Louise Hickman, Cynthia Bennett
In this conversation, we reflect on what access work means to us as disabled scholars who conduct research on disability and technology. To arrive at this phrase, in addition to our empirical work, we draw from three sources. First, Access to Work is a U.K.-based government program that funds certain accessibility accommodations for workers with disabilities. Second, we gather terminology from Leah Lakshmi Piepzna-Samarasinha's book Care Work: Dreaming Disability Justice , which archives lineages of disabled people caring for one another by meeting access needs collectively. These narratives exemplify Mia Mingus's phrase access intimacy, which is "that elusive, hard to describe feeling when someone else 'gets' your access needs" . Finally, we bring in the interdependence disability justice principle, which argues that no one is independent but rather a part of networks, which may scaffold access-work analysis in HCI. During this discussion, we articulate what access work has meant in each of our research programs. In doing so, we draw out some methodological recommendations and tensions for HCI researchers and practitioners to carry out when pursuing collaboration and labor analysis.
→ Access work is an interdependent network of care that builds on the principle of disability justice: collaboration and intimacy building.
→ Non-innocent authorization affirms the practice of "working with friction" to understand both the messy encounters and the value of access work for disabled people.
→ Deploying automated access tools that lack the situated knowledge of human access workers to tasks requiring such nuance risks regressing, instead of enhancing, accessibility.
Louise Hickman: I find the term access work has many entangled meanings. How does access work show up in your research and everyday life?
Cynthia Bennett: I have encountered access work both formally and informally. To me, formalized access work refers to someone hired, often through employer- or government-sponsored schemes, to do particular work that remediates access to communication and daily living skills like sign language interpreting, captioning, and attendant care. But these formalized relationships are often unrealistically portrayed as transactional and static. Instead, through my informal experiences as a blind person who hangs out with other disabled people and from disability justice activism, I understand access work to be fluid and interdependent [1,2]. Further, and crucially, the access work I've experienced and read animates interdependencies by centering people with disabilities as access workers, not just as recipients of the aforementioned government-sponsored programs. Blurring who gives and receives access work by foregrounding the reciprocity in these relational acts has kept me accountable to write about people with disabilities as having agency, and about accessibility as networked and collective.
LH: I like that your response underlined how malleable the term access work is. Like you said, there are moments in our everyday interactions where we take on the responsibility of access work, and they do not always map onto our infrastructure.
The genealogy of access work in my research came from the U.K., where there is a government scheme by the Department for Work and Pensions called Access to Work. This allowed me to think more about retaining employment, disability, and technology. Quite often, disabled employees refer to their workplace support as access workers to differentiate it from personal care in the home.
Building on the politics of access work in the workplace and reflecting on my own experience teaching in the classroom with CART (communication access realtime translation) captions, I started to think about the relationship between technology, labor, and disability justice, especially Mia Mingus's teaching on access intimacy . I first experimented with a related phrase, collegial intimacy, to describe the complex and technical intersection of access work. For example, captioners may work closely with clients to code the names of speakers, which is often obscured by automated captions. Through this work, I focus on the intimate practice of dictionary building with particular captioners to find ways of describing the process. I produced a film with a New York City—based artist, Finnegan Shannon, called Captioning on Captioning, which featured some of the coding process done in real time . The film exemplifies intimacy building by showing captioners working with class syllabi and speakers' names during meetings. We asked captioners to describe the practice of captioning to situate their expertise. We are working on a new edition with audio description, which raises new research questions about the process of articulating the visuals and physicalities of such technical expertise to blind audiences.
The film itself is a good segue to reflect on how our disciplinary training informs our practice and so on. I completed my Ph.D. in communication, and my bachelor's is in American studies. My research methods have been largely shaped by disciplines across the humanities and social sciences, which means there is a practice of doing translation work across disciplines. What has informed your choice of method to study access work?
