When I was a child in India riding pillion on a scooter or motorcycle behind my dad or older brother, I would constantly be peeking over their shoulders braving the winds to get a view of what was coming up ahead. These peeks let me understand what was happening giving me trust in how my dad or brother was reacting to the conditions and making me feel safe.
This memory makes me think about what autonomous systems we trust today in comparison to the distrust that surrounds autonomous vehicles. We trust automated systems which do not have an operator on a day to day basis including trains, streetcars, shuttles, escalators, elevators, and so on.
We also get into taxis, Ubers or carpools with friends without any trust issues. In these cases we not only get to see how the driver is driving and generally responding to the vehicle’s environment, we also get to “see” the operational domain – where the car is, where it’s going, and the general path it is driving on.
How does the geo-fence – the operational design domain of the vehicle – apply in these scenarios and engender our trust?
For the train , it is the tracks and the location of the stations. For a motorcycle, it is the road ahead. The tighter the geo-fence – and the rider’s knowledge of it, the easier it is to create trust. In the case of the train, the bi-directionality of the train tracks makes it easy to trust. There are only two directions it can go. Plus, this geofence environment is naturally constrained and largely protected from outside variables. For instance, there is no overtaking and passing involved as you would see in highway driving.
When we think about AVs, the clearer the rider’s knowledge of the geofence is, the more they will trust the system.
Automated vehicle designers are applying a two-pronged approach to create trust in these systems:
- Start with a well-defined geo-fences. At a systemic level, creating geofences that are easily understood and definable, creates an entry point for users to build trust in AVs. Creating smart corridors in the interstate freeway system for AVs to operate within. Clear geo-fence boundaries will make people more inclined to trust them.
A great example is Waymo One, the driverless ride hailing platform available to the public in Phoenix, AZ. Riders who sign up to the platform are pre-screened before they can start requesting and using rides based on the zip-code they live in. Essentially only riders living within the geo-fence can use the service at the moment. (Source: https://techcrunch.com/2019/11/01/hailing-a-driverless-ride-in-a-waymo/).
Another example of a geo-fence is how DJI designates “no-fly” zones for its drones near airports. This is more important now with DJI drones having more automated features which let the drone pilot draw a path which the drone can then follow on its own.
- At the vehicular level, incorporating design that can help riders understand and perceive the vehicle’s interaction with the geofence to learn and accept he automation’s performance. This can include:
- Displays that clearly show what the car is not only seeing and perceiving but also, importantly, what it will do. This image of Waymo’s information screen shows passengers exactly what path (or fence) the vehicle will follow – even though passengers do not need navigation in a car they are not driving.
- Designing operations to replicate human behaviors and actions – for instance, creating a turning experience that feels “right” to driver/passenger
- Voice technology to keep riders informed and engaged with what is coming next.
Creating trust in AVs will be accelerated by designing automated systems from a collaborative POV thinking of the rider and the system as “co-pilots” in the experience with clear geo-fence definition, design and technology that inform and empower riders and onboarding/education initiatives to will build understanding and knowledge in automated systems.
Contributed by Shasank Nagavarapu, Senior Associate, Human Centered Design