The Perils of Multitasking Pilots Overestimate Their Abilities, As Well As the Benefits of Doing Several Things at Once
humanfactors
BY LOUKIA D. LOUKOPOULOS, R. KEY DISMUKES AND IMMANUEL BARSHI
As we started the taxi, I called for the taxi checklist but became confused about the route and queried the first officer to help me clear up the discrepancy. We discussed the route and continued the taxi. ... We were cleared for takeoff from Run- way 1, but the flight attendant call chime wasn’t working. I had called for the ‘Before Takeoff’ checklist, but this was inter- rupted by the communications glitch. After affirming the flight attendants were ready, we verbally confirmed the ‘Before Takeoff’ checklist. On takeoff, rotation and liftoff were sluggish. At 100–150 ft, as I continued to rotate, we got the stick shaker. The first officer noticed the no-flap condition and placed the flaps to 5. … We wrote up the takeoff configuration warning horn but found the circuit breaker popped at the gate.1
s this an example of recklessness? from Birmingham (England) Interna- [stabilizer trim set for takeoff] to go Complacency? Absent-mindedness? tional Airport at 155 kt after finding it unchecked.”2 Complex operating conditions? Com- impossible to rotate the aircraft. The Are these incidents exceptions to plicated operating procedures? Insuf- investigation revealed that “a number usual practice or symptoms of wider- ficientI crew experience? Or something of distractions, combined with unusual spread vulnerability? What do they say as subtle as multitasking? demands imposed by the poor weather, about the progress of an industry that During another flight, in Febru- led to a breakdown of normal proce- has suffered at least three catastrophic ary 2009, a crew rejected their takeoff dures and also allowed a missed action accidents when a takeoff configuration
The Perils of Multitasking Pilots overestimate their abilities, as well as the benefits of doing several things at once. © Chris Photogtaphy Sorensen
18 | flight safety foundation | AeroSafetyWorld | August 2009 HUMANFactors
warning system failed to alert the crew (FAA) to characterize the nature and pieces of information from external that they were attempting to take off demands of routine flight operations. sources. In the ideal world, everything without having set the flaps?3,4,5 Preliminary findings13 raised red flags occurs at specific, predictable moments In reviewing categories of accidents for an industry that, like many others, as the taxi phase of flight unfolds. for 2008 — spurred in part by the fatal had unsuspectingly accepted multitask- This is the way activities are laid Aug. 20 crash of a Spanair McDonnell ing as a normal state of affairs. out in the manuals, the way cockpit Douglas MD-82 during an attempted We argued that commercial and tasks are taught in training, and the way takeoff from Madrid, apparently with public pressures, organizational and pilots are expected to perform, on the improperly set flaps, according to pre- social demands, and the increase in line. The activity-tracing exercise can liminary reports6 — Flight Safety Foun- air traffic, mixed with a healthy dose be repeated for every phase of flight, dation decried the “unwelcome return of pilots’ overestimation of their own and in each case, the ideal perspec- of the no-flaps takeoff” and concluded abilities, were creating situations that tive portrays crew activities as linear, that “we are not making much progress were considered routine, although they or following a prescribed sequence; in reducing the risk of these [types of concealed appreciable risk. predictable; and under the moment-to- loss of control] high-fatality accidents” Our research at the Flight Cogni- moment control of the crew. (ASW, 02/09, p. 18). tion Laboratory at NASA’s Ames Re- The real world is not as straight- A quick search of the U.S. National search Center in California is based on forward. Observation of flight crews, Aeronautics and Space Administra- a combination of methodologies that from the vantage point of the cockpit tion (NASA) Aviation Safety Reporting through the years have included labora- jump seat, helps us understand the System (ASRS) database reveals more tory experiments, structured interviews full ramifications of that. During our than 50 reports of attempted no-flaps and surveys, in-depth analyses of flight observations, we recorded every event takeoffs in the last decade, as well as manuals, participation and observation that caused some perturbation — or reports of incorrectly set trim, airspeed of ground and flight training, incident disruption — of the ideal sequence of and heading bugs; cockpit windows not and accident report analyses, and many activities of the two pilots. It did not latched; and other omissions. In many hours of cockpit jump seat