Accessibility and inclusive design

Robots are often promoted as tools that will help people with disabilities live more independently, and some do. But robots can also create new barriers: a touchscreen kiosk robot that a blind visitor cannot use, a voice-controlled assistant that does not understand a person with a speech impairment, a delivery robot parked across a curb ramp that a wheelchair user needs. This lesson introduces ways of thinking about disability, the principles of universal and inclusive design, practical accessibility requirements for robot interfaces and bodies, including the Web Content Accessibility Guidelines for screens, and the main accessibility laws in Canada and the United States. Accessibility is not a feature added at the end; it is a quality of the whole design, and it benefits far more people than designers usually expect.

Disability and design

Disability is common. Statistics Canada's 2022 Canadian Survey on Disability found that 27 percent of Canadians aged 15 and over, about 8 million people, had one or more disabilities that limited their daily activities, and US public health surveys similarly estimate that roughly one in four adults lives with some type of disability. Rates rise with age, so as populations age, more of the people who meet robots in workplaces, hospitals and public spaces will have disabilities. Disabilities may be permanent, temporary, such as a broken arm, or situational, such as a worker wearing thick gloves or trying to hear an alert on a noisy floor.

Two models of disability shape design. The medical model locates disability in the person's body, as an impairment to be treated or compensated. The social model, developed by disabled activists and scholars in the United Kingdom in the 1970s and 1980s, distinguishes impairment from disability and argues that people are disabled largely by barriers in the environment: stairs rather than wheelchairs, inaccessible interfaces rather than blindness. For designers, the social model is empowering, because barriers are things we design and can remove. The disability rights movement's slogan "nothing about us without us" also makes a methodological point: people with disabilities should be involved as designers, testers and decision-makers, not only as subjects of study.

Universal and inclusive design

Universal design, a term associated with the architect Ronald Mace, aims to make products and environments usable by all people, to the greatest extent possible, without the need for adaptation. In 1997 a group of architects, product designers and engineers working with Mace at North Carolina State University published seven principles.

Inclusive design, as developed by Jutta Treviranus and her inclusive design research institute (IDRC) at OCAD University in Toronto, emphasizes recognizing diversity and uniqueness, using inclusive processes and tools, and seeing design within a broader system. Rather than designing for an imagined average user and then adding accommodations, inclusive design starts from people at the margins, whose needs reveal problems that affect many others. The curb cut is the classic example: ramps cut into curbs for wheelchair users also help people with strollers, delivery carts, luggage and bicycles. In robotics, a robot that announces its intentions by sound for blind pedestrians also helps sighted people who are looking at their phones.

Accessible robot interfaces

Many robots have touchscreens, apps or web dashboards. These should meet the Web Content Accessibility Guidelines, WCAG, published by the World Wide Web Consortium. The current version, WCAG 2.2, was published in October 2023. WCAG is organized around four principles, summarized as POUR: content must be perceivable, operable, understandable and robust. Each guideline has testable success criteria at levels A, AA and AAA, and laws and policies commonly require level AA. Examples include text alternatives for images, captions for audio, keyboard operability, sufficient color contrast, minimum target sizes, not relying on color alone, predictable navigation, clear error messages and compatibility with assistive technologies such as screen readers.

Color contrast is a good example of a precise criterion. WCAG defines the contrast ratio between two colors using their relative luminance \( L \), computed from linearized red, green and blue values:

Level AA requires a contrast ratio of at least 4.5 to 1 for normal text and 3 to 1 for large text and for meaningful graphical elements such as icons and the boundaries of controls. Status colors on a robot's screen, such as a pale green "ready" label on a white background, often fail this test. Color should also never be the only way information is conveyed: about 8 percent of men of Northern European ancestry have some form of color vision deficiency, most commonly red-green, so red and green status lights should differ in position, shape, text or flashing pattern as well.

# WCAG contrast ratio for status labels on a robot dashboard
import numpy as np
def lum(rgb):
    c = np.asarray(rgb, float) / 255
    c = np.where(c <= 0.04045, c / 12.92, ((c + 0.055) / 1.055) ** 2.4)
    return 0.2126 * c[0] + 0.7152 * c[1] + 0.0722 * c[2]
def ratio(a, b):
    la, lb = sorted([lum(a), lum(b)], reverse=True)
    return (la + 0.05) / (lb + 0.05)
white, dark = (255, 255, 255), (30, 30, 30)
status = {"ready green": (76, 175, 80), "dark green": (0, 100, 0), "amber": (255, 191, 0),
          "dark amber": (153, 102, 0), "red": (220, 38, 38), "blue": (37, 99, 235)}
for name, c in status.items():
    r1, r2 = ratio(c, white), ratio(c, dark)
    print(f"{name:11s} on white {r1:4.1f}:1 {'AA' if r1 >= 4.5 else '--'}   on dark {r2:4.1f}:1 {'AA' if r2 >= 4.5 else '--'}")

ready green on white  2.8:1 --   on dark  6.0:1 AA
dark green  on white  7.4:1 AA   on dark  2.2:1 --
amber       on white  1.7:1 --   on dark 10.1:1 AA
dark amber  on white  4.9:1 AA   on dark  3.4:1 --
red         on white  4.8:1 AA   on dark  3.5:1 --
blue        on white  5.2:1 AA   on dark  3.2:1 --

