Bachelor Thesis · UX/Product Design
Passage — Routes that fit you
An AI-assisted navigation concept helping wheelchair users trust accessibility information in a city shaped by war.
Background
Ukraine's disability population has grown faster than its cities have adapted. War adds layers most navigation apps were never built for: structural damage that turns a safe route into rubble, power outages that disable lifts for hours or days, and air-raid alerts that mean the fastest route to a shelter matters as much as the most accessible one.
3M+
officially recognised people with disabilities
8%
of streets are barrier-free
37%
of displaced households include a person with a disability
Problem
Existing tools weren't built for a city where accessibility itself is a moving target.
- Structural damage can make a previously accessible route impassable with no warning
- Power outages, common in winter, take lifts and elevators out of service — the exact infrastructure wheelchair users depend on
- Air-raid alerts add urgency: reaching a shelter quickly can matter as much as reaching it accessibly
- Curfews and unpredictable disruptions mean the "safe window" to travel isn't constant
Research
16 semi-structured interviews — wheelchair users, mothers with strollers, older occasional city visitors, and accessibility organizations — analyzed with Reflexive Thematic Analysis.
Trust, not tech, is the barrier.
People don't distrust apps — they distrust unverified data.
Binary labels erase real users.
"Accessible / not accessible" ignores everyone who doesn't fit the average case.
Accessibility is dynamic, not fixed —
reshaped day to day by damage, outages, and alerts.
Trust is visual.
Photos and timestamps outrank written descriptions every time.
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
Why an app
Ukrainians already navigate their cities on their phones. An app can update in real time, meet people where they already are, and layer personalization on top of existing data — something a print solution or website can't do fast enough for a wartime context.
Who it's for
Primary
Wheelchair users in Ukraine, from new to experienced.
Secondary
Parents with strollers, older occasional visitors — their needs cross-validated design decisions but weren't the design target.
Out of scope (deliberate)
Visual/hearing impairments — this project focuses specifically on physical/mobility navigation.
Benchmarking
Audited navigation and accessibility apps across Ukraine, Germany, Austria, Japan, and Canada.
- Onboarding personalization rarely reaches the actual route shown
- Accessibility data exists, but scattered — never as a route-level trust summary
- Reviews are either too shallow to trust or too demanding to finish
- Dynamic barriers (crowding, sudden inaccessibility) are ignored almost universally
- Binary labels consistently exclude anyone who isn't the "average" case
Bringing the research question to life
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
How can personalized, AI-assisted navigation make accessibility trustworthy for wheelchair users in unpredictable urban environments?
Validation
Presented in Odessa to accessibility auditors, then tested with 8 users in Think-Aloud sessions — several of whom were also original interview participants, giving a direct check between what people said they needed and what they actually did.
Presenting Passage at accessibility event in Odesa
Process & tools
Design Thinking process: research → define → benchmark → design → test → refine, repeating as new findings surfaced.
Designed in Figma, then taken further than a click-through — coded into a working front end with Claude Code. AI tools supported different stages: Lovable and v0 for ideation and UX/UI exploration, Claude Design for layout variants on data-dense screens.