Exclusion by Design - Michele A. Williams - CourtneyR.dev's 2nd Brain

Exclusion by Design - Michele A. Williams ![rw-book-cover](https://d34adp677peecb.cloudfront.net/static/images/article0.00998d930354.png) 🔗 **Source:** [Exclusion by Design - Michele A. Williams](https://www.youtube.com/watch?v=rTlzxDLkg8I) > [!note]+ Document Notes > tags: [accessibility, ableism, disability-inclusion, inclusive-design, ux-research, representation] > author: Michele A. Williams > source: Accessibility Talks (A11yTalks) > reference: https://www.youtube.com/watch?v=rTlzxDLkg8I > date created: 2026-10-06 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: Michele A. Williams / Exclusion by Design (A11yTalks, September 2026) >  ## 🧭 Table of contents - [📝 Summary](#-summary) - [📖 Main content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning styles assessment](#-learning-styles-assessment) - [📊 Accessibility assessment](#-accessibility-assessment) - [🌍 DEIB assessment](#-deib-assessment) - [✅ Logical validity](#-logical-validity) - [🛡️ DARVO analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🔄 Refactor Appendix](#-refactor-appendix--michele-a-williams--exclusion-by-design) ## 📝 Summary - **What happened.** A11yTalks published a 1:01:15 session (uploaded 2026-09-30; the listing was titled "(premiere)" before it published). Host Andrew Olsen, an engineer at Principal Financial per his own intro, opens it. Michele A. Williams (UX researcher and accessibility consultant, company "making accessibility work", author of *Accessible UX Research*) speaks from 00:03:39 to about 00:47:19, then takes three audience questions (00:50:12 to 00:58:56). The host's intro, housekeeping, reaction and close are kept in separate Notes entries. - **The claim.** Disability disappears from products before design starts, because it's missing from the pictures, personas, journey maps, user stories, product visions and AI-generated concepts that teams build from (00:11:37 to 00:12:21, 00:22:36). She adds that omission is sometimes deliberate, because ableism teaches that disability doesn't belong (00:16:45). Her fix is a four-step cycle: notice, question (name ableism), interrupt, normalize (00:43:40), with role-specific actions that take initiative and not permission (00:24:03 to 00:29:51). - **Why it matters.** It gives you a five-question audit of upstream artifacts (00:22:36) and a list of actions sorted by job role (researcher, designer, developer, PM). Both transfer directly to WordPress teams that write personas, mockups, Trac tickets and Learn lessons before any code exists. - **Key stakeholders.** Product, research, design, engineering and program staff; disabled users and colleagues; stock-photo and content teams; research recruiters; AI-concept tooling. Named people: Stella Young, John Lewis, Dr. Cynthia Bennett, Dr. Ashley Shew (caption: "Shu"). - **Risks and tradeoffs.** The talk cites no statistic, no WCAG version or success criterion, and no law or regulation. Its evidence is the speaker's own anecdotes (coffee machine, students versus professors, a social media manager, one post-layoff note). It never defines the title phrase "exclusion by design". The "accomplice" framing carries job risk that she acknowledges once, in one sentence (00:29:07). - **Bottom line.** The audit questions, role actions, screener wording and compensation guidance are specific and checkable. The causal story (imagination gaps and ableism produce exclusion) is plausible and unproven in the talk. Treat it as a practice playbook, not as research evidence. ## 📖 Main content Timestamps mark the start of a caption paragraph. An event can sit up to about 45 seconds after the stamp. ### Thesis and how the title is used The title phrase is not defined in the talk. The host reads the full title at 00:02:13: "exclusion by design. How disability gets erased from concept to product and how to fix it." Williams says at 00:03:39 that she titled it "exclusion by design" and begins "Here's what I mean," but the caption around that sentence is garbled ("I want to talk about what we talk about"). The working definition has to be assembled from three claims: 1. **The pipeline claim (00:11:37 to 00:12:21, 00:15:16).** Personas, journey maps, AI-generated content, product visions and user stories are "the foundations that lead to what we build"; if a perspective is missing there, it's missing downstream. "We can't design for who we haven't imagined." 2. **The cause claim (00:16:00 to 00:16:45).** Teams give "didn't know," "didn't think about it" and "don't need to" as explanations. She calls them excuses and offers a "harder truth": sometimes disability is intentionally left out because society teaches it doesn't belong, which she names ableism. 3. **The self-reinforcing claim (00:21:08 to 00:23:20).** Where exclusion is the norm, access feels exceptional, and where disability is seen as a problem, it's rarely part of the solution. Her closing reframe: disability isn't the problem, the viewpoint is. ### Segments **Noticing (00:03:39 to 00:12:21).** She states her language choice (identity-first: "disabled people") and tells how a mother's civil engineering firm and her own internship there taught her to read architectural renderings for who gets imagined in them. Working in accessibility changed what she noticed: cobblestones for mobility aid users (00:06:32), the interiors of high-speed trains (00:07:16), back-to-school commercials and stock photos (00:08:00 to 00:08:44), and a touchscreen coffee machine mounted high on a counter (00:09:26 to 00:10:10). A colleague's reply, that noticing it was "so typical of you," leads to her point that she was catching up with what disabled people already knew (00:10:55) and that she learned accessibility from disabled colleagues, research participants and friends more than from WCAG (00:11:37). **Students versus professors (00:12:21 to 00:15:16).** In a Boston office, her team prioritized student-facing software over educator-facing software for accessibility work, citing the ratio and turnover of students to instructors (00:13:49). Colleagues on the instructor side felt relieved. When she said professors can be disabled too, they went silent: it hadn't occurred to them. She adds that students with disabilities graduate and may become instructors, and that people acquire disabilities across a lifetime (00:14:32). **Why exclusion persists (00:15:16 to 00:24:03).** Representation shapes imagination, imagination shapes scenarios, scenarios shape products (00:15:16). She then defines ableism and lists three everyday behaviors that present as polite: lowered expectations ("he's blind but goes to work every day," 00:18:11), the same technology framed as innovation for some and as accommodation for others (Audible versus a student with dyslexia, 00:18:56 to 00:19:41), and "inspiration porn" (a term she credits to Stella Young; exoskeleton-walk videos, 00:20:25). She ends the segment with five audit questions about research, strategy, personas, personas that acquire disability, and AI (00:22:36 to 00:23:20). **Interrupting (00:24:03 to 00:34:11).** Inclusion rarely happens by accident, so she asks the audience to act as accomplices, with a nod to John Lewis's "good trouble" (00:24:03 to 00:24:46). She splits actions into "subtle" and "rogue" by role, covering researchers (00:25:29 to 00:26:57), designers (00:26:57 to 00:28:23), developers (00:28:23 to 00:29:07) and everyone else (00:29:51). A three-repository stock-photo segment follows (00:30:35 to 00:34:11). **Limits of allyship and shared responsibility (00:34:11 to 00:37:52).** Accomplices shouldn't constantly ask "is this accessible" without context, use disability as a gotcha, speak over disabled people or assume one disabled person speaks for all (00:34:11 to 00:34:56). She quotes Dr. Cynthia Bennett from her book and says the work shouldn't depend on disabled people being in every room (00:35:39 to 00:37:07). **Normalizing (00:37:52 to 00:46:35).** A social media manager's "I added captions before I knew you would ask" (00:37:52 to 00:40:01) becomes her model of normalization: people ask about captions, disabled participants and accessible meetings without prompting (00:40:01). She expects push-back about time, says you won't convince everyone because "everyone else is listening," and tells of a colleague who picked up her advocacy after layoffs (00:41:29 to 00:42:12). Her mechanism for norm change is a chain of repetition across three people (00:42:54). She recaps notice, question, interrupt, normalize (00:43:40), calls for intentional learning (00:44:24 to 00:45:50), and closes: exclusion is learned, so inclusion can be learned (00:46:35). **Q&A (00:50:12 to 00:58:56).