Tuesday, June 15, 2010

Mapping the intersection of mind and computer

Well my inbox is filling up again with emails, as it did last month when I reported this story for The New York Times on pilot complacency and cockpit automation.

Prompting the latest flurry of comments is today’s article in the Wall Street Journal by Andy Pasztor and Daniel Michaels about the crash in May of Afriqiyah Airways. Only one of the 104 people on the Airbus A330 from Johannesburg to Tripoli survived the accident.
According to Pasztor and Michaels the landing accident is being seen as one in “which confused pilots got out of sync with the plane's computerized controls and ended up flying an apparently functioning commercial jet into the ground.”
This is no one-off event. A number of studies over the past 15 years indicate pilots fail to adequately monitor what the airplane is doing in one-half to three-quarters of all accidents. So in the wake of the Afriqiyah Airways disaster, what’s the big idea being proposed? More automation. That’s right, Airbus is said to be working to “devise foolproof automated ground-collision avoidance systems” that in cases of emergency transfer control from the pilots to the airplane.
“This is very disturbing”, wrote Hugh Schoelzel, a retired captain who worked as director of safety for TWA. “The more automation we add, the more training and pilot qualification issues arise. I believe in automation, but as an adjunct to basic pilot skills, not as an ‘end-all’.”
While automation may be causing a decrease in piloting skills as Mr. Schoelzel suggests, Missy Cummings a professor at Massachusetts Institute of Technology says there is another reason to be concerned about cockpit automation; boredom.
Dr. Cummings a former Navy pilot, is director of the humans and automation laboratory at MIT’s department of Aeronautics and Astronautics. Not surprisingly she is a proponent of automation and envisions a future that will include at least some pilotless commercial flights. But first some extremely troublesome problems have to be wrestled to the ground, problems demonstrated by one of Dr. Cummings students, Master’s degree candidate First Lt. Christin S. Hart, who has found that too much automation can prove counter-productive.
“Increased automation can lower an operator’s workload too much, leading to mental underload, which can cause a decrement in vigilance, or sustained alertness, and lead to boredom. It has been shown that boredom produces negative effects on morale, performance, and quality of work,” she wrote in her paper, Assessing the Impact of Low Workload in Supervisory Control of Networked Unmanned Vehicles.
These findings do not surprise Dr. Cummings “The human mind craves stimulation”, she explained to me last week during a visit to her office in Cambridge. Failing to find that stimulation in the task at hand, the mind will wander.
This cuts to the heart of a number of events outlined by industry researchers but takes us at warp speed to the episode last October in which two Northwest Airlines pilots overflew their destination - the Minneapolis airport. The Northwest pilots were doing personal work on their laptops which is not allowed.
“It doesn’t have anything to do with automation,” F.A.A. Administrator Randy Babbitt told me. “Any opportunity for distraction doesn’t have any business in the cockpit. Your focus should be on flying the airplane.”
But if I’m reading Lt. Hart’s study properly, the automation itself is an opportunity for distraction, even as it assists pilots by reducing workload and increasing the precision of calculations and navigation.
This is a conundrum. In today’s cockpit, two highly complex systems – the mind and the computer – come together, even though the contours of that intersection are still being mapped. It is not only unwise to race to a fix that fails appreciate these systems in balance, but it is unlikely to result in success.

6 comments:

Frank Van Haste said...

Dear Christine:

I, too, winced upon reading in the WSJ article on the Libya mishap that the Airbus prescription was more of the same.

It would seem that sometimes aboard modern aircraft there is too much automation ("What's it doing now?") and sometimes not enough. So, can the automation be made capable of adapting to changing circumstances?

In a post a few months back, I made reference to some work by Mark W. Scerbo at Old Dominion University on Adaptive Automation. In his abstract he says: "Adaptive automation refers to systems in which both the user and the system can initiate changes in the level of automation. ...[S]ystems have been developed that follow the neuroergonomics approach and use psychophysiological measures to trigger changes in the state of automation. Studies have shown that this approach can facilitate operator performance..."

It may be that there is no static configuration of the mind/computer interface that will solve our problems, but Adaptive Automation may offer a way forward.

Best regards,

Frank

Anonymous said...

Christine:

You are entirely correct to point out the two toxic effects of "automation in the cockpit," i.e., decrease in pilot skills and boredom.

In my flying experience I have see this phenomena first hand. Today's pilots are nothing more than "video game operators" but when faced with an emergency that is not handled by their computer or their automated systems they have no fucking idea what to do. I have flown with pilots and ask them to simulate a partial panel or a loss of electrical power. I cover their electrical instruments and ask them to fly with just needle-ball-airspeed as well as a magnetic compass. They go into panic mode. I saw it myself when I contrast my experience in the F-15 (a fully computerized plane) with the P-51 (a completely analog plane). I would much rather fly the P-51 because I was in control, knew what to do and it kept me on my a toes.

Hector Andres Negroni
Fighter Pilot & Modest Hero

Greg Feith, Former Senior NTSB ASI said...

The scenario of the A330 accident is identical to the crash involving a USAir DC-9 at Charlotte in 1994. It was not automation that caused the crash but rather the effects of Somotagravic illusion during the power-pull up, right turn that occurred simultaneously. The F/O was the flying pilot and executed the missed approach at the captain's command. The captain was performing other duties and sensed the airplane had been pitched up at higher than normal rate/attitude (even though it was only at about 15 degrees NU) and told the F/O "down, push it down." The F/O responded and manueuvered the airplane from 15NU to 5 degrees below the horizon even though they were only several hundred feet above the ground. The airplane struck trees and off the right side of the runway. This sequence is identical to the A330 pilot reaction to executing the missed approach.

