W tym AI nie pomoże. Praktyka prowadzenia badań z niepełnosprawnymi użytkownikami.
About
Niepełnosprawności nie da się symulować. Modele AI są szkolone bez uwzględnienia potrzeb osób z niepełnosprawnościami.
Znajomość zasad WCAG jest przydatna, warto używać ręcznych i automatycznych narzędzi do testowania dostępności, ale nic nie zastąpi testowania z prawdziwymi, niepełnosprawnymi użytkownikami.
Nasze demo obejmie przykłady aplikacji zawierających nieoczywiste problemy, które mogą być wykryte tylko przez prawdziwego niepełnosprawnego użytkownika.
Podzielimy się naszymi praktycznymi doświadczeniami z badań z udziałem niepełnosprawnych użytkowników, ich zaletami i wyzwaniami z nimi związanymi.
Watch the full talk
Watch this WaysConf session, then continue with related talks or explore the current programme.
Talk in brief
In this WaysConf 2024 presentation, Dawid Górny, Chief Accessibility Officer at Accens, alongside his colleague Darek, shares practical insights on conducting usability research with users with disabilities. Moving beyond theoretical guidelines, they focus on the organizational, logistical, and human aspects of accessibility testing. They discuss how artificial intelligence, while a game-changer, exhibits a "disability bias" because it is trained on average data, making human-centered testing irreplaceable. The speakers outline the unique challenges faced by users with intellectual, hearing, and visual impairments, illustrating these with real-world testing anecdotes. They also address the corporate hurdles of recruitment, workplace adaptation, and travel logistics. Ultimately, they provide a concrete checklist for organizations looking to start accessibility testing, emphasizing that compliance with the upcoming European Accessibility Act (2025) and tapping into a massive global market make accessibility both a legal necessity and a major business opportunity.
Key takeaways
- 01
AI Cannot Replace Human Testers
While AI tools like ChatGPT and Be My Eyes assist disabled users daily, they suffer from "disability bias" and hallucinations because they are trained on average datasets, making real human testing essential.
Watch from 8:54 - 02
Language Barriers for Deaf Users
For many people born deaf, sign language is their first language, meaning written Polish or English is a second language that they must mentally translate, which makes clear, simple text crucial.
Watch from 19:02 - 03
The Pitfalls of Superficial Alt Text
Vague alternative text like "a group of people having fun" fails to provide meaningful context for blind users, proving that alt text must be specific and descriptive to be useful.
Watch from 24:37 - 04
Recruitment and Bureaucratic Hurdles
Hiring disabled testers involves navigating complex legal and bureaucratic obstacles, such as the risk of testers losing their pensions or facing impractical occupational health checks.
Watch from 33:08 - 05
The Business Case for Accessibility
The global market of people with disabilities is estimated to be the size of China's population, presenting a massive economic opportunity alongside upcoming legal mandates like the 2025 European Accessibility Act.
Watch from 31:01
Video chapters
- 0:30Introduction to Accessibility Testing
Dawid Górny and Darek introduce themselves, sharing their personal and professional backgrounds in digital accessibility and testing.
- 4:12The Limits of AI in Accessibility
The speakers explain how artificial intelligence assists disabled users but warn about its inherent "disability bias" and tendency to hallucinate.
- 14:53Testing with Intellectually Impaired Users
This section covers the unique behavioral patterns and UI challenges encountered when conducting usability tests with intellectually disabled individuals.
- 18:42Understanding the Deaf User Experience
The presenters discuss how deaf users interact with written text and digital interfaces, emphasizing the importance of text-based communication options.
- 22:48Common Digital Barriers for Blind Users
Dawid highlights frequent accessibility failures on websites, including poor heading structures, keyboard inaccessibility, and low-quality audio descriptions.
- 28:34Legal and Business Drivers for Accessibility
The talk outlines the upcoming European Accessibility Act of 2025 and the massive global market size of users with disabilities.
- 32:08Logistical and HR Challenges in Research
Darek and Dawid share the practical difficulties of recruiting, employing, communicating with, and transporting disabled testers.
- 43:02A Practical Checklist for Organizations
The speakers conclude with a step-by-step guide on how companies can gradually implement accessibility testing without becoming overwhelmed.
Read edited transcript highlights
These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.
The Disability Bias in Artificial Intelligence
Although artificial intelligence has become a massive daily aid for people with disabilities, it carries a significant risk of exclusion. Research from major universities shows that language models exhibit a "disability bias" because they are trained on vast, averaged datasets. Since accessibility is fundamentally about addressing the needs of non-statistical, non-average users who exist on the margins of typical data, standard AI models often fail to account for their specific scenarios.
Watch from 8:31Designing Interfaces for Intellectual Disabilities
When testing digital products with intellectually disabled users, standard design conventions often fail. For instance, clicking a top-left logo to return to the homepage is an unwritten rule for average users, but an intellectually disabled person may not realize it is a link unless it is explicitly labeled. These users also struggle with dense blocks of text, complex jargon, and implicit dates, often requiring an assistant during testing sessions to help navigate the process safely.
Watch from 14:53The Deaf Perspective on Written Language
People who are born deaf typically communicate using sign language as their primary, native tongue. Consequently, the written national language is actually their second language. When reading websites or emails, they must mentally translate the text into sign language, which can be highly exhausting if the content is overly complex. Furthermore, simple contact details like phone numbers are useless to them unless the organization explicitly states that the number can receive text messages.
Watch from 18:42The Reality of Bad Audio Descriptions
Audio description is vital for blind viewers, but its quality varies wildly. Dawid shares humorous yet frustrating examples from popular streaming shows where poor descriptions ruined the experience. In one instance, a romantic scene was described flatly as "a man groaning," completely changing the tone. In another, a character wearing a "narrow mermaid dress" (syrena in Polish) led Dawid to picture the classic Polish car brand Syrena, demonstrating how literal or poorly chosen words fail to convey the intended artistic imagery.
Watch from 26:32The Complexities of Hiring Disabled Testers
Organizations attempting to recruit disabled users for formal testing face surprising bureaucratic hurdles. In Poland, hiring a person with a disability can sometimes trigger a review that strips them of their social pension. Additionally, standard occupational medicine requirements can be absurdly literal, such as forcing a completely blind tester to undergo an eye examination by an ophthalmologist just to get cleared for computer work.
Watch from 33:08How to Write Effective Alternative Text
During the Q&A session, the speakers address how to write high-quality alternative text without a massive budget. They advise writers to be highly concrete and descriptive rather than abstract. A good alt text should clearly state who or what is in the image, the context, and the surrounding circumstances. They also warn against stuffing alt text with keywords for SEO purposes, emphasizing that its primary function must always be to help a blind user understand the visual content.
Watch from 46:06