** Three questions: examples of disability questions in research (00:50:12), recruiting and compensating disabled participants (00:53:06), and capturing intersections of disability with race, age and socioeconomic status (00:56:45). Details are in Notes. ### Concrete examples, in order | Time | Example | Type | Population or barrier named | |---|---|---|---| | 00:04:22 | Civil-engineering renderings of houses and communities, later with people added | Concept art | Who gets imagined: ages, cultures, genders | | 00:06:32 | Waterfront community rendering (she believes the Netherlands) with a cobblestone or brick road and a cyclist | Built environment | People using mobility aids | | 00:07:16 | High-speed (bullet) train concepts that "looked like a spaceship"; interiors with narrow aisles and traditional seats | Transport | Wheelchair users; people whose bodies don't fit the seating (caption garbled) | | 00:08:00 | Old Navy back-to-school commercials, diverse by culture, race and gender identity | Advertising | Disabled children and youth, rarely shown | | 00:08:44 | Business stock photos, diverse by gender, age, culture and race | Stock imagery | Abilities absent | | 00:09:26 | Denver-office coffee machine: touchscreen, high on the counter, controls higher still | Appliance interface | Low vision or blind users (touchscreen); reach and height (mobility); wheelchair users (no roll-up space) | | 00:13:04 | Student-facing versus educator-facing software, Boston office | Software prioritization | Disabled instructors; students who become instructors; people who acquire disabilities | | 00:18:11 | "He's blind but goes to work every day," said at an awareness booth | Language pattern | Blind adults; employed adults | | 00:18:56 | Audible's "listen to your books" marketing versus a student with a reading difficulty such as dyslexia | Product framing | Students who need text-to-speech | | 00:20:25 | Exoskeleton-walks-across-stage videos; Stella Young's TED talk "I'm not your inspiration" | Media | People who use canes, walkers, wheelchairs | | 00:23:20 | AI-generated concepts built on training data where exclusion is the norm | AI tooling | Disability in generated output (no example output shown) | | 00:28:23 | Drag and drop with documented keyboard functionality; touchscreens with an audio interface | Interface patterns | Keyboard users; people who can't see the screen | | 00:29:07 | Icon buttons with accessible names in web code; screen reader testing before handoff | Code and test practice | Screen reader users | | 00:31:18 | All In Frame photo: a woman at a laptop with a guide dog beside her; a standing colleague with dwarfism at the seated person's eye level | Stock photo | Guide dog user; people with dwarfism | | 00:32:01 | Disabled And Here photo: a man signing outdoors and smiling; repository also includes queer representation and varied cultures and races | Stock photo | Sign language users | | 00:33:28 | "Pixels" photo: a young woman with Down syndrome in a yoga studio | Stock photo | People with Down syndrome | | 00:37:52 | Social media manager adds captions before being asked | Practice change | Deaf and hard of hearing audiences (she notes captions help many people) | | 00:52:22 | Slack's intentional design for screen reader users, "in the news in a good way" | Product feature | Screen reader users | | 00:58:12 | Deaf users, native signers, as an under-studied research population | Research gap | Deaf users: contacting a company, content understandability | | 00:58:56 | Neurodivergence, from a talk she just attended | Research gap | Neurodivergent users | ### Standards, laws and frameworks cited - **WCAG:** one mention, at 00:11:37, with no version, level or success-criterion number. She uses it as a contrast: she learned accessibility from disabled people, "from engaging with disabled people", rather than from standards. [Inference: she means WCAG 2.x, but she doesn't say.] - **Laws and regulation:** none about accessibility. No ADA, Section 508, EN 301 549 or European Accessibility Act. The only legal reference is a US civil-rights-era analogy: discrimination against Black citizens was "even the law" and you "almost break the law" to disrupt it (00:24:46). - **Named concepts and frameworks:** identity-first language (00:03:39); her definition of ableism as "a disregard or a lessened view of disabled people" (00:16:45); inspiration porn (Stella Young, 00:20:25); "good trouble" (John Lewis, 00:24:03); "accomplice" versus ally (00:24:03); technoableism, via the book by Dr. Ashley Shew (caption "Shu"; 00:54:34); a four-part cycle called "pillars" at 00:37:07 and 00:37:52 (notice, question, interrupt, normalize; recap at 00:43:40); two research veins from her book, accessibility-focused research and research inclusive of disability (00:51:40); a cumulative knowledge-base approach to research (00:56:45); a consultant-versus-participant distinction for compensation (00:54:34). - **Not named by her (my inference, flagged):** the social model of disability fits "disability is not the problem, it's our viewpoint" (00:23:20); the curb-cut effect fits "captions benefit so many people" (00:40:01). She names neither. ### Statistics and their sources She states no statistic with a source. The only quantities: | Time | Quantity | Source named | |---|---|---| | 00:26:14 | "how prevailing disability is and how much it is present in our population", used to argue you've likely had a disabled participant already | None; no figure given | | 00:13:49 | "a ratio" of students to instructors, with student turnover, used to prioritize student-facing software | None; no figure given | | 00:51:40 | A participant "increasing their font on their mobile device by 25" | None; an illustration, with no unit stated | | 00:30:35 | "three in particular" stock-photo repositories | Self-evident count | | 00:48:02 | Host: she spoke "four years ago in September of 2021" | None; see the date note under Questions | ### Research methods described - **Screener (00:25:29, 00:50:56).** Add a disability question to the participant screener; offer checkboxes by category (her examples: vision, mobility), plus an open-ended description field; report the data in findings. - **Assistive-technology and adjustment questions (00:50:56 to 00:51:40).** Ask about assistive technology or device use without asking about disability, for teams uneasy about collecting disability data. Her example is a participant who raised font size on a phone. - **Unprompted accessibility-focused studies (00:26:57).** She ran them "whether prompted to or not" and presented the findings widely; she says she was often employed as an accessibility researcher, so this wasn't fully rogue. - **Three study shapes (00:51:40 to 00:53:06).** Accessibility-focused foundational research with many disabilities, "similar to how you might do an audit but from a user's perspective"; feature-focused research (Slack's screen reader work as the example); and "the dream," disabled participants included in all research, so every protocol gathers their feedback on every task. - **Recruitment (00:53:50).** A recruitment firm; the usual channels plus the screener; relationships with disability organizations, after-school organizations and volunteer organizations. - **Compensation (00:53:50 to 00:55:19).** Standard fair pay; extra for transportation or longer sessions; consultant-level pay and credit for people doing ideation or deep-dive design, including a path to hire. - **Cumulative sampling (00:56:45 to 00:58:56).** Treat research as a knowledge base; every study adds participants the last one lacked; periodically review who you've recruited historically. - **Her own epistemology (00:11:37).** She describes learning from colleagues, participants and friends. It's participant-observer learning, not a formal method. ### Recommendations, in the order she gives them 1. **00:11:37** Learn accessibility by engaging disabled people (colleagues, participants, friends), not only through standards. 2. **00:11:37 to 00:12:21** Check personas, journey maps, AI-generated content, product visions and user stories for missing perspectives. 3. **00:14:32** Don't treat the second product (instructor-side) as exempt; people acquire disabilities and students become instructors. 4. **00:17:27 to 00:21:08** Recognize three ableist patterns (lowered expectations, "innovation versus accommodation" framing, inspiration porn) and shift the viewpoint to "there shouldn't be so many obstacles." 5. **00:22:36 to 00:23:20** Ask: did foundational research include disabled participants before direction was set; who influenced the product strategy; who do personas envision; when did a persona last acquire a disability; what does AI concept generation do when exclusion is the norm. 6. **00:24:03** Interrupt; act on initiative, because inclusion rarely happens by accident; act as an accomplice. 7. **00:25:29 to 00:26:57** Researchers: add disability questions to screeners, report them, and run accessibility-focused studies and share them widely. 8. **00:26:57 to 00:28:23** Designers: annotate design files (heading levels), apply low-hanging considerations in hardware (colors in appliances and vehicles), document keyboard function for drag and drop, and pair touchscreens with an audio interface. 9. **00:29:07** Developers: give icon buttons accessible names; test with a screen reader or other assistive technology and fix issues before handoff. She adds a caution: don't do this in ways that cost you your job. 10. **00:29:51** Add captions before being asked; PMs add accessibility acceptance criteria; put accessibility findings in research and heuristic reports; use inclusive imagery in concept renderings. 11. **00:30:35 to 00:34:11** Use disability-inclusive stock photo repositories (All In Frame, Disabled And Here, "Pixels") and their alt text. 12. **00:34:11 to 00:34:56** Don't ask "is this accessible" without context, don't use disability as a gotcha, don't speak over disabled people, don't assume one person represents everyone; create space and share influence. 13. **00:36:24 to 00:37:07** Draw on disabled colleagues' perspectives, and don't require disabled people to be in every room. 14. **00:40:01** Normalize by expecting captions, recruited disabled users, accessible meetings and alt text without prompting. 15. **00:41:29 to 00:43:40** Expect push-back, accept that you won't convince everyone, and stay the course. 16. **00:43:40 to 00:46:35** Notice, question, interrupt, normalize; keep learning intentionally, with research as "the biggest pillar," plus talks, books and disabled people on social media; aim for presence in conversations, not perfection. 17. **00:50:56 to 00:51:40** Screener with checkboxes and an open field; ask about assistive technology and adjustments separately from diagnosis. 18. **00:51:40 to 00:53:06** Choose among foundational, feature-focused and fully inclusive protocols. 19. **00:53:50** Recruit through firms, ordinary channels with screening, and disability organizations. 