than merely as condemnations. In staying with the friction and accepting these invitations, researchers may avoid the pitfalls inherent to dominant approaches to disability design research in HCI (see, e.g., Shew, 2020Shew, , 2017.
The third commitment of crip technoscience is to interdependence as a political technology . Hamraie and Fritsch write: We position the crip politics of interdependence as a technoscientic phenomenon, the weaving of relational circuits between bodies, environments, and tools to create non-innocent, frictional access. Mainstream disability technoscience presumes disability as an individual experience of impairment rather than a collective political experience of world-building and dismantling. This perception has two primary consequences. First, disabled people are perceived as dependent and the goal of technoscience becomes to encourage independence. Second, disability and technology are both perceived as apolitical and stable phenomena, rather than material-discursive entanglements that take shape through struggle, negotiation, and creativity. [22] Traditional approaches to autism technology research support behavioral interventions concerned with rendering autistic people independent. Yet, despite Western society's extreme valuation of the notion of independence, independence is a lie. None of us are independent. We are all interdependent on one another (Mingus, 2017). As Mingus writes, very few of us grow our own food, produce our own clothing, or create our own energy sources. Interdependence is a fundamental condition of human existence, yet our self-narratives of and preferences for individuality obscure this truth. In committing to interdependence as a political technology, HCI researchers have an opportunity to -in concert with autistic people -define alternative outcomes to autism technology research that promote care and support (see Piepzna-Samarasinha, 2018) rather than independence and normalization (see, e.g., Bennett, et al., 2018). Cripborgian framings of autism can help identify forms of interdependence that might warrant exploring (see, also , Bennett, et al., 2018). As computers and robots, autistics are fundamentally solitary and asocial, sharing more affinities with machines than with people. As cripborgs, autistics share affinities not only with one another, but with other groups of cyborgs. Cross-and trans-identity cyborg affinities represent coalitional and intersectional technology research opportunities with broad impact beyond any single identity group. Working on technology research to support coalitional forms of interdependence can, perhaps, satisfy HCI researchers' desires to "do good" while also maximizing the number of people who might benefit from their research.
The fourth commitment of crip technoscience is to disability justice . The term disability justice was coined by disabled queer and trans people of color comprising the Disability Justice Collective (see Piepzna-Samarasinha, 2018). Disability justice differs from disability rights. Rather than focusing on state-granted privileges, it seeks to dismantle the root causes of ableism: the intersecting forces of heteropatriarchy, white supremacy, colonialism, and capitalism (Sins Invalid, 2019). For some academics, particularly those within fields like HCI oriented around industrial progress and innovation (Dourish, 2018), perhaps adopting anti-capitalism as a tenet of good research is a bridge too far. But, researchers might operationalize a commitment to disability justice in autism technology research by ensuring research projects include and are led by those most affected by intersecting forces of oppression. Historical autism research has focused on autism as a male phenomenon (Moore, 2021). Researchers might look for ways to support the techno-political practices of, for example, autistics who are women, queer, trans and/or Black, Indigenous, and people of color. This commitment calls on researchers to sort out what it means to work in close collaboration with autistics with significant intellectual disabilities while remembering always to presume competence (see Bascom, 2014). As Costanza-Chock (2020) explains, design justice is a framework for understanding and designing against hegemonic power structures. The framework is significantly built around the precepts of disability justice (see Costanza-Chock, 2020) but, unlike disability justice, has been adapted to be directly applicable to HCI. Following the tenets of design justice may help HCI researchers operationalize disability justice within their work.
Rauchberg (2022) elaborates neuroqueer (see Roscigno, 2019;Walker, 2021; technoscience as an extension to crip technoscience, focusing on the liberatory and interdependent technological futures of autistic and otherwise neurodivergent people, particularly with respect to digital and communication technologies. Building from the standpoint that crip technoscience and examples of its implementation (see  privilege neurotypicality, Rauchberg's framework explicitly extends crip technoscience's commitments to interdependence and the recognition of disabled people as knowers and makers to disabled minds, as well as bodies. Too, Rauchberg's (2022) framework goes beyond positioning itself in opposition to disability technoscience, clearly and specifically rejecting curative violence (see Kim, 2017). Holding in mind these tenets of neuroqueer technoscience, and an opposition to curative violence in particular, HCI researchers might recognize the potential of autistic-led autism technology research to champion liberation and adaptation, rather than assimilation and control.

Conclusions
Popular autism discourses are often widespread and deceptively pernicious. They become so naturalized that they go unquestioned. But, as with any assumptions, particularly those about groups of people to which someone does not themselves belong, critical examinations are warranted from time to time. We should avoid justifying research on or with autistic people in claims that sound true. While mechanical constructions of autism may align with broader, cultural autism narratives, claims about "natural affinities for technology" lack support and may cause harm. Positioning autistic people as or like robots (i.e., things) may normalize forms of coercive intervention that researchers would question if only they more clearly saw their subjects as people. We suspect that despite any particular benefits they may reap from their research, most autism technology researchers are acting with good intentions. In outlining how those intentions are misguided and complicit in violence, we hope those researchers will reconsider their future research approaches. In providing alternative ways to understand autistic people (as cripborgs, rather than computers) and suggesting crip technoscience as a framework for research grounded within this new understanding, we hope to enroll these well-intentioned researchers in the project of dismantling anti-autistic ableism, both in and beyond HCI.