humanitarian · digital · family: designed for the wrong person
digitized aid that disabled refugees can't reach
Refugees with disabilities cannot access the digital services that humanitarian organizations built to reach them
Problem statement
Humanitarian organizations have increasingly digitized their services — registration, status updates, appointment scheduling, information dissemination, complaint mechanisms — and delivered them through websites, mobile apps, and messaging platforms. This digital shift was accelerated by COVID-19 and is now standard practice. But refugees with disabilities, particularly those with visual impairments, cognitive disabilities, or limited motor function, cannot access these digital channels. The UNHCR Innovation Service names this as the challenge behind one of its 2024 projects, a refugee-led observatory in Poland: "Refugees with disabilities, particularly those with visual impairments, struggle to access critical information via online channels extensively used by humanitarian and governmental actors, undermining their ability to access social and protection services." The digital accessibility problem in refugee contexts can be more severe than in general populations where displaced people have fewer alternative channels — if the digital service is inaccessible and there is no in-person alternative, the person is simply excluded.
Why this matters
Refugees with disabilities are among the most marginalized within already-marginalized populations — they face compounding barriers of displacement, disability, and often poverty, gender, or age-related discrimination. When critical services go digital without accessibility, these individuals lose access to protection, healthcare, legal aid, and livelihoods information. This is not a niche problem: UNHCR expects that its project in Poland alone could benefit 38,000 Ukrainian refugees with visual impairments. The digital accessibility gap creates a two-tier system where the most vulnerable refugees have the least access to the services designed to help them.
What’s been tried and why it hasn’t worked
Humanitarian digital platforms were built under extreme time and resource pressure, prioritizing reach and speed over accessibility. The sources cited here do not say whether UNHCR and partner organization websites and apps meet WCAG 2.1 Level AA accessibility standards, or how well screen readers work in the languages used by displaced populations. UNHCR's WhatsApp-based messaging service, piloted since 2020, reached more than 280,000 users across 16 countries and is now being scaled. In Ecuador, UNHCR added audio versions of all content for users who cannot read. The article describing the service does not mention users with visual impairments or other disabilities. Information sessions held in camps are primarily designed for hearing and sighted attendees. Physical information boards and printed materials assume literacy and visual access. Assistive technology that exists in high-income contexts (screen readers, voice interfaces, adaptive controllers) is designed for widely spoken languages, assumes stable internet connectivity, and costs far more than displaced families can afford. Attempts to add accessibility features to existing platforms after launch have been superficial — retroactive fixes rather than accessible-by-design approaches. The sources do not say whether refugees with disabilities were consulted when these services were designed. The Poland project sets up a refugee-led observatory to advise humanitarian and other actors on how to improve online community engagement, based on how accessible information is as experienced by refugees.
What would unlock progress
Accessibility-by-design principles embedded from the start of humanitarian digital service development, combined with assistive technology adapted for displacement contexts. Specific needs include: (1) voice-first interfaces in languages spoken by displaced populations (Arabic, Dari, Tigrinya, Somali, Rohingya, Ukrainian, etc.) that don't depend on reading or visual navigation; (2) AI-powered processing of text and audio submissions to UNHCR, as identified in the 2024 innovation portfolio — enabling people to communicate in their own language and modality rather than adapting to the system's requirements; (3) community-based accessibility support where trained refugee volunteers assist people with disabilities in navigating digital services; (4) offline-capable tools that don't require stable connectivity; and (5) participatory design processes that include refugees with disabilities as co-designers, not just end-user testers. The UNHCR Refugee-led Innovation Fund's model of funding displacement-affected communities to design their own solutions is structurally aligned with this approach. The Fund's page says it gives additional attention to organizations led by people with disabilities, and it lists projects for refugees with disabilities, such as art therapy for children in Yemen and a centre in South Africa that trains people with disabilities to repair phones, wheelchairs, and other assistive technology. None of the projects listed there is about making digital services accessible to people with disabilities.