CB: My undergraduate degree is psychology. I now understand that I predominantly learned to frame mental health disabilities as deficits, as maladies, rather than taking a more nuanced and empowering approach that recognizes neurodivergence as generative. My graduate training was in HCI, but I also selected coursework and did self-study in disability, race, and gender studies. I am grateful this self-study was supported by my mentors, as it was also very personal. I have been disabled all of my life and started formal activism as a teen, but I didn't get access to coursework that framed people like me in an empowering way until I was 25—and even still, I feel lucky to get the access and space to learn disability studies perspectives.
I didn't make methods decisions in my research intending to uncover access work. Their disciplinary roots range from HCI to feminist reflexivity, but more specifically I chose to read what I noticed through interdependence and access intimacy logics [1,2]. So I will summarize two projects that revealed access work to me, and that can advise readers' methods choices.
We observed that the blind travelers did access work creatively and intimately—access work and friendship were indistinguishable.
Access work became apparent to me during a previous collaboration. The formative project outlined a design space for AI tools applied for accessibility purposes. We observed blind people traveling together as a case because previous academic research had charted nonvisual navigation as a grand accessibility challenge. We observed that the blind travelers did access work creatively and intimately, in that access work and friendship were indistinguishable. The travelers made sense together and communicated status updates about what they were understanding to the group, and collectively they would make decisions with this sensemaking. This sensemaking involved touch, by hand, exploring objects when possible, but also by tactile feedback gained through their long canes. They also tuned into auditory and spatial cues. Discovery communications to the group were often done verbally, (e.g., "I found the path [having discerned with their cane that the smooth surface was likely pavement]," or "I hear people walking this way"). At a tourist monument, for example, the direction of pedestrian traffic can cue where the ticket counter is. So the travelers were putting all these pieces together. Everyone does this, but since the forms of sensemaking that these blind travelers were doing was nondominant and often done in the absence of accessible infrastructure, it stuck out.
Methodologically, witnessing people getting by in their everyday lives, even if the vacation context was unique, was generative. We came in with preconceptions about AI as a potential solution, and the impact of this work was to establish to us researchers that blind people and their tools are part of a network, and any AI solution would not be such if it did not integrate into that network.
Another methodological learning was the value of subject matter expertise. As a researcher who is also blind like the travelers, I translated knowledge and contextualized what we noticed. For example, if an observation stood out to a sighted person, I might have responded that it's something blind people commonly do. It might be extraordinary in how it contrasts with dominant visual sensemaking norms, but it's not that interesting to blind people. I also brought things to the attention of the research team that may have gone unnoticed otherwise, pointing it out as meaningful sensemaking and access work. The implication for methods decisions is the value of having knowledge translators on the team and foregrounding their expertise.
Sometimes providing access becomes an act of curating other people's experiences, which is antithetical to the spirit of accessibility.
Next, I led a similar project , but each set of participants consisted of one blind and one sighted person whose access work was sponsored by a government scheme. We observed that often the blind and sighted colleagues experienced access intimacy and became companions in the ways Mingus described. Some hired known contacts whom they already trusted; for some the companionship came through their common understandings of and respect for access. These observations, along with many moments shared with the blind travelers, idealized interdependence, and that had some trade-offs. From our observations, we drew out a practice called non-innocent authorizations.
LH: I find the term non-innocent authorization so interesting. Can you elaborate?