20. **00:54:34 to 00:55:19** Pay for transportation and extra time; treat ideation and deep-dive design contributors as consultants with attribution and a path to hire. 21. **00:56:45 to 00:58:56** Build a research knowledge base, plan recruitment across studies, and review who you've historically recruited; add deaf and neurodivergent perspectives. ### Garbled terms and likely corrections (my inference, not verified) | Caption | Likely meaning | Time | |---|---|---| | Ali Talks / Alley Talks / "comp11 tks" / "teamallettalks.com" | A11yTalks; linkedin.com/company/a11ytalks; [email protected] | 00:00:00, 00:59:38, 01:00:22 | | "Alycat Club" | A11yCat Club (monthly members) | 00:00:42 | | "Michelle" | Michele (title spelling; she says "one L" at 00:46:35) | throughout | | "Dr. Cynthia Bennett" / "Cindia" | Cynthia Bennett | 00:35:39, 00:36:24 | | "accessessibility is narrow" | the Bennett quote is unreliable as captured | 00:36:24 | | "Levant consulting group" | producer of All In Frame; spelling unverified | 00:31:18 | | "disabled and here h e r e" | Disabled And Here, spelled out | 00:32:01 | | "pixels" | possibly Pexels | 00:32:44, 00:33:28 | | "Dr. Ashley Shu", "technoabbleism" | Ashley Shew, technoableism | 00:54:34 | | "abbleism" | ableism | 00:16:45 onward | | "riding on the pike" | riding a bike | 00:06:32 | | "Why does disability inclusion keep happening?" | exclusion (the sentence reads as a contradiction otherwise) | 00:16:00 | | "mawconultinglc.com" | mawconsultingllc.com (unverified) | 00:46:35 | | "Key Wells and Dan Edge" | names unverified | 00:48:46 | | "Vikingsque", "dark ribed glasses" | host's description; "dark-rimmed" is plausible | 00:00:00 | ## 🗒️ Notes #### ⏱️ [00:00:00](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=0) Host intro, visual description and code of conduct (host) - Andrew Olsen welcomes viewers to the monthly virtual meetup on accessibility, inclusion and usability, and says he's an engineer at Principal Financial. - He gives a visual self-description. Host (Andrew Olsen), not the speaker: > [!quote] [00:00:00] > I am a white man with long red Vikingsque hair and wearing a dark ribed glasses and a blue shirt. - He reminds everyone of the code of conduct, which he says applies in chat and during the conversation (00:00:42). #### ⏱️ [00:00:42](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=42) Host housekeeping and sponsors (commercial context, not content) - Thanks bronze supporter Principal Financial Group (his employer) and monthly club members; asks for sponsorships that go toward compensating speakers and improving events (00:01:28). - Mentions employer matching through Benevity. Kept out of the summary. #### ⏱️ [00:02:13](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=133) Host: how to ask questions, speaker intro, fun fact (host) - Question routes: live premiere chat, YouTube comments after, LinkedIn and other social channels. - Introduces Michele A. Williams, "UX and accessibility consultant at making accessibility work," and the full title and subtitle. Fun fact: she's an only child, "but not a spoiled one" (00:02:56). #### ⏱️ [00:03:39](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=219) Framing and disability language (speaker) - Says the talk is less technical and more about mindfulness and thoughtfulness. > [!quote] [00:03:39] > I tend to use identity first, which means using phrases like disabled people rather than people first as in people with disabilities. - Her reason (00:04:22): she aligns with people who treat disability as a descriptor, not a word to avoid. #### ⏱️ [00:04:22](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=262) Origin story: renderings of houses and communities - A parent's civil engineering firm and her own internship there. Renderings later added people; growing up on the East Coast in a revitalizing area, she looked for diversity of ages, cultures and genders (00:05:05). - Those are the usual DEI characteristics, so mindful representation shows up for them. Accessibility work changed what she noticed (00:05:49). - Examples: waterfront rendering and cobblestones (00:06:32); bullet-train exteriors and interiors (00:07:16). #### ⏱️ [00:08:00](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=480) Everyday gaps: commercials and stock photos - Old Navy back-to-school ads show diversity in culture, race and gender identities but not disability, "at least not often" (00:08:00). Business stock photos likewise (00:08:44). - The hedge ("not often," "not necessarily") matters: she gives no count. #### ⏱️ [00:08:44](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=524) The coffee machine - She was visiting the Denver office from Boston. Her check (00:09:26 to 00:10:10): the touchscreen interface (vision), the height of the machine and its controls (mobility, and users' height), and the position on the counter (no roll-up space for a wheelchair). She also notes that a self-serve single-cup machine can help independence. > [!quote] [00:09:26] > the coffee is okay, but the machine's not accessible. - The reply she got (00:10:10 to 00:10:55): that's "so typical of you" and nobody else would point that out. > [!quote] [00:10:55] > So what was happening basically I was noticing what disabled people already knew. #### ⏱️ [00:11:37](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=697) Learning from people, and what's missing from the foundations - Her stated source of learning: colleagues, research participants and friends who are disabled. - The upstream list (00:11:37 to 00:12:21): personas, journey maps, AI-generated content or concepts, product visions, user stories. #### ⏱️ [00:12:21](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=741) Students versus professors - Setting: the Boston office kitchen, a conversation with product-side colleagues she didn't work with. - Her team prioritized student-facing software because of the student-to-instructor ratio and turnover (00:13:49). Product-side colleagues who didn't work on the student-facing software lamented for those who did (00:13:04), and instructor-side colleagues showed relief that they didn't have to do that work (00:14:32). > [!quote] [00:14:32] > I interjected and I pointed out that professors can be disabled too. - The silence that followed was surprise, not disagreement, in her account. Her follow-up: students graduate and may become instructors; people acquire disabilities across their lives (00:14:32 to 00:15:16). #### ⏱️ [00:15:16](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=916) Representation, and why exclusion keeps happening - Representation shapes who and which scenarios we picture, then the design. > [!quote] [00:15:16] > So we can't design for who we haven't imagined. - Common explanations she calls excuses: didn't know, didn't think about it, don't need to. She grants "a little bit of grace" (00:16:00 to 00:16:45). > [!quote] [00:16:45] > sometimes disability is intentionally left out because society teaches us that disability doesn't belong. #### ⏱️ [00:16:45](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1005) Ableism and three polite-sounding patterns - Short definition: discrimination or prejudice against disabled people; fuller: "a disregard or a lessened view of disabled people." - Lowered expectations (00:18:11): she answered the "goes to work every day" remark with "adults go to work." > [!quote] [00:18:56] > there shouldn't be so much so many obstacles to doing ordinary tasks simply because someone has a disability. - Audible's efficiency framing versus a student told they aren't reading if they use audio for dyslexia (00:18:56 to 00:19:41). The student's treatment is described as what "is going to be told," a constructed scenario with no case cited. - Inspiration porn (00:20:25): she credits Stella Young, in a TED talk, with coining the term; the talk title she gives is "I'm not your inspiration. Thank you very much" (00:21:08). > [!quote] [00:21:08] > Basically, if exclusion is the norm, access will always feel exceptional. #### ⏱️ [00:22:36](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1356) Missed opportunities: the five audit questions - Research: how many foundational studies included disabled participants before direction was decided. - Strategy: who had the opportunity to influence it. - Personas: who they envision, and when one last acquired a disability. - AI: if it generates from what exists, what happens when exclusion is already the norm (00:23:20). She notes "a lot of controversy" and that companies are doing it anyway. - Her conclusion: disability isn't the problem, "it's our viewpoint." #### ⏱️ [00:24:03](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1443) Interrupt: becoming an accomplice - "Inclusion rarely happens by accident." > [!quote] [00:24:03] > many accessibility improvements don't necessarily require permission. They require initiative. - Why "accomplice": someone who actively disrupts an unjust status quo, and she cites John Lewis's "good trouble" and the civil-rights-era analogy (00:24:46). Actions are "subtle or rogue" (00:25:29). #### ⏱️ [00:25:29](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1529) Researchers, designers, developers - Researchers (00:25:29 to 00:26:57): screener questions, reporting, unprompted studies. > [!quote] [00:26:14] > you've likely had someone with a disability in your research study whether you knew it or not - Designers (00:26:57 to 00:28:23): annotate design files; hardware colors; documented keyboard drag and drop; audio interface with touchscreens. - Developers (00:28:23 to 00:29:07): accessible names for icon buttons; screen reader testing before handoff. The caution she attaches: > [!quote] [00:29:07] > I'm not saying do this in ways that will obviously get you in trouble or have an impact on job security #### ⏱️ [00:29:51](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1791) Everyone: just do the thing - Add captions before asked; PMs add acceptance criteria; findings in reports; inclusive imagery in renderings. > [!quote] [00:29:51] > Basically we don't need permission to disrupt exclusion. #### ⏱️ [00:30:35](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=1835) Disability stock photo repositories - All In Frame: "very new," produced by an organization the caption renders as "Levant consulting group" (00:31:18). Photo described: woman at a laptop with a guide dog; colleague with dwarfism standing at the seated woman's height. Goal: "neutrality and the lack of fanfare" (00:32:01). - Disabled And Here: free set, includes queer representation and varied cultures and races; man signing outdoors (00:32:01 to 00:32:44). - "Pixels" (possibly Pexels): she found it while building her own website; photo of a young woman with Down syndrome doing yoga (00:33:28). - Both of the first two come with alternative text she says you can reuse (00:32:44). She names no licenses, so check terms before reuse. #### ⏱️ [00:34:11](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2051) What accomplices shouldn't do - Every quiet demonstration weakens arguments against inclusion (an unsupported universal; see Logical validity). > [!quote] [00:34:11] > We're not trying to trip up our colleagues. We're trying to advocate. - Don't speak over disabled people or assume one experience represents everyone; "create space and also share influence" (00:34:56). #### ⏱️ [00:35:39](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2139) Dr. Cynthia Bennett and shared responsibility - She says she is non-disabled and wanted representation in her book beyond research participants: > [!quote] [00:35:39] > As I was writing the book, I being a non-disabled person wanted to make sure that I had some disability representation in the book. - Bennett quote as captured (garbled, so treat as unverified): > [!quote] [00:36:24] > accessessibility is narrow. Whatever they have learned about it, accessibility is there but often implemented in a way that people who don't use the accessibility feature would understand. - Her takeaway, as captured: > [!quote] [00:36:24] > accessibility is not for is not speaking for people. It's helping people be heard. - The work should come from the disability community's perspective but doesn't require disabled people in every room (00:37:07). #### ⏱️ [00:37:52](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2272) Normalize: the captions story - She calls normalize the fourth pillar and then says she doesn't remember the fourth; the captured list ("notice interrupt") is garbled, and the full four appear only in the recap at 00:43:40 (00:37:07). Context: a client engagement during COVID, an in-person conference green room in California, a social media manager preparing a post (00:38:34 to 00:39:17). > [!quote] [00:37:52] > I added captions before I knew you would ask. - She reads the remark as acknowledgment, not frustration, and predicts the manager won't forget it, though she doesn't know if they're still at the company (00:39:17 to 00:40:01). - Normal looks like asking, unprompted: are captions included, did we recruit disabled users, is the meeting accessible, is there alt text (00:40:01). #### ⏱️ [00:40:44](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2444) Push-back and the layoffs note - Accessibility is "not just the products, it's the world" (00:40:44). > [!quote] [00:41:29] > sometimes you won't convince the person you're talking to, but everyone else is listening. - At a company where she led this work, a colleague routed around her; after layoffs, a surviving colleague wrote that they'd carry on the advocacy (00:42:12). > [!quote] [00:42:12] > the thing is culture changes quietly and this is how we get to it. - The chain (00:42:54): one person asks, a second repeats it, a third expects it, then it's standard practice. #### ⏱️ [00:43:40](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2620) Recap, intentional learning, close - Recap: notice (take in disabled people's perspectives), question (acknowledge ableism), interrupt, normalize. - "You don't need to know every disability, but you do need to keep learning" (00:44:24). > [!quote] [00:45:07] > The goal isn't perfection in anything that we do. The goal is making sure disability doesn't disappear from conversations that shape our future. > [!quote] [00:46:35] > So we can unlearn exclusion and we can learn inclusion with that intentionality. - Contact, as spoken: Michelle A Williams PhD on LinkedIn; company "making accessibility work"; book *Accessible UX Research*, with a QR code on screen (00:46:35 to 00:47:19). #### ⏱️ [00:47:19](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=2839) Host reaction (host) - Host thanks her, recognizes her voice from a talk "four years ago in September of 2021" on how sighted and blind web navigation differs (00:48:02). The caption timeline conflicts with the SEP 2026 title; see Questions. - Points to an August 2022 A11yTalks session on representation (caption: "Key Wells and Dan Edge") and says seeing a disabled family member in a TV show without a backstory is inclusion to him (00:48:46 to 00:49:29). - His own practice: asking "can everybody see this? Can everybody experience this?" (00:50:12). #### ⏱️ [00:50:12](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=3012) Q&A 1 (audience via host): examples of disability questions in research - Question as asked by the host: "what are some examples of your disability questions in your research?" - Answer (00:50:56): screener with category checkboxes (vision, mobility) plus open-ended description; ask about assistive technology; ask about device use without diagnosis. > [!quote] [00:51:40] > someone is increasing their font on their mobile device by 25. - Two veins of research (00:51:40 to 00:53:06): focused accessibility research (foundational, audit-like, or feature-focused such as Slack's screen reader work) versus including disabled people in all research, "the dream." #### ⏱️ [00:53:06](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=3186) Q&A 2: recruiting and fairly compensating disabled participants - Question: "how do we recruit and fairly compensate disabled participants for user research and usability testing?" - Recruitment (00:53:50): firms, normal channels with the screener, disability-focused organizations. - Compensation: she has changed her view. Baseline is fair pay for all. > [!quote] [00:54:34] > there might be additional compensation if say there's transportation involved. - Then a challenge from the author of a book she names as "against technoabbleism," Dr. Ashley Shu (likely Ashley Shew): people giving ideation or deep-dive design input are consultants, not participants (00:54:34 to 00:55:19). Some participants want the credit on a resume, and teams should consider attribution, more pay and a path to hire. - Host response (00:55:19): a different pool, and a consulting opportunity for product teams. #### ⏱️ [00:56:45](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=3405) Q&A 3: intersections with race, age and socioeconomic factors - Question: "how do we ensure our research captures the intersection of disability along with race, age and possibly you know socioeconomic factors?" > [!quote] [00:56:45] > no one research study answers every question or addresses every person. - Answer: build a knowledge base; each study adds what earlier ones lacked (00:56:45 to 00:57:30). > [!quote] [00:57:30] > no one person answers for everyone else. - Her examples of gaps are within disability (deaf and signing users, neurodivergence), not intersections with race, age or income (00:58:12 to 00:58:56). #### ⏱️ [00:58:56](https://www.youtube.com/watch?v=rTlzxDLkg8I&t=3536) Host close, call for speakers (host) - Thanks; continues on LinkedIn; invites new speakers and "marginalized folks in tech"; email and social links as spoken (01:00:22): YouTube, LinkedIn, Bluesky, Mastodon. Ends 01:01:07. ## 📚 References Every item is named in the talk. The talk gives URLs for none, so none are invented here. - [Video: Exclusion by Design, A11yTalks (YouTube)](https://www.youtube.com/watch?v=rTlzxDLkg8I) - Michele A. Williams, *Accessible UX Research* (book; QR code on screen at 00:47:19; no URL spoken). - Stella Young, TED talk "I'm not your inspiration, thank you very much" (00:20:25 to 00:21:08). - John Lewis, "good trouble" (00:24:03). - Dr. Cynthia Bennett, blind UX researcher, quoted in the book (00:36:24). - Dr. Ashley Shew, book on technoableism (00:54:34; caption "Shu"). - Stock photo repositories: All In Frame (00:31:18), Disabled And Here (00:32:01), "Pixels" (00:32:44). - Slack's screen reader design work (00:52:22; no source given). - A11yTalks earlier sessions named by the host: Williams's 2021 talk on sighted and blind web navigation; August 2022 talk on representation (00:48:02 to 00:48:46). - WCAG (00:11:37; no version). ## 🗣️ Paraphrase - **The premise.** Williams says her accessibility education came from disabled colleagues, participants and friends more than from standards. That changed what she saw: a cobblestone road in a concept rendering, a high touchscreen coffee machine, an ad with no disabled kids. She concludes disabled people already knew exclusion was everywhere and she was catching up. - **The mechanism.** The raw material for products (personas, journey maps, stories, product visions, AI-generated concepts) shapes who the team can picture. Colleagues who hadn't pictured a disabled professor couldn't see why instructor-side software needed the same work. She says omission sometimes comes from ableism, not ignorance, and she lists polite-sounding versions of it. - **The test and the fix.