The other morning, I was driving to the airport and realized that I had forgotten my cell phone. I panicked. How would I stay in touch with my wife for the next four days? How would I be able to keep track of my schedule? How would I board the plane without my digital boarding pass? What important events would I miss on social media? How would I manage my hotel reservation? I was in the middle of rush hour traffic and knew that if I turned around and went back home to get my phone I might not make my flight. I turned around anyway.
Since the first primitive human picked up a rock and fashioned it into a tool to change his or her environment, we humans have set ourselves apart from other species by developing and using tools to advance our way of life. Nowhere has the use of complex tools shaped our lives more than in transportation and communication. We have developed increasingly complex tools that both improve and dictate our daily lives. Clearly these innovations have made our lives easier, connected us more with each other, and made the world a smaller place.
We live in an increasingly automated world. Digital assistants, self-driving cars, cleaning robots, package-delivering drones, and AI customer service agents are just a few of the automated features in our modern society. We interface with some form of semi-automated or intelligent machine almost from the time we wake up in our smart houses until the time we set our alarms on our smart phones before going to bed. We are never far from a host of other wireless devices designed to make our life easier, better, and more connected. But have the tools themselves transitioned from servant to master? What happens when these intelligent machines become smarter than we are?
The recent deadly accidents involving two Boeing 737 MAX8 airplanes highlights again that the epic tug of war between the human and the machine continues. At one end of the rope is the creative, adaptable, and intelligent human. At the other end is the increasingly cognitive and almost sentient machine. It is a battle that will continue as long as humans employ complex machines to do their bidding. The full investigation of these two accidents will yield pertinent information about the design, training, and employment of the aircraft, but already we can glean several important lessons from the accidents that can be applied anywhere we employ intelligent machines to improve human life.
Before I continue, let me state categorically that I love the Boeing 737! It has been the most successful airliner in the history of aviation. I have operated it safely for well over 12,000 hours across this great continent and to Hawaii and back. I have flown it into low visibility weather, short runways, and heavy rains. I have flown old versions ready for the boneyard and new ones fresh from the factory floor. I have also deadheaded (a passenger in the back being transported to my next assignment) or jumpseated (on the fold-out seat in the cockpit) for thousands of hours as well. I have flown multiple versions—the 200, 300, 500, 700, 800, and yes, the MAX8. The Boeing 737 has always brought me safely home.
Not only has the Boeing 737 always brought me safely home, it has been the workhorse that has afforded me a comfortable lifestyle as a professional pilot. It has put food on my table, put diapers on my young children, and put my young adults through college. It has allowed me to take my family on good vacations, drive nice cars, and pay for braces for all my children. The Boeing 737 has been the economic engine that put a nice roof over my head and funded my 401K. That machine has afforded me a lifestyle that otherwise may not have been possible. This discussion is not intended to cast stones but to highlight lessons we can all use as we develop and employ the intelligent machines in our lives.
In our modern world, we shouldn’t be thinking in terms of which is better—human or machine. Instead we should be asking, “How can we develop machines that humans can employ to advance our way of life?” It should not be a question of human or machine but more a question of human and machine. Then our focus productively moves from the question of why should we employ these machines, to the question of how should we employ them. How do we interface with smart machines in a way that retains our humanity and advances our way of life?
The world of aviation has struggled with this question for many years, particularly as the automation tools in the flight deck have become ubiquitous and dominant. The proper employment of automation has helped make aviation the safest form of travel on the planet, statistically speaking. Pilots have learned to use automation as a tool, without allowing it to become the master. In the process of improving and employing automation in commercial aviation, pilots developed automation philosophies, policies, and procedures that ensure the proper employment of, and interface with, the aircraft automation. These rules have not been developed in a vacuum or in some ivy-covered tower of academia, but in the real world where the cost of the lesson is paid in the blood of pilots and their passengers along with the capital of crumpled aluminum and broken glass.
Essentially, three basic principles provide a framework for managing the tug of war between the human and the machine.
· The human should possess the same basic skills performed by the machine.
· The human and the machine should be able to interface easily.
· The human should always have the power (and knowledge) to override the machine.
The human should possess the same basic skills performed by the machine.
Smart machines are wonderful tools that accomplish complex, repetitive, or mundane, tasks efficiently. They unburden the human from complex tasks such as in-depth mathematical equations. They reduce the workload of repetitive tasks such as routing incoming phone calls to the right department. They liberate us from mundane tasks like washing dirty dishes. However, what happens when the machine malfunctions?
