A VR study of the moment a car becomes an aircraft, built with Jiayi Chen at Georgia Tech’s Sonification Lab. The question: what makes a rider trust a vehicle that changes what it is, mid-trip.
We built a VR city where a car drives the streets, pulls into a VertiSpot, an urban pad for takeoff and landing, climbs to a flyway thirty feet up, then lands and keeps driving. The goal was never to look futuristic. It was to find what keeps a rider calm while the vehicle changes what it is.
“How might we build trust, comfort, and engagement in vehicles that switch between driving and flying?”
Trust research has a simple rule: people rely on a system only as much as it shows them what it’s doing (Lee & See, 2004). A flying car breaks that rule twice, once on the ground, once in the air. We needed to know exactly where.
How riders read safety while the car moves from driving to flying.
Which cues, visual, audio, haptic, keep a rider feeling in control mid-transition.
What changes between driving it and riding in it, and what road automation can teach aviation.
Rivian’s hands-free driving is smooth and predictable, closer to the calm handoff flying cars need than Tesla FSD, which still demands constant supervision and can turn abrupt. Super Cruise’s driver monitoring and the Boeing 787’s head-up display say the same thing: show people what the system is doing, and keep them in the loop.
We ran three scenarios: drive it yourself, ride as a passenger next to a visible driver avatar, or fly it with a game controller. We watched how people reacted to every mode shift, then asked them about comfort, trust, and risk.
The old Unity build ran on a paid replay library and undocumented features we couldn’t maintain. We rebuilt it on SteamVR for Meta Quest, HTC Vive, and Valve Index: a new render pipeline, every interface element redone for the headset, and a boundary system to keep riders inside the play space and the car.
NavMesh couldn’t drive a car. It ignored lanes and traffic rules, so we scripted the routes by hand instead. Then we built a record-and-playback system: drive it once, replay it for every participant, so everyone saw the same ride.
Trust hit 9 out of 10 on the ground, then dropped right before takeoff.
Trust landed at 7.5. A visible driver avatar beat an empty seat every time.
Felt like a game. Scored the lowest on feeling real.
“Trust rose when someone else was visibly in command, and fell when the participant drove or used a controller, since those roles felt closer to a simulation than a ride.”




Across sessions, one pattern held: predictability beat spectacle, and people wanted more signals, not fewer.
Riders wanted transitions signaled clearly and played out slowly. A sudden lane change with no warning killed trust fast.
People kept asking for speed, a predicted path, a confidence score for the system. Even a basic map helped them feel oriented.
A driver avatar, steady cues, and a smooth handoff built confidence. A silent one left people confused and uneasy.
A lighting shift or a directional cue did more for realism than better textures ever did.
Reading trust research. Rebuilding a VR platform from scratch under real constraints. Running a study across three roles. Then turning raw notes into interface rules for a vehicle category that doesn’t fully exist yet.
“A vehicle earns trust the same way a person does: by showing its intentions before it acts.”