In this section, we will explore a few specific underserved populations and their privacy challenges and needs. When we describe people with disabilities, we follow the ACM accessible writing guide . We recognize the important role that language can play in the marginalization of people as well as the language we use may or may not reflect the norms within a particular disability community.
13.2.1 Models of Disability
There are many models (certainly more than three or four) of disability that have been recognized by disability scholars, such as the medical model, the social model, and the critical realist model.
Traditionally, the medical model has been used by scientific communities, but is often considered problematic. This model sees disability as something wrong with a person that must be “fixed” and has contributed to the oppression of people with disabilities. Disability rights activists then proposed social models of disability that identify disability as socially constructed and grounded in society and culture . This refers to disability as a problem with a society’s lack of inclusiveness rather than a personal issue . This model also has been critiqued for its emphasis on independent living (which isolates the realities of many people with disabilities who require assistance), as well as supporting “normalization” rather than celebrating or acknowledging disability pride and difference . The cultural/postmodern model was created to address the medical realities, lived experiences, and social elements for some people with disabilities . This model sees disability as another way of being, a cultural standpoint or lifestyle . The critical realistic perspective, which emerged from disability studies and was proposed to inform accessible technology design, also centers on rich, lived experiences of individuals with disabilities . Sins Invalid, a group of artists with disabilities, proposed “A Disability Justice framework understands that all bodies are unique and essential, that all bodies have strengths and needs that must be met... We understand that all bodies are caught in these bindings of ability, race, gender, sexuality, class, nation state and imperialism, and that we cannot separate them” . It is important to note that these models of disability are often respected based on how an individual associates personally within these frameworks. Thus, researchers and designers need to recognize the complexity and personal tendencies and experiences attached to such models and will dissuade from bifurcating them (e.g., the medical model vs. other models).
In summary, there are many models of disabilities, such as:
Medical model: disability is medicalized as being deviate from the normal biological functions .
Social model: disability is socially constructed as being a problem with a society’s lack of inclusiveness rather than a personal issue .
Critical realist model: “disability as an interaction between individual and structural factors” where individual factors can include impairments and structural factors can include others’ attitudes towards disabilities .
13.2.2 People with Visual Impairments
There are a wide range of disability conditions. Our first example focuses on people with visual impairments. Visual impairments exist on a spectrum, ranging from partial to complete loss of vision. In clinical settings, the term “visual impairment” refers to a “visual acuity of 20/70 or worse in the better eye with best correction, or a total field loss of 140 degrees” . “Blindness” means that a person cannot see anything, whereas “low vision” denotes “sight that may be severe enough to hinder an individual’s ability to complete daily activities such as reading, cooking, or walking outside safely, while still retaining some degree of usable vision” . A person can have visual impairments since birth or after birth (e.g., due to accidents, medical conditions, or aging).
While computers and smartphones help improve the independence and quality of life of people with visual impairments, these technologies (particularly the mouse, visual input/output, and touch-based user interfaces) also pose significant accessibility challenges for this user group. In turn, people with visual impairments (especially those living with blindness) often use screen readers (e.g., JAWS, Window-Eyes, NVDA, VoiceOver) on their computers or phones to parse and read the information from the screen. Screen readers usually read a screen sequentially but also support keyboard shortcuts to allow users to skip certain elements of a page or extract a list of hyperlinks on a page for faster navigation. People with visual impairments (especially those with low vision) also use screen magnifiers (e.g., ZoomText, MAGic) to zoom into certain parts of the screen to make it more readable.
People with visual impairments face many accessibility challenges with information and computing technologies and are likely to struggle with current mainstream privacy user interfaces (e.g., https padlock), which heavily rely on visual representations.
Existing literature has highlighted a number of privacy concerns of people with visual impairments:
Shoulder surfing (e.g., during their usage of ATM).
Aural eavesdropping (e.g., screen reader reading aloud private content).
Asking others even strangers to read inaccessible documents (e.g., mails).
Using assistive technologies can attract unwanted attention and make people with visual impairments more noticeable to attackers.
