(Also published on Linked-In)
Imperfect automation, continually getting better? Or distracted drivers, continually getting worse? Choose.
I am fearful of the rapid rush toward full automation and have published numerous articles about the difficulties we will face because of the mismatch of the automation and human behavior. However, I am even more fearful of the rapid rise of distracting devices installed in automobiles, mounted on dashboards, worn on the wrist or body, or carried on seats, pockets, and laps of drivers.
Automation of automobiles today is imperfect. The best vehicles, for example those designed by Google, are prototypes, still far too expensive for mass manufacturing. The automation now on the road and under development by the traditional manufacturers is remarkably good, especially on the highways, but still unable to manage all the complexities of city driving (a task that average drivers also have difficulty with).
Human driving is imperfect, hence the large number of deaths and injuries: Driver distraction is on the rise. Whether it is adjusting the car's indoor temperature, changing the music being played, or reading or answering texts, distractions take the eyes off the road for considerable time. How much is too much? At 60 mph (100 kph), in one second the driver has gone roughly 90 feet (30 meters). In the city, at 6 miles/hour (10 kph), this translates to 9 feet (3 meters): one second of distraction is quite enough to collide with another vehicle on the highway or run over a pedestrian in a crowded street.
Each day seems to bring a new distraction. Heads-up displays (HUDs) that once were aids to minimizing distraction by making it easier for the driver to see navigation aids and speed, are now catching featuritis, that deadly disease which corrupts products. Now HUDs show information about the song being played and at least one company proposes being able to do videoconferences with the image of the other person hovering in the air in front of the automobile. We know from aviation's use of HUDs that when reading the display, people do not see objects on the roads (or runways), even though their eyes are pointed right at them.
Automation is imperfect, but it is continually getting better. Distractions are dangerous, and continually becoming more pervasive, more numerous. This leads to the obvious conclusion:
Today's imperfect automation is preferable to today's distracted and imperfect drivers. The sooner we can get automated cars on the road, the better.
We know that today's automation is not capable of handling all the situations that take place on the roads today, especially in crowded cities with erratic behavior and on badly maintained roads where lane markings are non-existent and even road boundaries are difficult to discern. The biggest hurdle facing full automation today is dealing with the unregulated, largely unpredictable behavior of other road users, coupled with the complexities of the transition, when we will have older, manually controlled cars along with semi-autonomous vehicles sharing the road with fully autonomous vehicles.
Partial automation has its problems, the major one being the requirement that a human driver always be attentive, ready to take over when things go wrong. This is impractical: the better the automation, the less attention drivers will pay to the road, and even a one or two second delay in regaining control is too long for safety and too short for most people to understand the situation, respond, and have the car behave appropriately.
Laws are ineffective against distraction. People will talk, text, read, sleep, eat, pick up dropped items, and even change clothes while driving. The evidence is there: it is already happening. I predict that within the next decade, automation will be good enough to reduce the number of accidents and deaths in the world. More importantly, automation will be on an ever-increasing trajectory of improvement in safety and reliability.
In the United States over 30,000 people are killed each year in automobile accidents. Over 1 million are injured. In the world, the World Health Organization estimates deaths at over 1 million and injuries between 20 and 50 million.
Think of automobile driving as a disease that kills and disables. If this were a biological disease that affected the same number of people, there would be an outcry and action taken to control the epidemic. When should we switch to automated vehicles? As soon as we are determine that automation can significantly reduce deaths and injuries. We will need to do controlled tests in order to know that the cure - automation - is indeed better than the ailment - human drivers.
After we introduce full automation, there will still be accidents, injuries and deaths. Worse, because of the tight coupling of one automated auto to another, when there is an accident it is apt to involve multiple vehicles with a large number of deaths and injuries. But the fundamental question is whether there has been a significant reduction in deaths and injuries. Even if 10,000-20,000 people per year die in automated vehicles, that would be a huge decrease in deaths. And every year the automated vehicles will become safer.
I have long argued that we need to go slow with automation in the automobile (e.g., The Human Side of Automation). There were still too many unsolved problems. I have now changed my mind. Why? Because there are far more problems with the increasing number of distractions for drivers, too many new devices, too many new temptations. Imperfect driving is potentially more dangerous than imperfect automation. Add to this the other benefits to those today who are unable to drive: the elderly, the handicapped, and of course the blind.
Automation versus distraction? I bet on automation, and the sooner the better.