What’s at stake:
There is still a large gap between the aviation industry’s knowledge about the human factor in automation, and what automakers are implementing as they begin to design more human-centric vehicles.
When a crisis strikes either in air or surface transportation, is it automation technology that comes to the rescue, or is the hero a well-trained human?
The aviation industry, whose aircraft fly as much as 95 percent of the time on autopilot, has figured out the formula: the more automation operates in the cockpit, the more support and training becomes mandatory for the pilot. Automation does not reduce the need for human involvement.
As the automotive industry rolls out highly automated vehicles, meanwhile, it is pushing a contradictory narrative: automated vehicles make roads safer by replacing unreliable human drivers. Even if not fully self-driving, partially automated cars minimize human access to the steering wheel.
The recent web panel, “The Human Factor in Aviation and Vehicle Autonomy,” revealed the two industries’ divergent views on the human factor. It highlighted the gap between the aviation industry’s experience over the last three decades and the incomplete knowledge of automakers as they attempt to design more human-centric vehicles. Experts on the panel included Mike Lenné, chief science & innovation officer at Seeing Machines, Matt Gray, former captain at Qantas, and Bryan Reimer, research scientist, Massachusetts Institute of Technology, AgeLab.
Improving safety in the air
Aside from the introduction of the jet engine in the 1950s, the introduction of digital instruments – known as the “glass cockpit” in the 1970s – and the advent of fly-by-wire technology in the 1980s, are the engineering achievements most responsible for safety improvements in aviation.
Nonetheless, as Captain Gray acknowledged, Australia in the ’70s recorded “a number of accidents with … pretty fully serviceable aircraft, and yet they still crashed.”
They knew something was wrong. Gray said. “It turned out that … about 70 percent of aviation accidents are human error.”
The industry realized that aviation had to change. The aviation industry went about fixing the problem by focusing on human dynamics. Improvements followed several generations of human-factor training programs. “Today, typically, we look for nine competencies for a pilot. Only two are technical. The rest are non-technical things like situational awareness, knowledge, decision making, teamwork, leadership. … These things underpin a pilot’s technical training,” said Gray. Human-factor training in aviation now is nothing like “when I got trained years ago in the Air Force in 1980,” Gray said. Then the emphasis was technical training.
Today, airline pilots operate aircraft that can mostly fly by themselves.
But, to think that automation reduces the need for human expertise is “a myth,” stressed MIT’s Reimer. “I think the automotive industry doesn’t quite appreciate what the aviation industry has learned over the last 30-plus years,” said Reimer. “As we automate, we need to be thinking about how we support the driver more and more.”
How best to support the driver in real time, however, is poorly understood, let alone codified.
Sensing technologies, like one that Seeing Machines is developing, should help, said Reimer, “to sense driver’s state, and begin in real time to deploy countermeasures to support humans making better decisions.”
This won’t be easy for carmakers. Although most operators of highly automated vehicles clearly need more offline training before they take the wheel, “we’re never going to be able to deploy around consumers driving automobiles a training regimen that the aviation industry has been able to develop,” observed Reimer.
Human factors science
The general rule of the human factor is human behavior, but what exactly is “human factors science”?
Seeing Machines’ Lenné explained, “Fundamentally, human factors [science] is about understanding people and designing systems to support them.”
Those who turn up at work or begin any activity bring their own experiences, expectations, abilities and limitations, he noted. “They’re doing a particular task, which might be easy, hard, long, short … in a particular environment.”
Human factors science studies “the intersection of people with the tasks they’re doing, and the environment they’re doing it in,” so that “we can put people at the heart of design,” he said.
As analog instruments gave way to glass cockpits in aviation, pilots had to completely change the way they scan instruments. A flight deck today is visually denser, presumably providing pilots more information.
That’s not all good news, however.
A pilot today must identify the information vital to the immediate outcome he is trying to achieve while disregarding information not relevant at the moment. “It takes a lot of practice to do it, and we are spending a lot of time training pilots, 16 hours per year in simulators,” said Gray.