Difficulty in using end-user privacy/security mechanisms (e.g., privacy settings).
Taking or sharing images/videos that might contain private or sensitive content.
These and other related privacy and security concerns/needs of people with visual impairments have been identified. Holman et al. conducted focus groups with blind users and identified their top 10 security challenges: (1) CAPTCHA, (2) auto logout, (3) auto refresh/reload webpage, (4) inaccessible PDF (i.e., the PDF is not marked up with tags that can be read by a screen reader), (5) inaccessible antivirus software, (6) auto install software, (7) auto software updates make software inaccessible, (8) SecurID (a random number display in the device used for logging in), (9) keyloggers, and (10) spams . Some of these are more general accessibility issues such as inaccessible PDF; others are addressed by existing antivirus or anti-spam software. In terms of challenges related to CAPTCHA and authentication, there are a number of mechanisms proposed to improve or replace them (e.g., [12,13,14]).
People with visual impairments have privacy concerns about using mobile devices when they are in the speakerphone or screen-reading mode or generally in public because others can see or hear what they are speaking or doing [15,16,17]. People with visual impairments can wear earphones, but that is sometimes inconvenient  and could limit their abilities to hear or sense the nearby environment, making them vulnerable to attacks . The iOS Screen Curtain allows iPhone users to blank their screen, but that does not help with the privacy issues caused by the screen-reading mode and users with visual impairments may forget to activate that feature. The use of assistive technology (e.g., a portable magnifier) could attract unwanted attention and make users more noticeable to attackers [15, 19]. People with visual impairments often have to compromise their privacy for achieving independence and/or convenience.
Ahmed et al. have conducted two interview studies specifically investigating the privacy needs and practices of people with visual impairments in online and offline settings [18, 20]. They found that these users face difficulties in detecting visual or aural eavesdropping, have physical privacy and security concerns (e.g., using ATM), and sometimes need to ask others (even strangers) to help (e.g., read documents, type pin in shopping) . There are proposed solutions for specific tasks (e.g., accessible ATM ), but no generic solution to address the privacy risks emerged from asking others to help.
These users also report difficulties in managing their social media sharing, citing the difficulties in using the privacy settings on social media sites (e.g., Facebook) . These privacy settings have been found to be difficult for social media users in general . Furthermore, we have found that people with visual impairments were also concerned about online tracking (i.e., their data or web browsing activities being collected by companies or governments) , which has been shown as a privacy concern of the general population . There are browser extensions such as script blockers (e.g., NoScript) and ad blockers (e.g., Ghostery) that block third-party content or scripts on a webpage, but they have usability issues for general Internet users .
A few studies have focused on privacy/security issues for people with visual impairments. Many privacy/security threats arise from the use of accessible technology, as these devices inadvertently generate new avenues for passersby to learn personal information. People with visual impairments have concerns about aural and visual eavesdropping in public when using screen readers and screen magnifiers, respectively [26,27,28,29]. Prior work also suggests that this user group may not notice privacy/security risks in their environment or inherent in the technology they use . The use of accessible technologies can also draw unwanted attention and potential exploitation . To mitigate some of these issues, people with visual impairments use privacy features (e.g., iOS Screen Curtain) and wear headphones to mitigate problems with screen readers . Ahmed et al. identified privacy/security concerns or challenges people with visual impairments face such as difficulties verifying the security of banking or shopping websites, maintaining privacy on social media, asking strangers for help  and physical safety/security challenges in public spaces and at home . Our most recent ethnographic research with people with visual impairments and their allies in their everyday lives found that they often work cooperatively to protect the privacy and security of people with visual impairments, yet most existing privacy/security mechanisms fall short of supporting this kind of cooperative behaviors .