** She asks teams to check whether disabled people shaped research direction, strategy and personas, and she asks teams to treat accessibility as an initiative any role can start. Researchers add screener questions, designers annotate files, developers name icon buttons, PMs write acceptance criteria, and anyone can add captions first. - **The pace of change.** She says change comes from repetition: one person asks, another repeats, a third expects. She tells a story in which a social media manager preempted her captions question. Her advice for advocates who feel alone is to keep going. - **The Q&A.** On research practice she favors screeners with a disability category list and an open field, assistive-technology questions that don't require a diagnosis, recruitment through disability organizations, extra pay for transport and time, consultant status for ideation work, and a cumulative research knowledge base. ## 📰 Newsletter **Who are your personas, exactly?** A11yTalks' September session has Michele A. Williams asking WordPress-shaped questions with no WordPress in them: who's in your mockups, who reviewed your strategy, and when did a persona last acquire a disability? Her practical list is short. Name your icon buttons, annotate your design files, add captions before anyone asks, and swap in stock photos that show disability without making it the story. Bring it to your next theme review or Learn lesson, and watch the video here: https://www.youtube.com/watch?v=rTlzxDLkg8I. A11yTalks runs monthly and says sponsor money goes toward paying speakers (00:01:28). ## ✅ Action items The talk doesn't mention WordPress, so every team mapping below is my application of a timestamped recommendation. Each item says what to check first, because I haven't confirmed the current state of any team's process. #### DevRel - Developer advocacy: caption every recorded demo and talk before publishing, and write image descriptions for each slide in the speaker notes. Basis: "add captions before we're asked" (00:29:51) and the social media manager story (00:37:52). Measure: share of published videos with checked captions. - Education: run the five audit questions (00:22:36) on one workshop's personas and examples, then publish the findings internally. Use notice, question, interrupt, normalize (00:43:40) as the frame for an accessibility talk. - Developer experience: no sample ships an unlabeled icon button (00:29:07), and every drag-and-drop sample ships with a documented keyboard path (00:28:23). - Developer marketing: replace stock photos in decks, blog headers and landing pages with disability-inclusive sets (00:30:35 to 00:33:28), and reuse the supplied alt text where the repository provides it (00:32:44). Check each repository's license first; the talk doesn't state them. - Community management: add optional disability, assistive technology and device-adjustment questions to community surveys (00:50:56 to 00:51:40) and report the answers in the results. - Internal and cross-project: add "Did we recruit disabled users? Is the meeting accessible? Is there alt text?" to pre-launch and pre-event checklists (00:40:01). #### OSPO - No signal. The talk covers product-team practice, not policy, licensing, security or inventory. #### Community - Maintainers: don't make one or two disabled contributors the only source of accessibility review; spread the checks across non-disabled maintainers (00:37:07). Add a line for accessibility acceptance criteria to issue and PR templates (00:29:51). - Contributors: pair onboarding with a "does this need a keyboard path or an accessible name" prompt for first contributions (00:28:23 to 00:29:07). - Sponsors: budget paid consulting for lived-experience review as a separate line, not volunteer labor (00:54:34 to 00:55:19). - Ecosystem: keep a shared log of who research and testing have reached so far, reviewed on a schedule (00:56:45 to 00:57:30). #### WordPress - Accessibility: (1) run the 00:22:36 questions against WordPress product personas and Make-blog examples and post the gaps; (2) run or request an accessibility-focused usability study with disabled WordPress users and present it to Core and Design, the "unprompted study" move (00:26:57); (3) tally past WP Accessibility Day sessions by disability population and use deaf and signing users and neurodivergence as program gaps (00:58:12 to 00:58:56). I haven't seen the program, so tally first. - Design: (1) annotate design files with heading levels, landmarks and focus order before handoff (00:26:57); (2) in a design system, make an icon-button component fail or warn without an accessible name (00:29:07); applies to your own system too; (3) add a photography and illustration rule: include disabled people shown neutrally (00:31:18 to 00:33:28); (4) put disabled personas, including one who acquires a disability, in concept renderings (00:22:36, 00:29:51). - Themes: add icon-only controls (menu toggle, search toggle, close) to the review checklist, requiring accessible names (00:29:07); check where bundled demo and screenshot images come from and prefer sets with disability representation and alt text (00:30:35 to 00:32:44). - Training: add a Learn exercise "test your design with a screen reader before handoff" (00:29:07); require at least one disabled persona in course scenarios, including one who acquires disability (00:22:36); caption and transcript every workshop (00:29:51); pay disabled reviewers for lesson review as consultants (00:54:34). - Documentation: audit Handbook and DevHub examples and screenshots for who's pictured (00:12:21, 00:15:16); add an alt-text examples section modeled on the repositories' supplied alt text (00:32:44); add "captions? alt text?" to the contributor checklist (00:40:01). - Community: add to the WordCamp and meetup organizer checklist: recruited disabled attendees and speakers, accessible meeting, captions on (00:40:01); add optional disability and assistive-technology questions to attendee surveys (00:50:56); pay disabled panelists and consultants and credit them (00:54:34 to 00:55:19). - Core: (1) check whether the Gutenberg PR template and Core Trac guidance carry an accessibility acceptance criteria line; add one if not (00:29:51); (2) audit editor interactions for drag-and-drop with a keyboard alternative and icon-only controls with names (00:28:23 to 00:29:07); (3) add a disability and assistive-technology item to usability-test screeners (00:25:29); (4) if Core AI ships image or pattern generation, test whether prompts for "users" or "team" produce disabled people, then log the results (00:23:20). - Marketing and Photos: use inclusive imagery in campaigns (00:30:35); for the Photo Directory, check whether it accepts and tags neutral depictions of disability and recruit submissions (00:31:18 to 00:33:28). - TV: audit caption coverage on uploads and make captions the default before anyone asks (00:29:51). - Other subteams (Support, Mobile, Playground, Polyglots, Meta, CLI, Hosting, Tide, Openverse, Plugins, Core Performance, Core Program): No signal. #### The WPOCC - Programs and funding experiments: define a consultant tier for lived-experience contributors doing ideation or design review, with pay, attribution and a path to hire, and keep participant-level pay separate (00:54:34 to 00:55:19). - Projects: No further signal. ### FAIR - No signal. ## 🧪 Hypotheses - **Hyp 1.** Teams whose personas include one disabled persona who acquires a disability catch more accessibility defects at design review than teams without one (00:11:37, 00:22:36). - **Hyp 2.** Adding a disability category list and an assistive-technology question to a screener raises the share of disabled participants without raising screener abandonment (00:25:29, 00:50:56). - **Hyp 3.** AI image or concept generators asked for ordinary workplace, school or community scenes depict visible disability in a small minority of outputs (00:23:20). - **Hyp 4.** After one person adds captions (or alt text) unprompted, the share of teammates who do so without a reminder rises within a quarter (00:42:54). - **Hyp 5.** Design files annotated for headings and focus order produce fewer handoff accessibility defects than unannotated files (00:26:57). - **Hyp 6.** Treating ideation contributors as paid consultants raises retention of disabled contributors in a research panel compared with flat participant pay (00:54:34). ## 📊 Methodology - **Hyp 1:** compare defect counts at design review across 10 or more matched projects, half with the persona change. Track defect type and WCAG 2.2 criterion. - **Hyp 2:** run an A/B on two screener versions with the same recruitment channel; measure disabled-participant share, completion and drop-off per question. - **Hyp 3:** generate a fixed set of 200 or more images across 20 prompts and several tools; two coders count visible disability markers, with inter-rater agreement reported. State the tool versions and dates. - **Hyp 4:** baseline the practice rate (captions per video, alt text per image) for 4 weeks, introduce the behavior, then track the rate and who initiates it over 12 weeks. - **Hyp 5:** count handoff defects against a WCAG 2.2 AA checklist for annotated versus unannotated files over two releases. - **Hyp 6:** pilot two compensation structures with matched panel cohorts; track repeat participation and satisfaction at 6 and 12 months. - Source limit: the talk offers no data, so every method above tests a claim the talk makes without evidence. ## ❓ Questions - What does "exclusion by design" mean in the speaker's own words? The talk never defines the phrase, and the caption near the title is garbled (00:03:39). - How much of omission is intent and how much is lack of awareness? Her examples (00:14:32) show the second; her claim (00:16:45) is the first. - What's the cost to a junior or contract worker who goes "rogue" (00:25:29)? She notes the job-security risk in one sentence (00:29:07). - How should a team collect disability data in a screener under privacy law, and what's the retention policy? She frames the question as a clarifying one (00:26:14) and doesn't discuss consent or storage. - Which of her stock repositories carry licenses that permit commercial or open-source reuse? - Does the host's "four years ago in September of 2021" (00:48:02) conflict with the SEP 2026 title? 