I, Robot

-->

The aviation world has been a bit aflutter the past two weeks over an article that recently came out in the mainstream press and was subsequently picked up by various media outlets. The article, available here, was reprinted with several alternate headlines, the most sensational and popular of which asked, “are pilots forgetting how to fly?” Predictably, the article has stirred up a variety of strong opinion among pilots, with attitudes ranging from outrage to derision to hearty approval.

I actually thought the article was fairly well done for something written by and for laymen. There were some factual errors and I don’t agree with all the author’s conclusions, but there’s a pretty big kernel of truth in there. I wouldn’t say pilots have “forgotten how to fly,” but there has been a pretty fundamental shift in our role and duties over the past 20 years, and we are only now seeing some of the unintended consequences of that shift.

Any discussion of airliner automation needs to begin with an affirmation of its value. I do not like that automation makes my job less interesting or that it has decreased the value of my skills as a pilot, but it has undoubtedly made aviation a great deal safer. There was a time when good pilots, real stick-and-rudder types, were crashing airliners with frightening regularity. No matter how “good” one is, humans are fallible creatures, and in the course of operating a complex machine will inevitably make mistakes. When these mistakes take place in times of stress, in environments that inundate one with information to be sorted and interpreted, they can compound unnoticed until no amount of airmanship can save the ship. Automation decreased the pilot’s workload, sorted the information, arranged it in more easily understood formats, and trained an unblinking eye on the pilot, alerting him to all his most common foibles. The results speak for themselves.

So while we’ve seen an uptick in accidents that can be attributed to automation – or rather, the way we’ve interfaced with automated systems – one would do well to remember how many lives have likely been saved by increased automation. That said, our strength as an industry has always been our willingness to ferret out our weaknesses and confront them head-on. A spate of loss-of-control accidents is disconcerting no matter how low the overall accident rate remains, and I for one am very happy to see “the experts” recognizing the problem and doing some hard thinking about how to solve it.

The degradation of stick and rudder skills is one negative side-effect of automation, but not the most serious one in my mind, despite the article’s emphasis of it. It varies according to the aircraft, airline, and personal preferences. The JungleBus has one of the more automated, integrated cockpits out there, but my airline takes a pretty reasonable line on the use of automation. They encourage its use but caution that it should only be used at the level most appropriate to attain the goals of safety, passenger comfort, and economy – in that order. To that end, it is left to the PIC to determine what level of automation to use at any given time, so long as the pilots remain proficient in the use of all levels of automation, including its complete absence. My last airline took a similarly common-sense approach. There are, however, airlines that require automation to be used to the maximum extent possible, particularly overseas. As an example, there is a well-known British airline that prohibits its pilots from using manual thrust on the A320 except in abnormal situations.

I’ve flown with pilots who prefer to turn on the autopilot at 1000’ on departure and leave it on until 200’ on approach. I’ve also known a few guys who, left to their own devices, would hand-fly everything raw-data from takeoff to touchdown. Most of those I’ve flown with, however, are like me: hand-fly down low when the weather is good, use the automation when the weather is crappy or when in busy, complex airspace. This typically results in anywhere between three and twenty minutes of hand-flying per leg. Beech 1900 types excepted, most regional and domestic pilots probably hand-fly anywhere from 2 to 10 hours per month, long-haul pilots considerably less. Most of this time is spent in takeoff, acceleration, climb, visual or easy instrument approach, and landing. Flight directors and autothrottles (where installed) are most commonly left on.

This alone does not lead to anyone “forgetting how to fly,” least of all anyone who had a decent bit of flight time before the airlines. The most autopilot-loving JungleBus pilot has to turn it off to land, and most do a beautiful job of doing so. It does mean, however, that if automation is unexpectedly lost – especially in an unfamiliar flight regime – handling the aircraft without benefit of autopilot, autothrottles, or flight director will probably not be second nature. It will require some concentration, potentially at a time when multiple anomalies of an ambiguous nature are clamoring for attention of their own. The master warning and caution lights have loud alarms that sound with each new occurrence, but only the quiet voice of your first flight instructor reminds you to “fly the airplane!” It’s understandable that sometimes the former get more attention than the latter.