There are a number of privacy-enhancing technologies (PETs) designed for people with visual impairments:
Accessible authentication (e.g., PassChords , UniPass )
Accessible CAPTCHAs (e.g., more accessible audio CAPTCHAs )
Privacy-enhancing assistive features (e.g., Apple’s Screen Curtain)
General assistive tools making content more accessible and people with visual impairment more independent (e.g., screen readers)
Prior research efforts primarily those from the field of accessible computing have proposed different mechanisms to support people with visual impairments in various privacy and security-related tasks. One notable example is more accessible CAPTCHA (i.e., Completely Automated Public Turing test to tell Computers and Humans Apart) designs, for instance, using pairs of images and sounds [35, 36] as well as moving the controls for audio CAPTCHAs within the answer textbox of the authentication interface . Another notable area is authentication. For instance, Azenkot et al. designed a password scheme that utilizes patterns of finger taps on a touchscreen . Barbosa et al. designed a password manager that allows visually impaired users to easily transfer their login credentials from their mobile devices to web-based services .
Previous research has also elicited feedback on proposed solutions to physical privacy and security threats. Ahmed et al. found that their visually impaired participants appreciated the idea of devices that could detect the number of people in their vicinity, assist them with navigation, and prevent shoulder surfing attacks . In their follow-up study, participants endorsed the idea of knowing others’ proximity, identity, and activities, as well as inferences about the intentions of others’ actions .
13.2.3 Are Existing Privacy-Enhancing Technologies Sufficient?
While there is not much research on the experiences of people with visual impairments in using privacy-enhancing technologies, there are anecdotes that suggest the existing PETs are insufficient for this user group. For instance, privacy/security indicators (e.g., the https lock icon in web browsers) might not be very accessible to people with visual impairments who use screen readers. Privacy settings have also been shown to be difficult for people with visual impairments . Some of these are difficult to use for the broader population. It is also worth noting that some assistive technologies can introduce privacy issues. For instance, visual question and answer tools such as Be My Eyes and VizWiz  allow blind users to take pictures and ask questions about the pictures to crowd workers or volunteers (who can be total strangers). These pictures might contain private or sensitive content (e.g., credit cards, medicine details) .
13.2.4 Intersectional Privacy
Empirical research on people with visual impairments in general and their privacy and security practices in particular tend to focus on their visual impairments, which often are, however, a single aspect of their multifaceted and intersectional marginalized identities. In our own experiences of working with people with visual impairments, we found that many of them often have other aspects of marginalized identities such as other disabilities, minoritized race, and gender identities . These multifaceted and intersectional marginalized identities often contribute to their challenges and influence their privacy and security practices .
Intersectionality is a key analytic framework proposed by Kimberlé Williams Crenshaw in the late 1980s and is situated in the lived experiences of black women, women of color, and intersecting identity structures of race, class, gender, and sexuality . Crenshaw’s intersectionality arose from a legal perspective regarding a case in which an African American woman sued a company for discrimination in not hiring her. The judge dismissed the case because the company claimed to hire African Americans and women. In response, Crenshaw problematized this claim because African American women have multifaced identities including both gender and racial identities, and suggested that the company did not hire African American women. Crenshaw wrote, “I used the concept of intersectionality to denote the various ways in which race and gender interact to shape the multiple dimensions of Black women’s employment experiences” . Among this initial definition, Crenshaw has specified three kinds of intersectionality: structural, political, and representational. Structural intersectionality refers to how “the location of women of color at the intersection of race and gender” informs identity and marginalized positions . Political intersectionality highlights how “feminist and antiracist politics have, paradoxically, often helped to marginalize the issue of violence against women of color.” Representational intersectionality refers to cultural norms that create certain minoritized positions regarding identities . It is important to note that the concept of intersectionality encompasses many important factors beyond race and gender such as class and sexuality (e.g., used in queer studies by LGBTQ+ activists).
Since the term intersectionality has been coined, it has been adapted by a large number of feminist, critical race, critical disablity, and queer studies scholars as a research framework to examine complex identity and social structures. While retaining the black feminist foundation of this framework, critical disability scholars such as Rosemarie Garland-Thomson have proposed a disability axis on an “intersectionality nexus” which views disability not only as socially constructed but also intrinsically multifaceted . In this foundational scholarship, she connects disability, race, gender, class, and queer theory . Specifically, she considers disability along these complex and multifaceted elements such as political, social, and personalized understandings of disability identity . As such, her ability/disability system is meant to show another identity perspective to Crenshaw’s notion of intersectionality .