observations take long to realize that the real opera- of these events, the crew was saved by during passenger-carrying operations. tional world is more complex and more the proverbial bell — a takeoff configu- Taking advantage of these sources of dynamic than represented in writing ration warning horn. That bell cannot data, we systematically analyzed and and in training. be relied on to always work, however. contrasted cockpit operations in theory Let’s look at the taxi phase of flight What leaves expert, conscientious and in reality.14 in more detail, as it often unfolds in the pilots — and their passengers — hang- Take any carrier’s flight operations real operating environment. The base ing by the thread of a last line of defense, manual (FOM) and draw out the flow layer (grayed-out and in the background such as a warning horn or a checklist? of activities required of each pilot from of Figure 2, p. 21) is the ideal representa- Articles abound in the daily news moment to moment while the aircraft tion depicted in Figure 1. Another layer about a multitasking society and the is taxied from the gate to the run- has been added, formed by some of the dangers inherent in our natural drive way for takeoff, and you will see the many disruptions that were observed to have more than one thing going on theoretical, “ideal” taxi phase of flight from the jump seat during routine at once.7,8,9,10 Most people know they (Figure 1, p. 20). flights. Ovals contain some of the pos- should not talk on their cell phone while The crew’s activities can be traced sible, additional demands that are not driving, although many do it anyway.11,12 from the moment the captain requests explicitly expressed in the FOMs. But what does multitasking have to do that the first officer obtain taxi clear- The disruptions listed in each oval with pilots on an airline flight deck? ance until the aircraft is lined up with carried additional task demands for the runway centerline, ready for takeoff. attention and action. Ice or snow on Complex Operations There are a number of procedures the ground meant that the captain In 2000, we embarked on a research that pilots conduct individually, two deferred calling for flaps prior to taxi to project sponsored by NASA and the checklists conducted by pilots together, avoid contaminating the wing surfaces U.S. Federal Aviation Administration monitoring requirements and other with slush, continued with other taxi www.flightsafety.org | AeroSafetyWorld | August 2009 | 19 humanfactors
activities, performed the taxi checklist is considerably more work than the Limitations calling for verification of the flaps set- ideal perspective suggests. But it is People often feel they are perfectly ting, and remembered to set the flaps not just a question of workload quan- capable of performing several tasks right before takeoff. Encountering a tity. It is also a question of workload simultaneously. There seems to be a busy frequency meant that the first of- management. Responding to the mul- popular myth that humans are good ficer had to continue monitoring all ra- tiple, concurrent demands of flight multitaskers. In reality, however, human dio calls in order to “jump in” when the operations requires interweaving ability to process more than one stream frequency became available, all while new activities with old ones, defer- of information at a time and respond monitoring the captain, maintaining ring or suspending some tasks while accordingly is severely limited. Truly situational awareness and carrying out performing others, responding to un- simultaneous performance is possible other pre-taxi preparations. expected interruptions and delays and only when tasks are highly practiced Again, the exercise can be repeated unpredictable demands imposed by and rehearsed extensively together. for each phase of flight. The resulting external agents — all while monitor- Performance in this situation becomes “real” picture reveals activities that are ing everything that is going on. This largely automatic, making few demands much more fluid, convoluted and vari- is multitasking in a pilot’s world. on the brain’s limited capacities for able than in theory: Activities are dynamic ‘Ideal’ Taxi Phase of Flight1 and not so linear, are