In my opinion, however, the real problem with over-reliance on automation has less to do with stick-and-rudder skills and more to do with how it has affected our habits as professional pilots. Automation was supposed to free us to think more, but instead has freed us to think less. Consider the enroute phase of flight: it’s been so thoroughly automated, there’s precious little left for the pilot to do. Your course is drawn out for you, there is no doubt as to your position, updated weather is at your fingertips, your fuel state at next fix, destination, and alternate is right in front of you, and ATC has become much better about routing aircraft around heavy weather. All of these are good things – nobody is looking to give up positional certainty, accessible weather data, or fuel planning aids - but they are not conducive to keeping one alert and engaged in the flight. It’s too easy to get lulled into passivity, to turn your brain off and let things happen.

This technology-induced torpor can linger beyond cruise into the descent phase. In Dave’s thoughtful post on this same subject, he mentioned that a strong pilot should, when told to cross 40 west of a certain fix at FL250, be able to mentally calculate his descent point in three seconds flat – “tired or not.” Not so long ago, such mathematical process was a baseline requirement of the job; you wouldn’t survive on the line without it. Technological progress has made it possible for weak pilots to fit in. Worse, it has atrophied the brains of some formerly strong pilots.

The JungleBus, like “Fifi”, has a Vertical Navigation (VNAV) system that is very capable – when it works. It can be led astray by operator error, it has some insidious failure modes, and occasionally it does things that are just plain goofy. More than once, manually calculating my descent point as I received a crossing clearance saved my bacon – not from crashing, mind you, but certainly from violating my clearance. Yet, I often witness otherwise good pilots whose very first instinct upon receiving a crossing clearance is to enter it into the FMS. Worse, when asked where they plan to start down, many do not have an answer ready. The steady stream of ASAP reports for busted clearances my company receives (and this is true of other companies operating VNAV-equipped aircraft) suggests that my experience is typical. Yet, you are far more likely to be busted on a line check for failing to enter a crossing restriction in the FMS than you are for failing to mentally back up the FMS.

I do not wish to imply that the majority of airline pilots out there are weak or lazy. The opposite is true: most wish to do their job to the best of their abilities, and expend considerable effort in doing so. I am saying that we have been unintentionally training our pilots to value the technology in their aircraft over their own common sense. We have made it passé for a pilot to fly by his wits and then wonder why those wits have become dulled with disuse. This is becoming evermore true as we start pilots on technologically advanced aircraft ever earlier in their careers, to the point that we have pilots being hired at the regionals today whose very first flight lessons were in Garmin-1000 equipped Cessna 172s.

So long as the technology works perfectly, it’s a mere academic issue. Like the JungleBus’ VNAV, technology seldom works perfectly, and is always subject to outright failure. The real question, then, becomes “what happens when the automation quits?” How our pilots handle that is the real test of our reordered priorities. The recent cases of Colgan 3407, Turkish 1951, and Air France 447 are not encouraging. In each case the crew had an abrupt and unexpected transition from full automation to partial or full manual control, in the latter two cases coupled with insidious or confusing failure modes, and they did not handle them well. I can honestly say that just about every messed up situation I’ve seen in both the Q400 and the JungleBus happened because the automation abruptly quit or did something we weren’t expecting, and we handled the transition to manual control poorly. In most cases it was because we spent too much time asking “what’s it doing now?” and “how do we fix it?” instead of just pushing the big red button and flying it like the big 172 it is.