2026 minus 2021 is five years, so either the recording date, the title or the host's arithmetic is off. - What happened to the social media manager's practice after the engagement (00:39:17)? She says she doesn't know. - How do you handle conflicting accommodations when personas multiply? ## 🔮 Future research - Measured audits of who appears in advertising, stock libraries and product imagery, with counts; she observes this (00:08:00 to 00:08:44) without data. - Evidence on whether disabled personas change design outcomes, which the pipeline claim (00:11:37) assumes. - An empirical study of AI concept generation and disability (00:23:20). - Research into deaf and signing users and neurodivergent users as under-studied groups (00:58:12 to 00:58:56). - Methods for intersectional recruitment (disability crossed with race, age, income), which the Q&A answer didn't supply (00:56:45). - Longitudinal tracking of norm change (00:42:54) beyond anecdotes. - Compensation models for lived-experience consultants in open-source and volunteer settings (00:54:34). ## 💡 Implications - **Practice.** Upstream artifacts become audit targets: personas, journey maps, user stories and imagery (00:11:37, 00:22:36). Each role has an action it controls (00:25:29 to 00:29:51). - **Policy.** Accessibility acceptance criteria and caption defaults can enter templates and checklists without waiting for a mandate (00:29:51, 00:40:01). - **Security and privacy.** Collecting disability and assistive-technology data in screeners (00:25:29, 00:50:56) creates sensitive data that the talk doesn't address. Collect it with consent and a retention limit. - **Education.** Lesson design should include learning from disabled people, not only standards (00:11:37). - **Funding.** Pay structure splits: participants versus consultants with attribution (00:54:34). - **Community.** Don't make disabled members the only reviewers (00:37:07), and don't ask "is this accessible" as a gotcha (00:34:11). ## 🧠 Learning styles assessment - **Reading level estimate.** Flesch-Kincaid grade 10.1 and Flesch reading ease 59.3, computed with a vowel-group syllable heuristic on 7,183 words of Williams's own speech (00:03:39 to about 00:47:19), 346 caption-punctuated sentences, 20.8 words per sentence, 1.49 syllables per word. The host and Q&A are excluded. Spoken text with filler (53 "um" or "uh" in the sample) runs longer per sentence than edited prose, so the grade overstates difficulty. #### 👀 Visual learners - Present: slides with renderings, ad and photo snapshots, a coffee machine photo, stock photos and a QR code (references at 00:06:32, 00:09:26, 00:31:18, 00:47:19). I couldn't see any of it; only the transcript was available. - Improve: before-and-after pairs of a persona or rendering with and without disabled people, and a one-page diagram of the notice, question, interrupt, normalize cycle. #### 🔊 Aural learners - Present: conversational delivery and a clear spoken recap at 00:43:40. The host also reads questions aloud. - Improve: a 3-minute audio summary of the four steps and role actions. #### 🗣️ Verbal learners - Present: stories (coffee machine, students versus professors, captions story), a spoken ableism definition (00:16:45), and a named term, "inspiration porn." - Improve: a short glossary (ableism, identity-first, technoableism, accomplice, inspiration porn), since several terms arrive through captions that garble them. #### 🧩 Kinesthetic learners - Present: concrete tasks by role (screener checkboxes, annotate headings, name icon buttons, screen reader test). - Improve: a worked exercise: take a persona set and rewrite it with one persona who acquires a disability; a screener template with sample wording. #### 🧠 Logical learners - Present: a clear chain (representation, imagination, design, product) and five audit questions. - Improve: statistics and sources, which are absent; a decision tree for when to run a focused study versus an inclusive protocol (00:51:40). #### 🧑‍🤝‍🧑 Social learners - Present: audience Q&A, the host's reactions, and her stress on conversation with disabled people (00:11:37) and repetition among colleagues (00:42:54). - Improve: discussion prompts for team reading groups; a peer-review rubric for personas. #### 🧘‍♀️ Solitary learners - Present: reflection cues ("who are we imagining?", 00:22:36) and a call for intentional self-education (00:44:24). - Improve: a self-audit worksheet of the five questions. #### 🌿 Naturalistic learners - Present: built-environment examples (waterfront, roads, trains; 00:06:32 to 00:07:16). - Improve: a link between those physical examples and digital ones; no sustainability angle appears in the talk. ## 📊 Accessibility assessment I assessed this video (auto-captions only; transcript has ">>" speaker-change markers but no names, and I could not see slides or the video description, chapters or caption settings). #### Screen reader semantics - Present: the host gives a visual self-description in the first minute (00:00:00). Williams gives no self-description in the captured audio. She does describe most slide images aloud: cobblestone road and cyclist (00:06:32), the coffee machine (00:09:26), each stock photo (00:31:18, 00:32:01, 00:33:28), the exoskeleton videos (00:20:25), and she reads out her contact details (00:46:35). - Improve: descriptions are partial and unlabeled (the Old Navy slide at 00:08:00 is described as "a lot of children... diverse"), and the QR code at 00:47:19 is on screen only; she asks the team to post a link. Maps to WCAG 2.2 1.1.1 (Non-text content) for the QR and 1.2.5 (Audio description, prerecorded, AA). #### Keyboard navigation - Present: the YouTube player supports keyboard operation; nothing in the audio indicates in-video controls. - Improve: no gap I can identify from the transcript. WCAG 2.1.1 and 2.4.7 pass via the platform; chapter markers (a navigation aid for a 1:01:15 video) are unverified. #### Contrast and readability - Present: unverifiable from a transcript. - Improve: she says a later slide "has a lot of stuff on it" (00:46:35), a density signal; check 1.4.3 (contrast) and 1.4.12 (text spacing) on the deck if published. #### Text resizing and reflow - Present: captions resize in the player; slide text burned into video doesn't reflow. - Improve: publish the deck as accessible HTML or tagged PDF (1.4.4, 1.4.10) if it's shared. #### Media and captions - Present: auto-generated captions exist. - Improve: they garble proper nouns the audience needs: the host's group name ("Ali Talks"), the speaker's name ("Michelle"), "accessessibility is narrow" (a quote in a talk about accessibility), "abbleism," "technoabbleism," and the repository names (00:31:18 to 00:33:28). Fifteen garbles are listed in Main content. That fails 1.2.2 (Captions, prerecorded) in substance; for the live portion, 1.2.4 (Captions, live) applies. No speaker labels: Q&A turns are marked ">>" only, and the host's questions and Williams's answers are separable by content, not by name. Human-corrected captions and a downloadable transcript would close this. #### Forms and interactions - Present: three question routes (live chat, YouTube comments, LinkedIn and other social, 00:02:13), so a viewer who isn't live still has a path. Questions are read aloud. - Improve: the contact email is garbled in captions ("teamallettalks.com", 01:00:22), so it can't be reliably copied from audio alone. A visible, text link would help (3.3.2 Labels or Instructions is the nearest equivalent; no form is in the video). #### Standards reference - Gaps against WCAG 2.2 AA from what I could assess: **1.2.2** (caption accuracy), **1.2.4** (live captions, unverified), **1.2.5** (audio description; most slides are narrated but not all), **1.1.1** (QR code). Unverifiable: **1.4.3**, **1.4.10**, chapters. - The talk's own subject sharpens the finding: the host modeled a visual self-description; the speaker narrated her images but didn't self-describe. ## 🌍 DEIB assessment #### Cultural breadth - Present: a Netherlands rendering (00:06:32), US civil-rights history (00:24:46), sign language and native signing (00:32:01, 00:58:12), queer and multi-cultural representation in one repository (00:32:44). - Improve: examples are US-centered (Boston, Denver and California offices; Old Navy; Audible; Slack; John Lewis). No non-US disability law, language or context is discussed. #### Inclusive language - Present: she states identity-first language at 00:03:39 and explains why (00:04:22); defines ableism (00:16:45); credits Stella Young for "inspiration porn" (00:20:25); and uses "disabled people" consistently. - Improve: she says "sneaky person" and "rogue" for assertive behavior, which is playful but not defined; "accomplice" is explained (00:24:03), but its legal-risk connotation is not. #### Neurodiversity support - Present: a clear four-step recap at 00:43:40 and an openly modeled slip ("I don't remember the fourth," 00:37:07). Dyslexia appears in the Audible example (00:19:41). - Improve: examples are mostly mobility, vision and hearing. Neurodivergence enters at the end as a recent learning (00:58:56). The talk has no cognitive-accessibility examples. #### Representation - Present: she names her own position ("a non-disabled person," 00:35:39) and sources perspectives from a blind researcher (Dr. Cynthia Bennett) and a disabled advocate (Stella Young). The host describes himself (white man, red hair, glasses). - Improve: the speaker doesn't state her own identity beyond that, and I can't tell from the transcript whether any presenter at this session is disabled. She also says disabled people shouldn't have to be in every room (00:37:07), but this session's speaker is non-disabled, as she says. #### Psychological safety - Present: the code of conduct is read at 00:00:00; she grants colleagues "a little bit of grace" (00:16:00) and warns against gotchas (00:34:11); the Q&A is cordial. She models changing her mind (00:53:50). - Improve: no content warning, though she recounts ableist remarks. The "rogue" advice at 00:25:29 carries risk for people with less status; she acknowledges it once (00:29:07). ## ✅ Logical validity #### Strong reasoning - **Checkable decomposition (00:09:26 to 00:10:10).