unpredictable, and are Monitor not fully under the Monitor ATC (ground), Captain ATC (ground) company First o cer control of the pilots. Request taxi clearance Obtain clearance Taxi clearance Pilots are routinely (taxi) Taxi Procedure Taxi Procedure forced to deviate Recall Check xxxx xxxxx xxxx x from their linear, xxxx xxxxxx Flight controls Check Flaps (green light) xxxx xxxx xxxx well-practiced and xxxx xxxx xxxx Flaps (green light) xxx xxxxx xxxx xx xxxx xx xxxx xxxxxxxxx habitual execution of Cabin door Lock Monitor procedures. Neither xxxx xx xxxx xxxxxxxx captain taxiing Takeo brie ng Review the nature nor the Ask for checklist Begin checklist Taxi Checklist timing of tasks and xxxx xxxxx xxxx x events can be antici- Flight controls Checked xxxx xxxxx pated with certainty. Flaps Set, green light Takeo brie ng Completed Essential information xxxx xx xxxx xxxxxxxxxx (at or near runway) Checklist complete and/or the individuals Before Takeo Procedure Before Takeo Procedure Engine start switches CONT xxx xxx xxxxxxx xx xx x required to perform Takeo clearance Landing lights and As FMC position update As desired strobe light switches desired some activities are not Transponder On xxx xxxxxx xxxx xx always available when Ask for checklist Begin checklist Before Takeo Checklist expected. Tasks often xxxxx xxx xxx xx Packs Set must be initiated Transponder TA/RA earlier or later than Master caution Checked xxx xxxxxx xxxx xx Checklist complete planned. Pilots must Line up with runway continually find ways to fit more than the “ideal” activities into ATC = air traffic control; CONT = continuous; FMC = flight management computer; TA/RA = traffic advisory/resolution advisory the allotted time. 1. The ‘ideal’ phase represents that described in a flight operations manual. Source: Loukopoulos, L.D.; Dismukes, R.K.; Barshi, I. The Multitasking Myth: Handling Complexity in Real-World Operations. Burlington, Vermont, U.S.: Ashgate One implica- Publishing Co. 2009. tion of the real picture is that there Figure 1
20 | flight safety foundation | AeroSafetyWorld | August 2009 HUMANFactors
attention and working memory. But human information processing attri- ad hoc, inventive ways. Too many of these when an individual tries multitasking butes, and the experience, skill and goals seemingly benign situations bore a strik- in a situation that involves novel tasks, of the individual — always within the ing resemblance to stories recounted by complex decision making, monitoring, context of prevailing standard operating pilots in incident reports or that we read or overriding habits, it all falls apart. procedures and operational restrictions. about in accident reports.15 In principle, pilots, like all people, However, the approach people take to For example, the crew cited in the have limited choices when called to multitasking demands is not necessarily first paragraph of this article received a multitask: They can interweave steps deliberate or well thought out. stick shaker warning after rotation and of one task with steps of other tasks, During our observations, we spent realized they had inadvertently omitted or defer one task until the other task is many hours watching pilots handle rou- setting the flaps to the takeoff posi- completed, or even purposefully omit tine multitasking situations, apparently tion. This crew had been multitasking, one task. The choice and the degree to without much effort or many errors — attempting to concurrently address a which any of these proves successful but we became increasingly uneasy with discrepancy in their route and an inop- depend on the interaction of the char- the risks they were unknowingly accept- erative call chime. acteristics of the tasks being performed, ing each time they were called to react in The crew in the Birmingham event rejected their takeoff, after finding it impos- ‘Real’1 Taxi Phase of Flight sible to rotate the air-
Monitor craft, because they had Monitor ATC (ground),Busy frequency Ask first officerATC (ground) to set flaps company inadvertently omitted Captain Defer communication First o cer Request taxi clearance Obtain clearance setting the stabilizer Taxi clearance Monitor frequency and (taxi) Verify ramp contact when available clear area trim for takeoff. This Taxi Procedure Taxi Procedure Acknowledge clearance Recall Check Ice/snow xxxx xxxxx xxxx x crew was also multi- xxxx Deferxxxxxx setting flaps Consider need for de-icing Flight Confirmcontrols Checkcaptain’s understanding tasking: They had to FlapsRemember flaps(-, before green light) takeoff De-icing checklist xxxx xxxx xxxx System configuration? (APU, packs) Unfamiliar with xxxx xxxx xxxx Flaps (green light) deice the aircraft, were Monitor airport/taxi route xxx xxxxx xxxx xx Remember/identify xxxx xx xxxx xxxxxxxxx airport traffic Review charts Monitor preoccupied by the Cabin door Lock turns/intersections Monitor xxxx xx xxxx xxxxxxxx hold-short instruction taxi progress captain taxiing visual “clear” turns weather conditions, Takeo brie ng Review aircraft to follow Ask for checklist Delay (traffic) Begin checklist were trying to meet a Consider shuttingTaxi down Checklist one engine New/additional First officer busy Restart enginexxxx before takeoffxxxxx xxxx x taxi instruciton takeoff time constraint, Defer checklist Flight controls Checked After start checklist Acknowledge instructions and were focused on Perform