The good news is this: pilots like to be pilots. With some changes I think we can restore a proper balance no matter how advanced our aircraft are. Firstly, an adjustment in automation philosophy is in order (at some airlines more than others). I think my airline’s policy is a good starting point: fly at the level of automation most conducive to safety, passenger comfort and economy, and leave it up to the PIC which level best attains those goals. However, I think we should also emphasize the importance of regular exposure to all levels of automation, including manual thrust and raw data, in various phases of flight, in order to maintain proficiency in all levels. Training and checking should be conducted at all levels of automation, whether the FAA requires it or not.

Secondly, we should adjust procedures to keep pilots engaged in the more automated phases of flight. In cruise, I would suggest pilots be required to complete a navigational log, perhaps collocated with the expanded flight plan section of the release. Passing over or abeam each planned waypoint, the pilots would be required to enter time, fuel load, difference from planned time enroute and fuel burn, estimated time and fuel to the next waypoint & destination, and nearest suitable airport. This is already SOP at a few airlines, but not many. It may smack of busy-work to some, but I think it would be useful as a means of keeping pilots’ brains alert and focused on the flight, as well as serving as an invaluable backup in case of unexpected loss of navigational capability. With newspapers, crossword puzzles, laptops, Angry Birds, and napping all banished from airliner cockpits, I will only half-jokingly suggest that each pilot be issued a sextant and mariner’s almanac to take sightings to back up the GPS.

Lastly and most importantly, we need to adjust our training and checking to emphasize the necessity of brainwork. Technology and mental skill ought to be mutually beneficial and neither should be employed to the exclusion of the other. Simulator instructors and check airmen should make a regular practice of failing the automation in unexpected and artful ways as a means of ensuring that pilots are actively backing up their technology and are continuously prepared to revert to lower levels of automation. Ultimately, the most difficult thing about all this is that it will require a certain change in the training mindset at many airlines. With training footprints slashed to a bare minimum, the goal has become preparing the pilot to pass his checkride in a minimum of time. The focus needs to shift back to preparing the pilot for whatever life on the line throws at him, in particular the sneaky problems that have a way of snowballing unnoticed. Vee One cuts are serious and it’s good that we practice them, but they’re not particularly subtle, nor do they require much thought beyond rote repetition. We need to move beyond “checking the boxes” mode and include opportunities for real learning in every training and checking event. This will require more simulator time and therefore increased training budgets, but I believe the result will be more thoughtful pilots more attuned to their aircraft and better equipped to handle unusual problems.

While the industry sorts out the big-picture stuff, we as individual pilots can take a few simple steps to maintain a balanced relationship with the automation in our aircraft. First, use it as a backup to your own airmanship and common sense. Calculate your descent profiles mentally and then use VNAV. When ATC clears you direct to a fix, have a pretty good idea of what your heading should be so you can check the route your FMS or GPS is proposing for reasonableness. Secondly, make a regular practice of flying under various conditions with reduced levels of automation. You can put together a regimen. I try to fly at least one approach each trip without autothrottles, and one every other trip raw-data. I also occasionally practice fully-automatic RNAV approaches, something we do in real life very little. If you are an FO, explain your habit of practicing all levels of automation to your Captain at the start of the trip. If you’re a Captain, encourage your FOs to do the same.

My third suggestion is to ensure you are paying close attention to what is on your FMA (Flight Mode Annunciator) or equivalent display at all times. Return to it often. It should be as central to your scan as the attitude indicator. Many of the “what’s it doing now?” moments I’ve seen occurred because the autopilot and/or autothrottles were in a different mode than the pilot thought he had selected. And finally, when you do find yourself asking “what’s it doing now?”, make it your very first step to fly the airplane, reverting to lower levels of automation as necessary, and not succumbing to the temptation to troubleshoot until the airplane is going where you want it to.

All these guidelines are applicable to advanced airplanes from glass-equipped C172s on up through A380s. Flight instructors, drill them into your students from the very first flight lesson. I generally believe that glass cockpits in training aircraft are overkill or even counterproductive for early flight training. I may very well revise that opinion, however, if their use results in a new generation of professional pilots who start their careers with a healthy and balanced approach to automation.