Beyond screens: bodies, voices and spaces

Robots are physical, so accessibility extends beyond the screen. Physical controls such as emergency stop buttons, start buttons and touch panels must be reachable and operable from a seated position and by people with limited grip strength or dexterity. Building accessibility standards give reach ranges that can guide robot designers.

In the United States, the 2010 ADA Standards for Accessible Design give reach ranges for operable parts: an unobstructed forward or side reach between 15 and 48 inches (about 380 to 1220 mm) above the floor, with a clear floor space of 30 by 48 inches (about 760 by 1220 mm) in front of the control. Robot designers can use these as a guide for kiosks, lockers and controls that people operate from a wheelchair.

Speech interfaces must handle diverse voices. Speech recognition systems have been shown to make more errors for some groups, including speakers with certain accents and dialects and people with dysarthria or other speech impairments. A 2020 study by Allison Koenecke and colleagues found that several commercial recognizers made roughly twice as many errors for Black American speakers as for white speakers in their samples. Robots that rely on speech should offer alternatives, such as buttons, an app or gestures, and developers should test recognition with the full range of intended users. For Deaf and hard-of-hearing users, robots should show spoken output as text and avoid sound-only alerts; some Deaf people use sign languages such as American Sign Language or Langue des signes québécoise, which are full languages, not gestures added to English or French.

Mobile robots in public spaces affect people who never chose to interact with them. In 2019 a doctoral student who uses a wheelchair, Emily Ackerman, described being trapped in a street in Pittsburgh when a sidewalk delivery robot sat at the top of a curb cut, blocking her way onto the sidewalk as traffic approached. Incidents like this show that robots must treat curb ramps, tactile paving, doorways and narrow passages as spaces to keep clear, must yield to people using mobility aids, guide dogs or white canes, and must be tested with disabled pedestrians. Cities that permit sidewalk robots increasingly impose conditions on their behavior.

Cognitive accessibility matters too. People with intellectual disabilities, dementia, brain injuries or limited literacy, and anyone under stress, benefit from plain language, consistent layouts, step-by-step guidance, forgiving error handling and the ability to take time. WCAG 2.2 added success criteria aimed partly at cognitive accessibility, such as accessible authentication that does not depend on memorizing or transcribing codes, and avoiding redundant entry of information.

Accessibility law in Canada and the United States

In the United States, the Americans with Disabilities Act of 1990 (ADA) prohibits discrimination on the basis of disability in employment (Title I), state and local government services (Title II) and public accommodations such as shops, hotels and hospitals (Title III). In 2024 the Department of Justice issued a rule under Title II requiring state and local government websites and mobile apps to meet WCAG 2.1 level AA, with compliance dates phased by government size; check the current dates, as they may change. Courts have applied Title III to websites and digital services in many cases, and Section 508 of the Rehabilitation Act requires federal agencies' information and communication technology to be accessible, using standards that incorporate WCAG 2.0 level AA. Employers must provide reasonable accommodation, which can include adapting robot workstations. In Canada, the Accessible Canada Act and provincial laws such as Ontario's AODA play the corresponding roles.

These laws rarely mention robots specifically, but they apply to robots through the services and workplaces where robots are deployed. A hospital that uses a robot kiosk for check-in must make check-in accessible; an employer that introduces a collaborative robot must accommodate workers with disabilities. Designing for accessibility from the start is far cheaper than retrofitting.

Testing with disabled users

Accessibility checklists catch many problems, but testing with people with disabilities catches problems checklists miss. Recruit participants with a range of disabilities relevant to the robot, pay them fairly for their expertise, make the study location and materials accessible, and allow extra time. Automated tools can check screen contrast and code-level issues, and screen-reader testing should be done by experienced users. Study designs from lesson 8 apply, with care: small, diverse samples of disabled users may not support statistical comparisons, and qualitative methods such as think-aloud sessions and interviews are often more informative.

<!-- Accessible robot status panel fragment (listing) -->
<section aria-labelledby="status-h">
  <h2 id="status-h">Robot status</h2>
  <p role="status" aria-live="polite">
    <span class="icon" aria-hidden="true">&#9654;</span> Moving to Station 4, ETA 2 minutes
  </p>
  <button type="button" class="stop" aria-describedby="stop-help">Stop robot</button>
  <p id="stop-help">Stops the robot at once. You can also press the red button on its side.</p>
</section>