CB: Sometimes the access work the sighted person was doing was less collectively done with their blind client but opened up an opportunity. Maybe it happened through instructions on negotiating visually demarcated pathways or describing visual digital content. With that opening of an opportunity came an authority. It wasn't done from a negative place but rather from a genuine interest in fostering access. However, even with the best of intentions, the access worker still decided what information was necessary. For efficiency reasons, they often considered the desired outcome, such as completing a task and the time they had to work together, as a rubric to filter infinite visual phenomena into what they shared. This inevitably meant that the blind clients were missing information. This is not necessarily bad, but when understood along with the fact that this work was remediating access to fundamentally inaccessible infrastructure, specifically that which is visual dominant, the scales measuring who has agency in the given situation still tipped toward sighted people. I want to clarify that our noticing authorities occurring according to normed abilities was structurally supported, but non-innocent authorizations can and do occur among people with disabilities. Who we chose to notice—pairs of one person with low vision and one person with full vision as formalized through that government scheme—and how we make sense of it were themselves decisions that reflected our current thinking, priorities, and publishing contingencies, among other factors. Caveats made transparent, non-innocent authorization aims to preserve the value of human connection while staying accountable to the fact that access work, or more broadly actions done in service of supporting people with disabilities, is not self-evidently or universally good. It is messy like other interpersonal encounters. Sometimes providing basic access itself becomes an act of curating other people's experiences, which is antithetical to the spirit of accessibility—for people receiving access provisions to have more agency.
LH: I love the term non-innocent authorization. It touches on the ongoing tension between access work and technology and gives me a great jumping-off point to discuss how my work with captioning started. The project began with me as a misinformed graduate student who became focused on the failure of captioning. And why couldn't real-time captioners keep up with spoken speech in the classroom? Once, my captioner's hard drive failed and they consequently lost their dictionary software. We both had to start again with building new dictionaries for me to access the classroom. This was a key moment when I started to rethink my position to access work, the complexity of these networks, and how workers and technologies come together to produce good captions.
CB: Can you say more about the process of captioning?
LH: Captioners use steno form—digital shorthand—to capture spoken speech under real-time conditions. You can think of these word parts as phonetic shorthands, in which the captioner uses a stenographic keyboard with 24 keys, different from the QWERTY keyboard. The same steno form can mean different things depending on the context, which the captioner and captions reader are ever learning. For example, the steno form A-BG might have three words associated with the shorthand: access, academic, and accusation. They are vastly different and require a degree of embodied knowledge and understanding of how the speakers they supported interacted with each other.
Building personal dictionaries takes time. For example, there might be a student in a sociology class, with specific terminology used. This is different from the demands of a chemistry class, where the same captioner might work for another client. Training a new dictionary is an embodied practice and depends on people and the social context around the captioner.
Captioners often spend time preparing for meetings, for example, looking up commonly used terms on the class syllabus. They have to code authors' names and speakers' names, and provide local context to the spoken speech used in the classroom. In some sense, they are working ahead of time and trying to provide context for an event that has not yet taken place. It is an economy of anticipation that touches on the politics of inclusion. Planning ahead means there is work being done to determine the readiness of the dictionary before the event. For some, there are implicit assumptions that captioners should embody the demands of the real-time process and be more machinelike. The emphasis on the optimization of real-time writing applies pressure on access workers.
Understanding the labor, in some ways, we have to expect that captioners will make mistakes without repercussions. The pressure of real-time writing requires high cognitive demand from the workers.
Quite often, stenographers move between jobs. For example, courtroom transcription is less demanding regarding real-time writing; the transcript can be edited after the fact. This is not the case for broadcasting work like CART. Coping with these demands is another aspect of access work and underlines the precarity of access infrastructure. It's worth pointing out that CART captioners and sign language interpreters have a history of being contract workers rather than directly employed, which predates the rise of on-demand that we experience today. There has always been some degree of mobility for access workers when choosing the jobs they take on. It's interesting to see how captioners might reposition themselves to prevent burnout. Still, as I described, the process of dictionary building and steno form involves a lot of decision making by the captioner, which I now understand could be a type of non-innocent authorization. The labor demands and caption readers' agency are quite often in tension.
CB: How do you reconcile the current demands on access workers in your research?
LH: I now put access work into conversation with public policy recommendations toward infrastructure changes. For example, a guiding question is: How do we prevent further harm to access workers? I believe that attending to the well-being of access workers in addition to those receiving the support will ultimately improve accessibility.