"A spokesman for Airbus, a unit of European Aeronautic Defence & Space Co.,
said the company on Friday sent update notices about the investigation to all operators of A330s, but declined to give details.
A person familiar with the communication said it told airlines that investigators' analysis indicates the crashed plane, which was only nine months old, had no system malfunctions, still had sufficient fuel going to its engines, and didn't suffer a fire before impact.
When the jet was at roughly 1,000 feet in altitude and about a minute from touching down, according to people familiar with the details, the pilots reacted
to some type of ground-proximity warning. The co-pilot, who was flying, increased power quickly and pulled up the nose of the plane, typically the
correct steps to initiate a so-called go-around. But since there weren't many passengers aboard and the jet had relatively little fuel, it accelerated rapidly. The captain, according to people familiar with the sequence of events, may
have been looking at charts or was distracted by something else.


At that point, the co-pilot apparently believed the Airbus A330 was climbing at
a dangerously steep angle and pushed the nose down, quickly losing control of the plane. The captain subsequently tried to yank the nose up again when he heard more collision warnings, but the big plane was too close to the ground
for such maneuvers, these people said. That scenario is consistent with the widely scattered, small pieces of wreckage found at the site."

I have investigated several other events whereby the pilots executing a missed approach or VFR go-around have over controlled the airplane (not because of automation) but because the the procedure was conducted as trained "power-up, pull up." The typical training program, regardless of whether it is the airlines or GA, train pilots to shoot the approach (usually down to minumums) and then perform the missed approach which employs the use of full climb power and a normal pitch attitude for the the climbout. Unfortunately with the A330 event and others I have investigated, we do not train pilots to employ techniques (such as lesser pitch attitudes and the use of modulated power rather than full power) to execute the miss or go-around when it is conducted higher on the approach where the need to climb to a specific altitude is substantially less. A light airplane and full power, utilizing the normal power up, pull procedure will cause the airplane to zoom-climb and if the pilot only needs to climb 500 feet after abandoning the approach the airplane is going to climb through the desired altitude before the pilot can react and then he/she will get into an "aircraft/pilot coupling" issue that results in the pilot being out of sync with the airplane in inputting the necessary corrective flight control actions.

This is not necessarily an automation issue as it is a more diverse training issues instead of teaching a rote program and hoping pilots react robitically.

Greg Feith
Former NTSB Senior ASI

James said...

This discussion has touched the epicenter of aviation safety. As other esteemed contributors have opined, it is the interface of humanity and technology. The same human brain that is capable of assessing all the variables necessary to put a powerless airliner in the Hudson is also capable of of flying a perfectly good airplane into the ground.

We, as an aviation community, will not be able to say we have effectively transitioned to automated flight decks until we have the level of engagement with the automated airplanes that we had with manual ones. Monitoring the aircraft in both horizontal and vertical axis is a perfect example. Prior to the introduction of sophisticated Flight Management Systems crews were very focused on their desired altitude because the responsibility to capture and or maintain that altitude belonged solely to the pilot. With FMS type airplanes I have observed that a large number of crews delegate not just the task of altitude management, but the responsibility as well.

Todays automated flight decks are a much more cerebral environment. They require the pilots brain much more than their hands. On the manual flight deck, there are tactile feedback mechanism that keep the pilot engaged. But that feedback must be balanced with the flight instruments as the USAir accident demonstrated. The feedback on the automated flight deck is very subtle and must be verified to evaluate the effect of the inputs to the automation.

Mr. Feith is correct when he states, "This is not necessarily an automation issue as it is a more diverse training issues instead of teaching a rote program and hoping pilots react robitically." The relationship between procedures and decision making skills cannot be over-emphsized. Procedures are necessary for standardization and describing best practices, however, they should enhance and support good decision making skills, not replace them. Too often the industry reacts to incidents and accidents with another procedure or piece of automation. Those safety systems are valuable, but more so is the opportunity for learning and training. On the day and time of most airline accidents other flight were successfully managing the conditions and environment. It is always very beneficial to focus on what allowed that crew to end up with an undesired outcome. I believe that it is rarely the lack of a procedure or device, but rather a breakdown of basic pilot or decision making skill.

Thank you for the opportunity to contribute to this very important discussion.

James L. Blaszczak
777 Captain, Continental Airlines

Anonymous said...

These comments are some of the most balanced and insightful I have read in response to an article on a blog. Thank you. I learned a great deal not only from the article, but also, from those that provided their insights as comments.

Dave Moreau
ERAU Graduate Student

Kaitain said...

I find it strange that two and a half years on from the discoveries of the flight recorder and voice recorder, almost nobody has discussed the issue of the lack of sleep in this incident. Specifically, Captain DuBois had only had one hour's sleep the night before, and Robert reported having had unsatisfactory sleep on his rest phase of the shift cycle. That leaves only Bonin having (potentially) had enough sleep, and he's the least experienced, and spooked by the conditions. Could CRM have failed in part because the two senior officers were both functioning some way below their optimal level? Their responses seem indecisive and bizarre throughout. Sleep deprivation can give you reflexes and judgement so poor it is almost like being drunk.