** The coffee-machine claim breaks into three barriers (touchscreen interface, height of the machine and controls, position on the counter), each tied to a user group. A reader can test each against the machine. - **A falsifiable audit (00:22:36).** The five questions are answerable per team (did foundational research include disabled participants before direction was set; when did a persona last acquire a disability). - **Role-matched actions (00:25:29 to 00:29:51).** Each action sits inside an artifact that role controls: screener, design file, code, acceptance criteria. That's an inductive design for adoption, and it's why the advice is more actionable than a general call to care. - **Double-standard comparison (00:18:56 to 00:19:41).** The same text-to-speech technology is praised as efficient for one user and treated as cheating for another. The structure of the comparison is valid, although the second half is a constructed scenario. - **Revision on the record (00:53:50 to 00:55:19).** She revises her compensation view after a named author's challenge, and she says so. - **Honest limits in Q&A (00:56:45 to 00:57:30).** "No one research study answers every question" bounds her own claims. - **Clarity fix.** State the title's meaning in one sentence, and attach one number and one source to the pipeline claim. #### Fallacies - **Equivocation between intent and unawareness (00:16:45 versus 00:14:32).** She calls "didn't know" an excuse and says disability is "sometimes... intentionally left out," but her own evidence shows colleagues who hadn't thought of it. Weakness: the example doesn't support the stronger claim. Fix: separate three mechanisms (unaware, unimagined, deprioritized) and give one example of each. - **Unsupported causal claim (00:15:16, 00:46:35).** "We can't design for who we haven't imagined" and "exclusion is learned" are asserted without a study. The first is close to a tautology (a team that never considered a user hasn't designed for them). Fix: cite research on persona or team composition and defect rates, or mark it as practitioner experience. - **Hasty generalization from anecdote (00:08:00 to 00:15:16, 00:21:08).** A coffee machine, an office conversation, ads and stock photos support "exclusion is the norm." Her hedges ("at least not often," "not necessarily") limit this. Fix: count a sample of ads or stock results. - **Analogy with a scale mismatch (00:24:03 to 00:24:46).** Equating adding captions with civil-rights-era law-breaking overstates the stakes and the legal status. She softens it ("subtle or rogue," 00:25:29; "not... get you in trouble," 00:29:07). Fix: use "initiative without permission," her own phrase. - **Survivorship and unverifiable prediction (00:39:17 to 00:42:54).** Two anecdotes (one captions moment, one post-layoff note) support "culture changes quietly," and the manager "not going to forget that" is a prediction she can't check (she doesn't know if they're still at the company). In the layoffs story, the colleague who bypassed her still did what they wanted. Fix: track adoption metrics over time or report cases where practice reverted after the advocate left. - **Unsupported universal (00:34:11).** "Every time we quietly demonstrate that inclusion is possible, we weaken the arguments against it." Fix: "can weaken." - **Partial answer to a question (00:56:45 to 00:58:56).** Asked about intersections of disability with race, age and socioeconomic factors, she answers with cumulative sampling and gives examples within disability. The intersection itself is not addressed. Fix: name a recruitment method (quota by cross-cut, or targeted waves) and give one worked example. - **Unattributed source (00:36:24).** The Bennett quote as captured is unintelligible, and her "takeaway" ("not speaking for people. It's helping people be heard") can't be tied to it. Check the book. - **Not fallacies:** "I believe this is in the Netherlands" (00:06:32), "I don't remember the fourth" (00:37:07) and "maybe" (00:20:25) are hedged speech. #### Statistics and sourcing audit - No statistic has a named source. The prevalence claim (00:26:14) and the student-instructor ratio (00:13:49) have no number. The 25 in the font-size example (00:51:40) is an illustration with no unit. - Credit: Stella Young, John Lewis, Dr. Cynthia Bennett and Dr. Ashley Shew are named, and the stock repositories are named and checkable. #### Tone of voice - Reflective and anecdotal, with humor ("I know this is recorded and can be traced back to me," 00:25:29; "not a spoiled one," 00:02:56). Invitational (stay the course, 00:43:40) more than accusatory. Direct about ableism (00:16:45) while granting grace to colleagues. Conversational, with 53 "um" or "uh" in the main talk. ## 🛡️ DARVO analysis Not applicable to this source. There's no accusation, accused party or defensive response in the talk. ## 🎯 Conclusion - **Bottom line.** The talk argues that disability is erased upstream, in the imagery and artifacts teams build from, and that anyone can interrupt that by acting without waiting for permission. The strongest parts are the five audit questions (00:22:36), the role-specific actions (00:25:29 to 00:29:51) and the research guidance in Q&A (00:50:56 to 00:58:56). The weakest are the missing sources and the undefined title. - **Next steps for readers.** - Run the five questions on your next set of personas or mockups. - Add a disability and assistive-technology item to one screener, with consent and a retention rule. - Name every icon-only button, and test one flow with a screen reader before handoff. - Caption first, then ask. - Check the stock repositories' license terms and replace one set of images. - Verify the garbled names and the 2021 date before citing the talk (Fleeting Queue). ## 🔄 Refactor Appendix — Michele A. Williams / Exclusion by Design > Density: 2–5 Literature atoms (H3 pad), 0–3 Permanent candidates (H4 pad). > Emitted: 5 Literature, 3 Permanent. Resonance filter applied: the stock-photo list, > the Bennett quote and the Q&A specifics stayed in Notes because the captions garble > the first two and the third is procedure, not claim. > > Konik claim-header rules: present tense, declarative, ≤10 words, searchable, unique. > > Atomicity: one declarative sentence per atom. "and"/"also" → split. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Split Pad #### [H1] Accessibility is learned from disabled people, not standards > [!quote] [00:11:37] > I realized I wasn't really learning accessibility from learning standards like WCAG, I was learning accessibility from engaging with disabled people. **Context and use**: - 00:11:37, after the coffee-machine story; her account of how her own education happened, as a researcher and accessibility specialist. - Frames any training program: standards give criteria, and contact with disabled users gives the judgment about where they apply. No WCAG version is given. #### [H2] Teams can't design for users they haven't pictured > [!quote] [00:15:16] > Representation leads to what we imagine, who we imagine, the scenarios we envision that then lead to the design **Context and use**: - 00:15:16, after the students-versus-professors story. The pipeline claim: representation, imagination, scenarios, design, products. - Treated as practitioner experience in this note; no study backs it in the talk. #### [H3] Solutions rarely include what the team treats as a problem > [!quote] [00:23:20] > Basically, if disability is seen as a problem, it will rarely be included in the solution. **Context and use**: - 00:23:20, after the AI question; the closing of her missed-opportunities argument and the setup for "disability is not the problem, it's our viewpoint." - The causal direction is asserted, not shown, and it has no source. #### [H4] Inclusion rarely happens by accident > [!quote] [00:24:03] > In order to interrupt though, we need to take some action because inclusion rarely happens by accident. **Context and use**: - 00:24:03, opening of the "interrupt" step and the "accomplice" framing. - Justifies unrequested actions (captions first, screener questions, annotated files); she bounds it with a job-security caution at 00:29:07. #### [H5] Advocates change a room others are listening to > [!quote] [00:41:29] > sometimes you won't convince the person you're talking to, but everyone else is listening. **Context and use**: - 00:41:29, before the layoffs story and the three-person repetition chain at 00:42:54. - Practical for volunteer and open-source settings where the visible objector is not the whole audience; evidence is two anecdotes. ### 🧠 Atomic Split Pad #### [P1] A missing persona is a missing requirement **Claim** (one sentence, Courtney's voice): A product team can't review for a user nobody pictured, so a persona set without disabled people is a requirement set without them. **Evidence pointer**: [[#H2]] · [[#H3]] **Scope & conditions**: Holds where personas, user stories or concept art drive design decisions. Weakens in a team that tests with disabled users directly and has no personas. The talk offers anecdotes, not data. **Links**: - [[Design system documentation as audit surface]] — because