Intersectionality has been a recent topic in human-computer interaction (HCI) scholarship with the call for inclusion of critical theories such as feminism (e.g., ) and critical race (e.g., ). This line of work advocates that by using intersectional analyses, people can be better understood, thus leading to richer data and more ethical methodologies and designs (e.g., ). For instance, Schlesinger et al. point out that besides race, gender, and class, other dimensions such as disability or age are also good for intersectional analysis .
In our studies on visual impairments, we found many of our participants to have intersectional identities along with the visual impairments that shaped their experiences . Therefore, we adopted intersectionality as an analytic lens to unpack the everyday privacy/security experiences of people with visual impairments. Our intersectional analysis is akin to intracategorical intersectionality , which “focuses on a single identity category...and then analyze other dimensions of identity within the target community” . We focused on people with visual impairments while considering their overlapping identity dimensions (e.g., age, gender identity, and other disabilities) .
For instance, one of our participants unfortunately lost her sight in an accident. She is a mother and self-identifies with having bipolar disorder and a learning disability. She lives with her children and mother. She often asks their help with many things from emailing to managing her bank account. We observed that her privacy and security needs and practices were often influenced by these multifaceted and intersectional aspects of her identity. She gave a hypothetical example where it would be difficult for her to have another male friend to illustrate her inability to hide her phone conversations, which might in turn lead to misunderstanding from her boyfriend. She wants to control the visibility of the conversations on her phone herself, but she found the technologies too overwhelming to learn. There are typically functionalities such as deleting phone call records, text messages, or the contact information that allow phone owners to protect their privacy and communication on the phone. However, there is no simple one-click feature that “[hides] conversations on a phone.” In practice, users need to understand and use a combination of technical features to clean their conversations (in various apps) on the phone. This participant said she had a hard time knowing and learning how to use all these features. One key insight here is that the combination and intersection of her visual impairments, bipolar disorder, and learning disability probably all played a role in her challenges in using the technologies and achieving her privacy/security goals on her own .
13.2.5 People with Hidden Disabilities
We use the term “hidden disabilities” as an umbrella term to cover a wide range of disabilities such as learning disabilities (e.g., dyslexia), autism spectrum, and attention deficit hyperactivity disorder (ADHD) as well as psychosocial, internal conditions such as chronic pain, and mobility disorders as well as anything “not obvious” to others. Disability and HCI research tend to focus on physical disabilities such as visual and motor impairments rather than hidden disabilities. Even fewer studies have been done to understand people with hidden disabilities and their privacy and security needs.
In our own research, we conducted focus groups with people with hidden disabilities to understand their information disclosure practices, in particular the disclosure of their disability identities. Similar to our study on visual impairments, we found that people with hidden disabilities often also have multiplicity in identity. In addition, since their disabilities are often not obviously visible to others, their (marginalized) identity disclosure is a key aspect of their everyday privacy practices.
We identified two main domains (i.e., professional and informal/social) in which our participants with hidden disabilities make decisions about whether, when, and how to disclose their disability identities. Professional domains include academic or job settings, whereas informal/social domains include online (particularly social media) and in-person settings (e.g., family and friends). We observed that our participants exhibited various types of identity disclosure behaviors in these two domains. We adopted MacDonald-Wilson et al.’s definitions for these different disclosure behaviors:
Forced disclosure (e.g., students with disabilities requesting accommodations)
Selective disclosure (i.e., a person with disabilities chooses when and whom to disclose which aspects of the disabilities)
Nondisclosure (i.e., a person with disabilities chooses not to disclose any aspects of the disabilities)
Disclosure by others (e.g., a friend or family member discloses one’s disabilities often without one’s permission)
Forced disclosure refers to a situation where individuals are “required by circumstances” to disclose their disability identity to another person, typically an employer or a supervisor, “or in the need for accommodation” . Selective disclosure denotes “sharing information with specific or a limited number of people, or sharing specific or limited information with others” . MacDonald-Wilson et al. also consider selective disclosure as being “used to access protections under the ADA [Americans with Disabilities Act] while minimizing risks related to stigma, and allows the person the option to ‘blend in’ or ‘pass for normal’” . Nondisclosure means “a choice made by individuals to keep private any information” regarding their disability identity, which they mention “may result in additional stress and lower self-esteem because one is hiding an aspect of one’s life, but it also protects the individual from potential stigma and discrimination” .