Entry points for student teams
A student team could conduct an accessibility audit of a specific humanitarian digital service (UNHCR registration portal, a refugee-serving NGO's mobile app, or a WhatsApp-based information channel) against WCAG 2.1 Level AA, following the W3C's published evaluation methodology at https://www.w3.org/WAI/test-evaluate/, then design and prototype accessibility improvements for a specific disability type and language community — auditing a public-facing service needs no partner, license, or restricted data. Recruiting displaced people with disabilities as test participants is out of reach for a semester team, since it is a doubly protected population reachable only through agency-mediated recruitment and ethics review, so the deliverable is the audit findings, the redesigned prototype, and a usability evaluation run with disabled users reached through a domestic resettlement or disability-services organization, with the displacement-specific critique supplied by practitioners who work in those settings rather than by their clients. A team with NLP or voice interface skills could prototype a multilingual voice-first interface for a specific humanitarian information service on top of open speech models that already reach displacement languages: Meta's MMS checkpoint (https://huggingface.co/facebook/mms-1b-all) does ASR for over 1,000 languages including Tigrinya, Somali, and Amharic, and NLLB-200 (https://huggingface.co/facebook/nllb-200-distilled-600M) covers the translation leg. The languages those models still miss — Rohingya among them — define the harder half of the problem and are worth documenting as a finding in their own right. Skills in accessibility design, UX research, NLP, or human-computer interaction would be most relevant.
Genes
Structural cousins: stuck the same way, in other fields
Sources
UNHCR Innovation Service, "Our 2024 Projects," (accessed 2026-02-14). UNHCR, "Refugee-led Innovation Fund," Supplemented with: UNHCR Innovation Service, "Taking innovation global with two-way communications with refugees," Medium, UNHCR, "Mental Health and Psychosocial Support," go to source 1 ↗ go to source 2 ↗ go to source 3 ↗ go to source 4 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- The `failure:wrong-stakeholder` tag applies because humanitarian digital services were designed for an assumed "default" user — an able-bodied, literate, sighted person with a smartphone and connectivity. Refugees with disabilities, who have distinct and foreseeable needs, were not included as stakeholders in the design process.
- The `failure:not-attempted` tag applies because humanitarian organizations acknowledge the problem but have not made systematic efforts to address it. Individual accessibility improvements are ad hoc rather than part of a coordinated strategy.
- The `constraint:equity` tag is particularly strong here: disability intersects with displacement, poverty, gender, and age to create compounding marginalization. Inaccessible digital services don't just fail to help — they actively exclude the most vulnerable from systems designed to protect them.
- Cross-domain connection: this brief shares deep structure with `health-assistive-tech-aging-adoption-gap` — both involve digital services designed for able-bodied/tech-literate users that systematically exclude the populations with the greatest need. The failure mode (design assumptions that exclude the target population) is identical; the domain context (humanitarian vs. health) is different.
- The UNHCR's 2024 innovation project on AI-supported processing of text and audio submissions is a partial solution — it adapts the system to accept diverse input modalities rather than requiring users to adapt to the system.
- The one figure in the cited sources for the scale of this problem is on the 2024 projects page: 38,000 Ukrainian refugees with visual impairments in Poland could benefit from the observatory project.
Reconciliation 2026-08-21: The entry-point triage flagged this brief as needing usability testing with displaced people with disabilities. That flag is partly right and partly wrong: the brief already offered two distinct doors, and the WCAG-audit door was already facility-free, so the ≥2-door standard was met and no door needed adding. What did fail was the design-the-trial default — the deliverable required test results from a doubly protected population (disability plus displacement) reachable only through agency-mediated recruitment and ethics review that outruns a semester. The audit and prototype arms are kept verbatim in substance; the testing clause now routes usability work through disabled users reached via a domestic resettlement or disability-services organization, with displacement-context critique from practitioners rather than clients. The audit door is now anchored to a verified public method — W3C WAI evaluation resources, https://www.w3.org/WAI/test-evaluate/ (fetched 2026-08-21), alongside the WCAG 2.1 quick reference at https://www.w3.org/WAI/WCAG21/quickref/. The voice door previously named target languages without naming any tool that reaches them; it now cites two models verified by fetch on 2026-08-21: Meta MMS (https://huggingface.co/facebook/mms-1b-all), whose model card confirms 1,000+ ASR languages and whose language list includes `ti` (Tigrinya), `so` (Somali), and `am` (Amharic), and NLLB-200 distilled 600M (https://huggingface.co/facebook/nllb-200-distilled-600M). The same fetch confirmed Rohingya is absent from the MMS list, so the brief now states that gap rather than implying a tool exists for it. Mozilla Common Voice was considered and not cited: the site renders client-side only, so its per-language coverage could not be verified by fetch.