extends - [[Persona spectrum]] — because contrasts on whether the persona is a person or a permanent trait - [[WordPress accessibility acceptance criteria]] — because applies-in Core and Gutenberg process  #### [P2] Accessibility norms spread through repetition by a second person **Claim** (one sentence, Courtney's voice): A practice becomes standard when a second person repeats it unprompted and a third expects it, not when the original advocate persists. **Evidence pointer**: [[#H5]] **Scope & conditions**: Observed in two anecdotes (captions, a post-layoff note); no adoption data. Fails if the practice reverts when the advocate leaves. **Links**: - [[Contributor onboarding]] — because applies-in first-contribution checklists - [[Maintainer advocacy fatigue]] — because extends - [[Default settings as policy]] — because contrasts on whether norm change needs a default  #### [P3] Pay lived-experience input by its role **Claim** (one sentence, Courtney's voice): Participant pay fits task feedback, and consultant pay with attribution fits ideation and design input, so the role decides the rate. **Evidence pointer**: [[#H4]] — compensation segment at 00:54:34 to 00:55:19 stays in Notes because the captions garble the source's name. **Scope & conditions**: Applies to research panels and community programs with a budget. Needs a clear line between feedback and co-design, which she leaves to judgment. **Links**: - [[Open source sustainability funding]] — because applies-in paid contributor tiers - [[Research participant compensation]] — because extends - [[Unpaid accessibility labor]] — because contrasts on who carries the cost  ### ✏️ Fleeting Queue - [F1] Verify speaker name: title says "Michele A. Williams", captions say "Michelle", and she says one L at 00:46:35 — check the A11yTalks speaker page #fleeting #inbox - [F2] Verify All In Frame, its producer (caption "Levant consulting group"), and license terms (00:31:18) #fleeting #inbox - [F3] Verify "Disabled And Here" (spelled out in captions, 00:32:01) and "pixels" (possibly Pexels, 00:32:44) and each license #fleeting #inbox - [F4] Find the exact Cynthia Bennett quote in *Accessible UX Research*; the caption at 00:36:24 is unintelligible #fleeting #inbox - [F5] Verify "Dr. Ashley Shu" (likely Ashley Shew) and the book title (technoableism), 00:54:34 #fleeting #inbox - [F6] Date check: host says "four years ago in September of 2021" (00:48:02) against the SEP 2026 title; confirm the recording date #fleeting #inbox - [F7] Verify the August 2022 A11yTalks session on representation and its speakers (caption "Key Wells and Dan Edge", 00:48:46) #fleeting #inbox - [F8] Find the source for Slack's screen reader design work "in the news" (00:52:22) #fleeting #inbox - [F9] Look up disability prevalence figures from a primary source (CDC or WHO); the talk gives none (00:26:14) #fleeting #inbox - [F10] Confirm the company website and A11yTalks contact addresses (captions "mawconultinglc.com", "teamallettalks.com") #fleeting #inbox - [F11] Article seed: "Who's in your WordPress persona set?": run the 00:22:36 questions on Make-blog personas #fleeting #inbox #articleseed - [F12] Check whether the Gutenberg PR template has an accessibility acceptance criteria line #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Run the five audit questions (00:22:36) on my design system's personas and example imagery, and list gaps — design system review, 30 min #things - [ ] Verify the repositories (All In Frame, Disabled And Here, Pexels) and license terms, then swap one demo image set with supplied alt text — content review, 20 min #things  📄 **Full text:** [[Full Document Contents/Articles/Exclusion by Design - Michele A. Williams.md|Exclusion by Design - Michele A. Williams by Michele A. Williams]] ## Highlights - [00:11:37] So basically what was happening was in the course of learning accessibility, learning digital accessibility, I realized I wasn't really learning accessibility from learning standards like WCAG, I was learning accessibility from engaging with disabled people. whether they were my colleagues, whether they were research participants, whether they were friends, I was starting to get that perspective. Well, what does this all mean for those of us who primarily work in the tech field, which I presume would be most of the audience for these talks? Well, what it means is that there might be representation or or perspective missing in some of the key aspects of what we do to build our tech products. For instance, in our personas, in our [00:12:21] journey maps, in AI generated content or concepts, in product visions, in user stories, these are the foundations that lead to what we build. But that is going if if there are aspects and perspectives missing, that is going to impact what we build and how we go about building it. Here's an example. that same position that where I said I traveled to the different offices. I had another encounter in my home office in Boston and I've dubbed this students versus professors. So essentially we were in the kitchen of that space. So apparently these kitchen spaces are important but ([View Highlight](https://read.readwise.io/read/01m48z20t18zr0bc5fsz7vench)) - Note: [H1] Accessibility is learned from disabled people, not standards - Tags: atom/literature/h1 - [00:14:32] checks. that's how they were viewing it. But in the process of doing that and again hearing the sort of almost relief that they didn't have to do that work because they worked on the instructor side. I interjected and I pointed out that professors can be disabled too. This then led to them basically falling silent. It wasn't that they disagreed. It's just that it simply had not occurred to them that professors or instructors can also have disabilities in the vein that uh and I continue to explain students who have disabilities [clears throat] for instance graduate and they may want to get a job somewhere and that job may look like becoming an [00:15:16] instructor or people also acquire disabilities throughout their throughout their lives. So either way, you all should also be thinking about this and are going to be asked to do the same task. This is not a separation. So that leads back to this idea of a phrase that you may have often heard. Representation matters. Representation leads to what we imagine, who we imagine, the scenarios we envision that then lead to the design and lead to the products that we put out into the world. So we can't design for who we haven't imagined. So this is why it is important to understand that not everyone notices the world in the same way if you've not ([View Highlight](https://read.readwise.io/read/01m48z23c9kmcdnn1y0nvda87c)) - Note: [H2] Teams can't design for users they haven't pictured - Tags: atom/literature/h2 - [00:23:20] but nonetheless, we know that companies are doing it. Then if AI generates concepts based on what already exists, for instance, based on training models, then what happens when exclusion is already the norm? Basically, if disability is seen as a problem, it will rarely be included in the solution. So hopefully at this point, maybe you've experienced some of this, maybe this is resonating with you, maybe you had not yet noticed it, but now you do. Hopefully within that space, you ultimately see this as not the right way to go, not the right way to think about disability, not the right way to have it showing up in our world. Basically, you're seeing that disability [00:24:03] is not the problem, it's our viewpoint. So therefore, what we want to do next is interrupt. In order to interrupt though, we need to take some action because inclusion rarely happens by accident. So really many accessibility improvements don't necessarily require permission. They require initiative. So how I position that is we should become accompllices. And I use this word uh intentionally because an accomplice in this instance would be someone who actively helps disrupt an unjust status quo. So I think about uh in the US there is a late civil rights leader named John Lewis who came up with the phrasing good ([View Highlight](https://read.readwise.io/read/01m48z25mh51sbzwm7k7mkqkh2)) - Note: [H3] Solutions rarely include what the team treats as a problem - Tags: atom/literature/h3 - [00:40:44] we're putting together. Because we also know that accessibility again is not just the products, it's the world. So thinking about again how I've been viewing the world, uh everything from the coffee machine to transportation to representations on TV. This isn't, excuse me, this isn't just um our websites or our digital products. It's everything. But that leads to the push back that you often get of someone who it seems like you're really having a tough time to convince to do this. You're getting that push back that it takes time and we don't have the time to do this. We don't need to do this. And there are going to be occasions when you don't actually sort of win the argument, let's say. But the [00:41:29] way that I position that is sometimes you won't convince the person you're talking to, but everyone else is listening. And I have a very much have a a story around that as well. I worked again at a company. I was in charge of doing this work. I had some folks that did have that push back and we would have back and forth because I am who I am and I'm going to advocate how I advocate. There were often some back and forth in meetings around the prioritization of the considerations that needed to be included. I often did not convince that person. They would end around me. They would do what they, you know, felt they wanted to do. But not long after a lot of layoffs happened and a lot of the dust settled after that, I ([View Highlight](https://read.readwise.io/read/01m48z282dwv42bect3gm9c4yj)) - Note: [H5] Advocates change a room others are listening to - Tags: atom/literature/h3