B3 · 2025 · Behavioral design
Raahi
A voice companion for long-distance drivers that keeps them present, talks them into breaks at the right moment, and gives the passenger a way to help without nagging.
Read the full write-up on Medium- What I did
- Research with taxi drivers, the behavioural framing, the driver and passenger flows, an MVP video, prototype testing.
- Result
- Fifty drivers interviewed, a working interaction prototype on Arduino, tested with the drivers it was for. The fatigue sensing behind it was designed, not built.
Fifty drivers, and every one knew the risk
Highway hypnosis is the thing where you arrive and cannot remember the last twenty minutes of road. On a long, flat, monotonous drive it comes for everyone, and after bad rest it becomes one of the main reasons long journeys end in accidents.
The surprising part was that the drivers already knew. They recognise the risk and keep going anyway.
You arrive and cannot remember the last twenty minutes of road
Fifty conversations turned up four things. Stopping feels like the expensive option, because the fare is fixed and the clock is not, so drivers reach for a cigarette, an energy drink or loud music instead. None of that is rest. Break avoidance has three parts: underestimating your own tiredness, overconfidence, and not knowing where a safe stop is or what it costs you. Monotony is cumulative, so alertness drops and stress rises the longer the road looks the same. And past a point it becomes microsleeps, which is the moment nobody gets back.
An alert treats the driver as a hazard
Alerts arrive from a machine that does not know you, and drivers learn to swat them. The lever that worked in conversation was different. Something that keeps you company, notices when you have gone quiet, and suggests a stop the way a passenger would, with a reason and a place attached.
Scroll the diagram sideways for the rest of the unit. The unit sits on the dashboard and is read at a glance: a strip for energy, two keys for progress, one button for mode.
What it does, in the order a drive needs it.
- Fatigue tracking. Energy read from HRV sensors, fed back in real time rather than at the end of the trip.
- Games that break the monotony. Spot the landmark, lyrics, memory, count the billboards. Talking, not tapping.
- Rest stops with a reason. Suggested against the fatigue reading, so the stop is argued rather than nagged.
- Adaptive modes. Solo and passenger drives are different conversations, and one button says which one this is.
- Rewards. Incentives through the aggregator for safe, consistent use.
- Connectivity. Dash cams, sensors and the phone, so the feedback has more than one signal to work from.
A voice companion, not an alarm
Talk to the driver instead of warning them
Instead of Drowsiness alerts and forced break timers
Alarms treat the driver as a hazard to be managed. A companion treats them as someone to talk to, which keeps them engaged on a monotonous road, and monotony is the actual cause.
After The break arrives as a suggestion with a rest stop attached, so 'where' and 'how long' are answered before they are asked.
It rides along with the trip
Live in the car and the journey, not in another app
Instead of A standalone driver app
The solutions that get used slot into what people already do. A driver on a highway is not going to start a new routine.
After Using it costs nothing extra, so there is no new habit to sustain.
How a driver actually gets one
Raahi only works if it reaches the driver, and a driver is not going to hunt for it. So the flow starts inside the app they already open for work: the aggregator prompts them, points them at the nearest centre to collect the unit, and the pairing happens in the car.
The rewards screen is where the research actually lands. Fatigue advice that costs a driver income is advice they will ignore, so taking breaks has to pay: the tier is built from breaks taken as much as from trips completed.
The passenger gets a screen
Turn the source of the pressure into the source of the help
Instead of Designing for the driver alone
The passenger is the reason the driver does not stop, and also the person best placed to help. Giving them a view of the trip and the driver's state changes which of those two they are.
After Support instead of nagging, without asking anyone to police anyone.
What the road taught me that Figma did not
What felt like a gentle suggestion in Figma landed differently spoken aloud in a moving car. The pattern I keep: solutions that fit an existing habit get used, and solutions that ask for a new one get tolerated for a week.
I would test on real drives, in real cars, earlier. Every important finding came from behaviour, and behaviour only shows up on the road. I would also explain the fatigue score, because drivers wanted to know how it decided they were tired before accepting being told to stop, and a number you cannot argue with is a number you ignore.