Reconciliation 2026-09-29: The whole write-up was read against all four cited sources. The live addresses at unhcr.org and medium.com refused automated requests, so each page was read in full on 2026-09-29 from the Internet Archive's saved copy of the same address: the 2024 projects page as saved on 2026-02-14 (https://web.archive.org/web/20260214133210/https://www.unhcr.org/innovation/2024-innovation-projects/), the Refugee-led Innovation Fund page as saved on 2026-02-09, the Medium article as saved on 2025-06-21, and the mental health page as saved on 2025-08-11. (1) The text said the UNHCR Innovation Service "identifies this as a 2024 priority challenge." The quoted sentence is the challenge statement of one project on the 2024 projects page, a refugee-led observatory in Poland funded through the Digital Innovation Fund. The page does not call it a priority. The sentence now says where the quotation comes from. (2) The text said the problem "is more severe than in general populations because displaced people have fewer alternative channels." None of the sources makes that comparison, so the sentence now says "can be more severe" where alternatives are fewer. (3) The text said an estimated 15% of the world's population lives with a disability, that displacement creates further disabilities, and that there are 18+ million displaced people with disabilities. The Source Notes derived this from 15% of 123.2 million displaced people, giving 18.5 million. None of these figures appears in any of the four sources, and no source was named for them. They were taken out. The one figure the sources give for the scale of this problem is now used in their place: 38,000 Ukrainian refugees with visual impairments in Poland could benefit from the observatory project. (4) The text said most UNHCR and partner websites and apps do not meet WCAG 2.1 Level AA and that screen reader support is poor or untested. The sources do not mention WCAG, screen readers, or any accessibility testing. The sentence now says the sources are silent on this. (5) The text said UNHCR's WhatsApp service had been "scaled globally as a primary communication channel" and that it excludes people with visual impairments and people who cannot read. The Medium article says the service was piloted from 2020, reached more than 280,000 users across 16 countries, is being scaled, and was built to supplement existing channels such as the hotline and the HELP site. It also says that in Ecuador audio versions of all content were added for users who cannot read. It does not mention users with disabilities. The text now reports those points. (6) The text said "the core problem is that refugees with disabilities were not consulted during service design." The Medium article says community members were consulted in Algeria and that feedback was sought from communities during the pilot, but no source says whether refugees with disabilities were consulted. The text now says that, and describes what the Poland project does. (7) The text said the Refugee-led Innovation Fund "has not yet specifically targeted disability accessibility." The Fund's page says it gives additional attention to organizations led by people with disabilities, and it lists projects for refugees with disabilities in Yemen and South Africa. No listed project is about accessible digital services. The text now says that. Confirmed against the sources and left as written: the titles on the Source line; the quoted challenge statement, word for word; the 2024 project on AI-supported processing of text and audio submissions (a project in Jordan, which the page presents as a way to answer requests faster); the Fund's model of funding organizations led by displaced people; the link between COVID-19 and the move to digital channels. NOT found in any source and left as written because they are general statements without a number or a named organization: that platforms were built under time pressure with reach put ahead of accessibility; that camp information sessions, notice boards and printed materials assume hearing, sight and literacy; that assistive technology is built for widely spoken languages and costs more than displaced families can afford; that accessibility fixes after launch have been superficial; and the list of languages under voice-first interfaces. The Source Notes line saying humanitarian organizations have made no systematic effort is the reasoning for a tag and was left as written, though the Poland project is one such effort. The Entry Points section was checked on 2026-08-21 and was not checked again. Genes untouched.