Artificial intelligence and adaptability—not just automation—need to be key components in the design and functionality of autonomous vehicles
As a frequent Uber user, I’ve experienced every kind of driver on the spectrum. Some take great care to provide a relaxing experience and a smooth, leisurely ride. Others say very little and drive assertively as a way of demonstrating their focus on getting me to my destination quickly. Depending on my mood, the weather, traffic levels and a host of other interfering factors, there are days I appreciate the first driving style, and others when I strongly prefer the latter. And when a driver’s style doesn’t match my current preferences, it can make for an uncomfortable or even agitating ride.
A driver’s ability to make contextual adjustments to a passenger’s needs is so determinant of the passenger’s experience that it arguably affects the quality of the ride more than shock absorption, climate control, or other elements we traditionally consider central to the driving experience. So, if autonomous vehicle engineers want to achieve long-lasting adoption of their artificially intelligent “drivers,” they have to provide contextual learning abilities.
Perceived safety vs. engineered safety
So far, much of the chatter around vehicle automation has focused on the engineering challenges of getting passengers safely from point A to point B, what we might call Engineered Safety. However, autonomous drivers will have to go beyond just “keeping it between the lines” and behave in ways that build Perceived Safety, the experience of feeling safe amid ever-changing external and personal conditions.
For example, let’s say you’re riding in a self-driving car on a foggy evening. Your car’s sensors are not affected by fog in the way the human eye is. Therefore, it can drive relatively normally in these conditions, navigating sharp turns, passing slower vehicles and driving at speeds typical of a bright, sunny day. But as the human passenger, how does this make you feel? You may completely trust your vehicle’s sensory capabilities and feel perfectly comfortable. Or your physical fear centers may override your logic, making you clutch the seat with white knuckles. If you don’t yet trust autonomous vehicles, your reaction may be much more severe.
Perceived Safety is a subjective measurement, and it also isn’t static—the same individual may feel comfortable driving rapidly in the fog but want to take more caution in the rain. Furthermore, the same person can be adventurous one day and anxious the next. We are not logical machines, and we do not have uniform perceptions of danger or reactions to it.
Perceived Safety plays a major role in our experience when being chauffeured by an autonomous driver or human one, but it isn’t the only factor. Our mood can also affect our expectations for the way we’re driven.
We see mood-based choices available in luxury vehicles even today, with many offering drivers the option of either a “sport” or “eco” mode depending on their current preference. Imagine you’re running late for work. You may appreciate your autonomous vehicle positioning itself more assertively within the flow of traffic and passing slower vehicles sooner in order to get you to the office on time. By contrast, if you’re sharing a ride with your family on a Sunday afternoon, you might prefer the vehicle to simply follow the flow of traffic.
In both of these scenarios, it would be reasonable to expect your preferences be met, whether you’re being chauffeured or driving yourself. Likewise, for autonomous vehicles to provide a comfortable experience, they’ll need to be highly adaptable to feedback from passengers.
Consider as well the diverse and dynamic needs of delivery companies. Autonomous trucks will need the ability to adjust their driving style based on the payload they’re carrying — while they may want to prioritize speed over smoothness when carrying a load of textiles, they’ll need to do just the opposite when transporting animals, liquor or hazardous materials.
Artificial Intelligence, not automation
Because of the overwhelming number of nuanced scenarios that can affect Perceived Safety and Contextual Comfort, the ability to engineer an autonomous chauffeur depends on machine learning capabilities. Autonomous vehicles will not only need to learn their passengers’ baseline comfort levels but the most common variations for each person who uses the car. They’ll also need the ability to adjust on the fly based on the context of the moment.
To guide them in this learning, passengers will need the ability to communicate with their vehicles as they would a human chauffeur. This may mean vocally interacting with a Siri-style A.I., or it may mean selecting from a menu of options. It will also mean remembering settings for specific scenarios and defaulting or suggesting those settings when the context appears again. For instance, your vehicle should revert to the assertive driving style the next time you’re late for work.
Perceived safety and contextual comfort as brand differentiators
In the same way that many luxury automotive brands have built their reputations on offering the smoothest ride, Perceived Safety and Contextual Comfort will be prime differentiators among autonomous vehicles.
Vehicle brands will be ranked and categorized on their ability to react and learn their owners’ preferences. If Brand A syncs your vehicle’s OS with your calendar so it can drive more assertively when you’re late for an appointment and more leisurely when you’re on vacation, it’s going to have an advantage over Brand B that drives safely, but consistently in every situation. And if Brand C slows down when your mother is in the car but Brand D’s racecar driving style frightens her away from autonomous vehicles forever, there’s no question Brand C will find a larger audience.
In other words, contextual decision making powered by deep learning will offer the opportunity for not only wider adoption of driverless vehicles, but brandable experiences.
So as manufacturers increasingly solve the Engineering Safety challenges of vehicle autonomy, it’s imperative they also address Perceived Safety and Contextual Comfort. These will be the factors that ultimately decide which autonomous vehicles succeed and fail, and how long they will take to reach widespread adoption.
Even as we begin to trust artificial intelligence, we will continue to run on survival instincts, which can be tuned differently from one person and one day to the next. Contextual control is an essential part of the freedom that comes with driving or riding in a passenger vehicle. Moreover, without the ability to control the experience, individual vehicles become less practical in many circumstances, both for passengers and commercial cargo.
As we enter this new era of smart mobility, it’s vital that engineers recognize a simple truth: We aren’t likely to become logical machines based on the limitations of driverless vehicles. So it’s essential that driverless vehicles become more human.