Human drivers in automobiles also face similar challenges. For drivers licensed at 16, 17, or 18 years old, there is precious little training available. Reimer said, “I’m a big believer that a little information can go a long way.” So, even if 16 hours of annual simulator training a year is too much for the average consumer with a car, one hour every five years to at least update people on new technologies that are available and how to use them in a changing environment, would be very important, he added.
Asked if carmakers might start removing unimportant information from the digital cockpit, Reimer said, “Unfortunately, no.” Carmakers are moving in the opposite direction. Feature creep is the norm. “As consumers, we are all different, wanting different things. And carmakers are making sure that everything exists in the vehicle.”
Nonetheless, it is increasingly important for carmakers to tailor all that information to provide what’s needed and relevant at a given time, added Reimer.
The aviation industry is aware that pilots are subject to “mode confusion.” The antidote is “flight mode annunciation.” When a pilot pushes a button on the glare shield and switches to autopilot, an instrument will inform him that the plane is now in this particular vertical mode. The pilot is required to announce the change followed by a cross-check with other crew members saying “Yes, we are both in agreement,” explained Gray.
All these procedures are effective in normal operation. “As soon as the workload starts to peak and you get a time-critical problem, your working memory gets stuck and your capacity shrinks. And one of the things that you will miss is probably the automation mode,” said Gray. This is complicated by the number of sub modes within the automatics.
As Gray noted, “Automation, in some ways, has actually increased our workload from time to time. Poor mode awareness – generally close to the ground on takeoff or landing – has completely increased the complexity of the operation, and there are a number of incidents where automation has not helped.”
Consumer drivers are much like pilots, observed Reimers. “Consumers are visual learners. They look for visual confirmation when they turn on autopilot.” But when the icon change is too subtle or it suddenly disappears, the driver can get confused.
Different carmakers have different strategies to mitigate mode confusion. According to Reimer, GM, for example, with its Super Cruise, a hands-free driving system, appears to have decided early on to minimize mode confusion by not offering lateral assistance when hands are not on the wheel. On the other hand, Ford’s Bluecruise has both hands-on and hands-off features. One worry is whether consumers understand when to grab the wheel and help the car, said Reimer. Such Human-Machine Interface issues are a subject of human factors science and one of the things the AVT Consortium, organized by the MIT AgeLab, has been studying, said Reimer.
Who’s the boss?
One of the moderators of the Web Panel asked Gray who’s the boss in an aircraft – automation or the captain? Gray was unequivocal: “You’re asking a captain, he’s a captain, he’s always in charge being supported by the automation. You never want the automation to take you anywhere.”
He stressed, “Automation is only as smart as the information that I put into it. I am ultimately the aircraft captain, totally responsible.”
Gray added, “But I really have to know exactly what that automation is doing at the time. I need to decide if it’s appropriate for me to use automation at a particular point. That’s what I learned from training.”
Gray’s assertion begs the question. How many human drivers operating highly automated vehicles would speak with such clarity, especially bearing in mind that carmakers of L3 models are advertising vehicles that encourage human drivers to do something with themselves other than driving?
The bottom line:
Human-centric design is the only way to make both aircraft and automobiles safe. But there exists a huge difference between the aviation and automotive industries in regulatory environments, the number of vehicle manufacturers, and the availability of off-line training programs. If consumer drivers aren’t willing to go for off-boarding training, automakers will have to figure out a way to use the systems to train the individual on-board. That requires getting the computer scientist, the AI engineers, and those who understand humans together to develop much stronger human centered systems.
Junko Yoshida is the editor in chief of The Ojo-Yoshida Report. She can be reached at firstname.lastname@example.org.
Copyright permission/reprint service of a full Ojo-Yoshida Report story is available for promotional use on your website, marketing materials and social media promotions. Please send us an email at email@example.com for details.