Advokit

Co-design workshop session with stroke survivors with aphasia, seated around a table reviewing materials

Visit Advokit Code Paper

People with aphasia are routinely excluded from the digital welfare systems they depend on. Benefits portals, housing applications, GP apps — these services assume literacy, digital fluency and communicative competence that many stroke survivors simply do not have. The consequences are severe: lost entitlements, failed appeals, financial hardship.

Screenshots of five inaccessible welfare and healthcare apps including Southern Housing, Swiftqueue and Accurx — Private welfare and health providers — inaccessible by design to people with communication disabilities

P4, a co-designer living with aphasia, called the DWP to migrate his Universal Credit claim. Automated robocalls repeatedly misinterpreted his speech. After extended hold, a dropped call and being told to restart the next day, he attended the Job Centre in person — undermining the premise of digital access entirely.

Detailed journey map of P4's phone call to DWP, showing each barrier encountered from poor mobile coverage to dropped call to in-person attendance required — P4's Universal Credit phone call — a journey map of compounding failures for someone with aphasia

Advokit is a response to this — a generative AI-enhanced digital toolkit co-designed with stroke survivors, speech and language therapists and welfare advisors at Aphasia Re-Connect. Documented in peer-reviewed papers at CHI 2026 and ASSETS 2026.

Community Web Toolkit

Advokit is live at advokit.uk. I designed and built the toolkit end-to-end — full-stack architecture, accessibility infrastructure, LLM-powered assistance and a community stories pipeline. Every feature traces back to barriers named by co-designers: traffic-light difficulty ratings, progressive disclosure of tips, read-aloud with word highlighting and structured feedback controls in Benefit Buddy.

Toolkit Overview

Five core sections: a homepage, benefits list, comparative table, stories page and Benefit Buddy. The homepage opens with an overview video followed by four illustrated tiles for common pathways: considering whether to apply, continuing an application, exploring community stories and learning about the project.

Homepage — four illustrated pathway tiles with embedded explainer video

Benefits List and Table

21 welfare benefits as a navigable list and sortable comparative table — chosen over a linear document after co-designers described feeling overwhelmed. Difficulty uses a traffic-light metaphor (easy · medium · hard); perceived value a five-star rating, enabling rapid low-demand comparison.

Comparative benefits table — sortable by difficulty and perceived value

Benefit Pages

Benefit pages minimise textual density, with key information upfront and additional detail in interactive drop-downs. All text includes read-aloud with highlighting. Community-generated tips use progressive disclosure — revealing incrementally or all at once — and are printable for offline access.

Benefit page — progressive disclosure tips with read-aloud and print support

Benefit Stories

YouTube-embedded interviews with community welfare veterans, paired with commissioned illustrations, captions and printable transcripts. Each story includes a synopsis, top application tips and personal reflections — combining video, illustration and text to support varied literacy levels.

Benefit Buddy

An LLM-powered assistant configured specifically for welfare application writing — breaking questions into steps and coaching applicants to articulate functional impacts in assessor-facing language. Structured onboarding establishes context before the conversation begins. All responses are read-aloud enabled and copy directly into application forms. Feedback buttons — Good response, Bad response, More detail, Less detail — allow refinement without additional typing.

Benefit Buddy — structured onboarding, first-person draft answers and feedback-driven refinement

Co-Design Process

Four stages, May–July 2025: 8 workshops, 7 interviews, over 60 participants across SLT centres, community venues and online.

Four-stage study timeline from board formation and interviews through three rounds of co-design workshops, showing participant counts and design materials at each stage — Study timeline: four stages, eight workshops, seven interviews — 23–31 co-designers with aphasia per stage

Wider view of a co-design workshop room with participants at multiple tables — Co-design workshops across SLT centres and community venues

Stage 2 used accessible card decks — Help, Where, Barrier and Feelings cards — to rank welfare service experiences. Value/difficulty grids surfaced what mattered most: being understood, getting help quickly, not repeating yourself.

Help card showing PIP (Personal Independence Payment), Feelings card showing 'very negative', workshop photos and a value/difficulty ranking exercise with sticky notes — Stage 2: accessible card decks and ranking exercises to surface barriers and priorities across welfare services

Stage 3 used journey mapping to trace specific moments of failure. Co-designers placed cards along a squiggly-line emotional arc, producing maps dense with compound failures: inaccessible forms, phone queues, confusing letters, staff unfamiliar with aphasia.

Journey mapping toolkit with squiggly line format, Where card showing laptop computer, Barrier card saying 'I have low confidence', workshop photos and a completed journey map covered in sticky notes — Stage 3: journey mapping welfare experiences — from the first touchpoint to the final outcome

Stage 4 used speculative design fiction — short video prompts depicting near-future AI welfare scenarios — to ground discussion in vivid possibility. Co-designers responded with their own visions for empathetic, accessible AI welfare services.

Video prompt sequence strip showing five scenario stills, plus in-person workshop photo and online workshop showing AI Meeting Bot on screen — Stage 4: speculative design fiction video prompts — eliciting visions of AI welfare futures from co-designers

Five video prompt stills side by side on a dark background depicting AI welfare scenarios — The five video prompts — from eligibility denial to AI meeting bots to inaccessible healthcare systems

Open-Source Workshop Toolkit

A fully open-source toolkit enabling others to run co-design workshops on digital welfare access. Includes accessible card decks, printable facilitation materials and five speculative design fiction video prompts. Download and use freely.

Four Advokit card types shown in a 2x2 grid: yellow BARRIER card 'It's physically difficult', cyan WHERE card 'Basic phone or landline phone', purple HELP card 'Blue Badge scheme', and pale yellow FEELINGS card '++ very positive' — The four Advokit card types: BARRIER (yellow), WHERE (cyan), HELP (purple) and FEELINGS (pale yellow)

Advokit Card Deck PDF · Print & cut Difficulty–Value Ranking Grid PDF · Printable Journey Mapping Sheets PDF · Printable

Video Prompt 1 — A Future of Care Pam, Nadia and Yanni navigate AI-mediated care systems including driverless cars, telepresence robots and voice-activated smart homes. Produced by Superflux.

Video Prompt 2 — Press 1 to be Denied Eileen, 75, from Haringey, calls the National Mobility Support Line to request a Blue Badge.

Video Prompt 3 — Your Eligibility Could Not be Confirmed Maurice, 73, from Stratford, is assessed for PIP over a video call with PIPA, an AI assistant.

Video Prompt 4 — Just Want to See the Doctor Mark, 82, from Islington, struggles to book a GP appointment through an NHS app that redirects him to a pharmacist.

Video Prompt 5 — We'll Sort it Out Together, Brenda Brenda, 65, from Bromley, uses Homely, her Council Housing chatbot, to request home adaptations.