I'm currently working on a report focusing on several case studies on automated delivery devices, automated vehicles, and wheelchair-accessible vehicles. Working across these case studies means I can look at the broader impact of access work increasingly falling into the private marketplace. Most of the case studies focus on the U.S. For example, as independent contractors, Uber drivers in California are held responsible for providing an accessible service, not the company itself. If Uber explicitly directed its drivers to provide access for disabled users, this would tip the ABC test—a formula to calculate employee versus contractor status—toward it being noncontractual work , which is seen as an unworkable business model for on-demand driving. This is another example of the ambiguous status of access work. From the perspective of the worker, their work does not fit neatly into the marketplace or the welfare state, and this uncertainty places the onus on the workers to do the right thing and determine what good access looks like. This reflects some of the anxiety around good captioning that aligns with your earlier point on non-innocent authorization. For captioners, good captioning means taking on more responsibility and nonpaid work to prepare the dictionary ahead of time. What counts as "good access" extends into other areas that range from the captioning agency, training schools, courtroom reporters, and so on. The gig economy has brought to the fore unacknowledged labor that has long befallen access workers who have long been self-employed.
In the first couple of weeks of the Covid-19 pandemic, captioners working in higher education were quite concerned about the loss of work. But this did not come to be. Teaching staff were confronted with the failure of automated captions and the harm this caused to deaf and disabled students. Cloud captioning, in this instance, didn't have the capacity to care about social contexts in the same way. The demand for captioners grew in this period, and there were not enough captioners to meet the demand, meaning that deaf and hard-of-hearing users were not always getting the support they needed. This situation has created a new politics of accessibility around the future of access work and assistive technology, since there is a risk that institutions with less funding will have to make do with automated captions, whereas institutions with funds can invest in access with more prep work and care with a smaller pool of workers. The possibility of this two-tier system doesn't sit well with the future aspirations of disability justice. We need community-led projects that understand and highlight the harms involved.
Through this conversation, we have shared our perspectives on access work. We have expanded on research informed by critical disability studies, accessibility, and HCI research to reveal the integrity of access work.
The distribution of access work has many configurations, including blind travelers and people with disabilities in mixed-ability collaboration and captioners working closely with their clients. Access work is a form of intimacy building. It is a practice of forming social relations and maintaining accessible infrastructure undervalued by automation and casting human access workers as independent contractors. Access work aligns with the collective work of Leah Lakshmi Piepzna-Samarasinha's writing  to emphasize that accessibility is not a universal good but instead weaves together with an abundance of non-innocent authorizations, which coexist with the precarity and exploitations of gig work. The automation of access, on these terms, must reckon with the futures where labor is undervalued and ephemeralized by recognizing only certain activities as access work. In conclusion, how can we build toward an ethics of access that supports models of interdependence and access intimacy put forward by disability justice activists? And how can access work provide an analytical framework for HCI researchers to deepen the discussion around the future of work more broadly? These questions may act as scaffolding to not only nuance but also clarify how interactions—even those that are mundane—become generative to the contributions of one's specific research.
2. Mingus, M. Access intimacy, interdependence and disability justice. Aug. 2017; https://leavingevidence.wordpress.com/2017/04/12/access-intimacy-interdependence-and-disability-justice
3. Hickman, L. and Finnegan, S. Captioning on Captioning. LUX, London, 2020; https://lux.org.uk/event/louise-hickman-and-shannon-finnegan-captioning-on-captioning/
5. Bennett, C.L., Rosner, D.K., and Taylor, A.S. The care work of access. Proc. of the 2020 CHI Conference on Human Factors in Computing Systems. ACM, New York, 2020, 1–15; https://doi.org/10.1145/3313831.3376568
Louise Hickman is a research associate at the Minderoo Centre for Technology and Democracy at the University of Cambridge. Her work looks at the ethics of access, feminist labor, and digital infrastructure for disabled people and access workers. [email protected]
Cynthia Bennett is a senior research scientist at Google. Her work is at the intersection of accessibility, disability representation, and responsible AI. [email protected]
Copyright held by authors
The Digital Library is published by the Association for Computing Machinery. Copyright © 2023 ACM, Inc.