We found that disclosure by others was more prevalent in informal social settings. For instance, one participant explained an unpleasant experience: “Yeah, I think the experiences that I have mostly had with that-with people telling other people without my permission-is with my family. I know that they have good intentions, but I’d prefer not to go to a family gathering and have everyone come up to me saying, ‘Oh, I’m sorry you had a breakdown yesterday.’ That’s a little awkward for me, not knowing where that information is going to. I know that my mom tells her twin sister, they’re really close so she feels like she can share that kind of information. But when it comes back to me, or she texts me, they have good intentions, but I’d prefer to be able to come to them when I feel like I want to.” As noted, this participant did not mind sharing her disability condition but did not accept nonconsensual disclosure because she lost the agency in controlling her information.
Academic and professional settings, on the other hand, often exhibit forced disclosure or nondisclosure of their disability identities. Many of our participants felt forced to disclose their disabilities in schools to get accommodation or at workplaces to perform their work and to ensure their supervisors or colleagues that they are not lazy. For instance, one participant shared, “The other times when I’ve had pressure to disclose was when I was at work and I would have crazy anxiety. No one could tell because I’m extroverted introvert, no one could tell. So I’d be like, ‘ok, I have anxiety.’ Or I’m not meeting deadlines, and I know I have a problem, but again, I don’t have a formal ADHD diagnosis so I don’t really know how to explain it. There are times when I really think I’ve gotten in some trouble because of my disability, but I wasn’t able to advocate because I wasn’t able to express that, you know, I’m not lazy.” Sometimes, our participants experienced the opposite where they felt “forced” not to disclose their disabilities by their families, who believed that disclosing the disabilities would, for instance, negatively affect one’s chance of employment.
Our participants also disclosed specific disabilities and other marginalized identities differently. In general, the participants were comfortable disclosing their physical disabilities more than their hidden disabilities, citing stigma associated with hidden disability identity disclosure. One participant said, “I post about when I’m in pain, physically. I don’t post about being emotionally in pain, but I have posted about my history with suicide a little bit.” Some participants segmented social media platforms and used different platforms/accounts for different purposes. For instance, one participant had three Instagram accounts and used one of them, a private account, to relate to her disabilities. She explained, “For Instagram, I actually have three accounts for Instagram. I have a public account that I rarely ever post on, and I rarely ever look at, but it’s just nice pictures of me doing things. And then I have a private Instagram account, where I follow a lot of chronic illness accounts, I don’t post anything personally about my own chronic illnesses, usually, but I’ll follow and comment and stuff. Then I also have a service dog one that’s private where I can talk to other service dog handlers, but both of those are closed so I have control over who I allow to see those.” Even though she shared nothing about her disability, she understood that the pages she follows can be visible to others and hence preferred another account rather than using the same one. Anonymity on these accounts also helped. If the target audience did not know about their real identities, the participants felt more comfortable sharing about their disability. In terms of future design, designers should explore ways to further support these users to avoid forced disclosures and to facilitate selective and nondisclosure as well as “blending in” or “passing” if these users choose to.
13.2.6 People with Other Disabilities
While there is a large body of research that demonstrates various web accessibility challenges for people with other disabilities (e.g., motor impairments, cognitive impairments), little is known about their privacy challenges and needs. There is some empirical evidence that people with Down syndrome struggle with remembering mnemonic passwords  and people with intellectual disabilities had difficulties remembering passwords . Future research is needed to uncover additional privacy challenges for these user groups.