Because of the more powerful, and more efficient engines on the Boeing MAX8, an adjustment had to be made to the flight controls to inhibit the aircraft from entering a stall (a condition where the wings are no longer producing lift and the aircraft begins to fall out of the sky) in certain unusual situations. Designers of the new airplane relied on intelligent technology operating in the background (without the knowledge of the pilot) to keep the aircraft safely within the flight envelope. The engineers and designers called it Maneuver Characteristics Augmentation System (MCAS). If the thrust of the new, powerful engines forced the nose of the aircraft past a predetermined angle, the MCAS adjusted the pitch by moving the large horizontal stabilizer and pushing the nose of the aircraft downward to prevent a stall. In and of itself, this design feature is robust and almost transparent to the pilot, until something goes wrong.
Computers, no matter how intelligent, are only as good as the sensory information they can gather. Garbage in. Garbage out. In the case of both the Lion Air and the Ethiopian Air accidents, it appears that the angle of attack sensor (the sensor that warns of an approaching stall) was giving the flight computer erroneous information. With no other sources of information to crosscheck, the computer activated the MCAS and began forcing the nose downward to prevent a stall that was not occurring. Additionally, other false warnings sounded and displayed in the cockpit. The pilots became disoriented and did not analyze the problem correctly or apply the designated corrective action. Like the computer, the sensory information they were receiving did not correlate with their knowledge and they were unable to save the airplane. Garbage in. Garbage out. Tragically, everyone onboard the aircraft paid for the breakdown of both machine and human with their lives.
When the machine malfunctioned, the human in the equation did not apply the basic skills to continue safe operation of the machine. Why they were unable to do so is not clear yet. The training and ability of the pilots will come under scrutiny, as it always does in an accident. Likewise, the distractions caused by the malfunction and the design features of the pilot interface with the aircraft will be analyzed. Changes in the aircraft design will not be enough to ensure safety. After the investigation, professional pilots will study the report. Every pilot of the Boeing 737 MAX8 will receive training to handle the malfunction if it occurs again. They understand that their lives, and the lives of their passengers, depend on the basic skills of the human operator.
What happens when we no longer know how to do simple mathematical equations, drive a car, or fly an airplane? If we rely entirely on the increasingly capable and intelligent machines to perform complex and dangerous tasks, we may find that the human ability to perform those same tasks will atrophy and perhaps disappear entirely. We must ensure that the human in the equation possesses the basic skills to perform the tasks performed by the machine, or when the machines fail, we will not be able to survive.
The human and the machine should be able to interface easily.
When the automation does something unexpected, pilots will often jokingly say, “What’s it doing to me now?” It’s an indication that the pilots either lack knowledge, or the system does not allow for proper interface, or both. Either way, the result is technology in charge of the process, and pilots that have become passengers.
When the ill-fated Lion Air 610 took off, the stick shaker (a small vibrating motor attached to the base of the pilot’s yoke) immediately activated on the captain’s side. The annoying motor is a warning to the pilot that the aircraft is dangerously close to stalling and that a recovery is needed. In addition, the pilot’s screens indicated “IAS DISAGREE” a warning that the right and left side airspeed indicators (indicators that let the pilot know how fast the airplane is flying through the air) did not agree. Most likely, the captain’s instruments also displayed other unusual and erroneous information. The noise of the stick shaker and the unusual display information are powerful distractors, but professional pilots train for these situations and should be able to follow emergency procedures and safely land the aircraft. But what if you add one more thing?
In the case of the Lion Air accident, the one extra thing was the moving horizontal stabilizer trim. The horizontal stabilizer on the tail of the aircraft is controlled by an electric motor that can be operated manually by the pilot or automatically by the autopilot. The stabilizer is a larger aerodynamic control surface than the elevator controlled by the pilot’s yoke. Because it is a larger control surface, the horizontal stabilizer can override the elevator. In other words, if the nose of the aircraft is pushed down by the horizontal stabilizer, the pilot can pull back on the yoke as hard as he wants and the aircraft will still enter into a dive. This is what happened when the MCAS, being driven by erroneous information from a bad sensor, pushed the aircraft into a dive using the horizontal stabilizer. The pilots had the power to turn it off (and should have), but because of poor interface and the distractions caused by the malfunction, they did not. (Note: A jumpseating pilot on the same airplane the day before recognized the malfunction and instructed the crew on how to alleviate the problem. This is most likely because of the vantage point he or she had when the malfunction occurred.) (https://reports.aviation-safety.net/2018/20181029-0_B38M_PK-LQP_PRELIMINARY.pdf)
The more that automated systems confuse or distract the human, the more dangerous and ineffective those systems become. Ease of interface with the machine is essential.
