Understanding mobility across age groups and what it actually takes to design Voice AI for real people in real cities.
MY ROLE
UX Researcher
TEAM
UX Researchers, Team Lead
DURATION
3 Months (Feb 26 - May 26)
TOOLS
CustomGPT (ChatGPT), Zapier, Google Sheets, FigJam, Miro, G-Suite
MY RESPONSIBILITIES
Research Recruitment, CustomGPT Build, Synthesis, Design Guidelines, Presentation Narrative
Context
The upcoming "Silver Tsunami" in NYC
New York City is getting older, faster. 16.6% of its population is already 65 and over; that’s 1.55 million New Yorkers aged 62 and above as of 2023, a figure that has grown more than 50% in the last two decades. Adults over 65 represent nearly 45% of the city’s pedestrian fatalities.
Current infrastruture supporting them
At the same time, the MTA is consolidating public transit into a single digital platform, and voice AI is already embedded in city life - Accessible Pedestrian Signals (APS) & AI voice companions (ElliQ). The accessibility conversation in urban mobility is still mostly about physical infrastructure: curb cuts, crossing times, ADA stations. But the next frontier is conversational.
Research Focus
How can voice AI help older adults navigate physical spaces with reduced cognitive fatigue while preserving independence and awareness?
Understanding interaction, trust, and decision-making in real-world navigation
The Instrument
We couldn’t answer these questions from a usability lab or a post-hoc reflection form. Public, conversational AI interactions don’t happen frequently enough in the open to observe directly. If we wanted to understand how people actually interact with voice AI during real navigation, we needed to be inside the conversation when it happened and not reflecting about it afterward.
So we built the instrument into the conversation itself by building a CustomGPT that served three simultaneous roles:
Study Device: A functional AI navigation assistant participants could actually use to plan routes and errands.
Prompt Facilitator: Built-in reflection questions that transitioned naturally within the same chat; no tool switching, no separate form.
Research Assistant: A silent analyst maintaining a running scratchpad of metadata, Likert ratings, and interpretive notes, all exported automatically via Zapier to a master Google Sheet when the participant said “Submit Entry.”
From the participant’s perspective, they were talking to a travel companion that helped them plan errands and occasionally asked how it went. The reflection happened inside the interaction, facilitated by the same voice that had just helped plan the route. This collapsed the gap that conventional diary studies can’t close.
We recruited six participants across age groups and technology comfort levels: from a 25-year-old comfortable with AI tools to a 70+ year-old who had never used a voice assistant for navigation. Screener surveys helped us map their familiarity with ChatGPT and caregiving responsibilities before onboarding.
Participant recruited and their relationship with technology (ChatGPT)
What We Found
Six insights emerged from the transcripts and several of them directly challenged assumptions we started with.
Navigation is the last step, not the first
No participant opened with a destination. They opened with an activity, an experience, a social intention. One participant wanted kayaking spots where she could rent equipment and enjoy the day. Another was coordinating a multi-stop trip around a friend visit and a dinner reservation. The route came almost as an afterthought. Voice AI that starts by asking “where do you want to go?” is already behind.
Older adults navigate by landmarks, not coordinates
Every participant who encountered unfamiliar street names, compass directions, or distances in feet showed immediate friction: confusion, frustration, disengagement. When the AI referenced landmarks and neighborhood names they already knew, navigation felt effortless. “When you said Ohio Street, I was just like, what? I don’t know where that is.” The information was technically correct. The interaction still failed.
Navigation is social orchestration
Almost no one was navigating for themselves alone. They were coordinating for partners, families, groups. The destination was secondary to who would be there, what they would do, and when it needed to work for everyone. Voice AI designed for solo efficiency misses most of what people are actually trying to accomplish.
Autonomy is non-negotiable
Across every participant, a consistent boundary emerged: discovery and research could be delegated to the AI, but decisions were kept. The moment the assistant moved from presenting options to making a choice, trust eroded. “Let me just try it first, okay, and then you can correct me or whatever.” The AI earns its place by supporting, not overriding.
Cognitive overload comes from the system, not the city
Participants weren’t overwhelmed by New York City. They were overwhelmed by the AI. Verbose responses, invisible processing time, and unexpected complexity added friction on top of whatever the environment was already demanding. “I think you were just rambling. I would have liked a quick response.” Brevity isn’t a preference, it’s a safety requirement.
Older adults trust first; younger adults verify first
This one surprised us. Our assumption going in was that older adults would show more friction - less tech familiarity, more hesitation. The opposite was true. Older participants were more patient, more willing to rephrase when misunderstood, more open to sharing location details. Younger participants were more guarded, cross-checking outputs, withholding home addresses.
Translating Findings into Design Guidelines
From the six insights, we developed ten design guidelines for voice AI navigation, grounded in three core principles:
Autonomy — the AI assists; the human always leads.
Trust — earned through accuracy and silence, lost through a single wrong answer.
Personalization — the AI speaks your language, knows your landmarks, feels temporarily yours.
The guidelines range from practical guideline like speak landmark language, not coordinate language; shortlist options, never decide — to systemic: design the exit as carefully as the task, and earn the right to speak by staying silent when there’s nothing useful to add.
We recruited six participants across age groups and technology comfort levels: from a 25-year-old comfortable with AI tools to a 70+ year-old who had never used a voice assistant for navigation. Screener surveys helped us map their familiarity with ChatGPT and caregiving responsibilities before onboarding.
Project Takeaways
Building the diary study tool changed how I think about research design.
Every decision about how the CustomGPT prompted participants, when it shifted into reflection mode, how it handled an incomplete entry, was a UX decision with real consequences for data quality. The instrument and the study were the same thing. That’s not how research usually works, and getting it right required treating the tool with the same rigor we’d bring to any product.
Conversation and language became the biggest priority in setting the study up for success
Users needed to feel heard and validated ; not interrupted, not redirected. The moment the AI took the lead, people pulled back. Control had to stay with the user, not the tool.
For Woven City and the future of urban mobility
More broadly, these findings point to a design opportunity that goes well beyond accessibility features: building AI systems that treat navigation as a human experience: social, landmark-anchored, autonomy-driven rather than a routing problem to be solved.