The human should always have the power (and knowledge) to override the machine.
There’s an old joke among pilots. The flight deck of the future will have one pilot and a dog. The pilot will be there to monitor the aircraft automation and ensure that it performs correctly. The dog will be there to bite the pilot if he tries to turn off the automation and actually fly the airplane. Humorous as this sounds, the human should always have the power, and knowledge, to override the machine.
Somewhere during the certification process of the Boeing MAX8 the engineers had to decide how much information to tell the pilots that would be flying the airplane. Too much information and the FAA might require lengthy and unnecessary training. Not enough information and the pilots would not understand what the aircraft was doing when an automated system took control. It is reported that Boeing “…decided against disclosing more details to cockpit crews due to concerns about inundating average pilots with too much information—and significantly more technical data—than they needed or could digest.” (https://theaircurrent.com/aviation-safety/what-is-the-boeing-737-max-maneuvering-characteristics-augmentation-system-mcas-jt610/) It was assumed that a malfunctioning MCAS would appear like the similar malfunction of a runaway horizontal stabilizer in which all pilots are trained. They walked a dangerous tightrope between education and information overload.
In Boeing’s defense, the decision to keep updating the B-737 and not redesigning it entirely provided pilots with experience in previous models a knowledge and experiential base from which to learn the new iterations. Although this philosophy comes with the inherent danger of not explaining significant changes to newer generations of the aircraft, it is countered by the improved safety of familiarity. In other words, familiarity of pilots with past versions of the aircraft increases the chances of safe operation of future versions of the aircraft, as long as differences are properly explained and trained.
According to the accident report, the pilots of Lion Air 610 counteracted the MCAS over twenty times before it caused the aircraft to crash. (https://spectrum.ieee.org/riskfactor/aerospace/aviation/indonesias-safety-committee-releases-preliminary-report-into-lion-air-crash) Each time the MCAS forced the nose of the aircraft downward by using the horizontal stabilizer trim motor. Each time, either the captain or the first officer used the manual switch on the yoke to override it. Boeing procedures dictate that when the stabilizer runs away, that the motors controlling it should be turned off using two small switches on the throttle quadrant. In effect the pilots revert to manually controlling the large horizontal control surface and don’t allow the automated system to make any further inputs. Because these switches are critical during a malfunction and runaway of the horizontal stabilizer, this cutout feature gets tested during every initial preflight check.
It appears that the pilots were overwhelmed by the other distracting events in the cockpit and did not recognize that the stabilizer was moving without their input. As the flight progressed, for some reason they stopped countering the automated response of the MCAS with the manual switch on the yoke and the aircraft became uncontrollable. They had the power to override the automation, but perhaps because of insufficient training, confusion, and distraction caused by other warnings, it appears they didn’t.
Automated systems should never have ultimate control of the machines we operate. Humans should always have the power, and knowledge, to override the machine.
In spite of these recent high-profile accidents and the tragic loss of life, air travel is by far the safest form of transportation. According to Ian Savage, a professor at Northwestern University, between 2000 and 2009 air travel accounted for 0.07 deaths per billion miles of travel. Cars accounted for 7.28 deaths per billion miles. (http://www.cityam.com/215834/one-chart-showing-safest-ways-travel) You are one hundred times more likely to be part of an accident driving to and from the airport than you are while flying on your flight. This safety record didn’t happen by accident. It is the combined effort of aircraft manufacturers, government regulators, airlines, and the pilots that fly the airplanes. We will take lessons from these tragedies as well to ensure that they don’t happen again.
Advances in automation have helped make aviation the safest form of travel today. It is the combination of human and machine working together in harmony that allows for that safety. The tug of war between man and intelligent machine will continue, but in the end if we don’t follow some of the basic principles learned through these tragic accidents, we will lose that tug of war and humans will pay for it with more loss of life. Even worse, we may lose our own human autonomy to intelligent machines.
The most optimum way to move forward in this age of increasing technology is a combination of human and machine where the machine serves the human, not the opposite. We must be able to employ increasingly complex and capable tools without fear that they will one day become our masters. Like the first human that fashioned a rock into a tool, we must maintain basic skills, learn to properly interface with tools, and when necessary, abandon the tool that no longer serves our best interest.
I hurried home that morning and got my cell phone because it has become an integral tool in my everyday life. I only made my flight because I employed another new technology. I drove my Tesla in the HOV lane. The combination of human intellect and intelligent machine saved the day.