checklist as soon as possible xxxx xxxxx Interruption Verify understanding Flaps Set, green light remembering (which RespondTakeo to interruption brie ng Completed Verify captain’s understanding Resumexxxx checklist xx xxxxSwitch xxxxxxxxxx to tower frequency they did) to set the (at or nearRunway runway) change Checklist complete Monitor flaps, which they BeforeConsider Takeo feasibilityProcedure tower frequency Before Takeo Procedure Engine start switches CONT xxx xxx xxxxxxx xx xx x Consult charts/performance data Takeo clearance Just-in or revised load data had deferred earlier Landing lights and As FMC position update As desired strobe lightBrief switches new runwaydesired Calculate new performance data Transponder On because of the slushy xxx xxxxxx xxxx xx Reprogram new data Have captain verify new entries Ask for checklist Begin checklist conditions. Before Takeo Checklist Continue to monitor captain xxxxx xxx xxx xx Inform company The Madrid ac- Packs Set Transponder TA/RA cident apparently re- MasterVisually caution “clear”Checked runway sulted from the crew’s xxx xxxxxx xxxx xx First officer’sChecklist turn to complete fly Line up with runway inadvertent omission Take control of aircraft while finishing checklist of setting the flaps for takeoff, coupled APU = auxiliary power unit with the failure of the 1. The ‘real’ phase represents the findings of jump seat observers. takeoff configura- Source: Loukopoulos, L.D.; Dismukes, R.K.; Barshi, I. The Multitasking Myth: Handling Complexity in Real-World Operations. Burlington, Vermont, U.S.: Ashgate Publishing Co. 2009. tion warning system. Was this crew also Figure 2 multitasking? There www.flightsafety.org | AeroSafetyWorld | August 2009 | 21 humanfactors
are indications that the crew was distracted by often so busy that it is extremely difficult for pilots an overheating probe, and had to return to the to pause long enough to review whether they have gate for maintenance, receive additional fuel, completed deferred or interrupted tasks. and start the engines anew. Pilots also are highly vulnerable to errors of Our research has focused on key aspects of omission when they must attempt to interweave human cognition that lie at the heart of multitask- two or more tasks — performing a few steps of a ing, namely remembering to perform tasks that task such as flight management system (FMS) data must be deferred (prospective memory), auto- entry, switching attention to another task such as matic processing and switching attention between monitoring taxi progress, back and forth. Much tasks. There is considerable scientific evidence of the time pilots can interweave tasks without that pilots, like all people, are highly vulnerable to problems, but if one task becomes demanding— inadvertent but potentially deadly omissions when the FMS does not accept the input, for example a situation leads them to defer a task that nor- — their attention is absorbed by these demands, mally is performed at a particular time and place. and they forget to switch attention to other tasks. Deferring a task breaks up the normal sequence of Monitoring, a crucial defense against threats and habitual actions and removes environmental cues errors, often falls by the wayside when pilots must that help pilots remember what to do next. interweave it with demanding tasks. In fact, moni- Interruptions create especially dangerous toring is far more difficult to maintain consistently prospective memory situations — by requiring than most pilots realize, as evidenced by studies of pilots to remember to resume the deferred, inter- automation monitoring.16,17 rupted task — but are so commonplace that pilots may not recognize the threat. Interruptions typi- Dispelling the Myth cally disrupt the chain of procedure execution so There is no single best technique to manage the A crane lifts the abruptly that pilots turn immediately to the source challenges posed by multitasking in flight opera- wreckage of a of interruption without noting the point where tions, but we have suggested various things that Spanair McDonnell the procedure was suspended, without forming an pilots and organizations can do.18 First, we must Douglas MD-82 that explicit intention to resume the suspended pro- dispel the myth that multitasking comes easily to crashed in August cedure, or without creating salient cues to remind humans, especially to pilots with “the right stuff.” 2008 during takeoff themselves to resume the interrupted task. Certain We must help pilots recognize typical multitask- from Madrid. phases of flight such as taxi-out and approach are ing situations that create vulnerability to error even in the most routine aspects of operations. Organizations must take a close look at the dif- ference between the ideal perspective and the real nature of actual flight operations and adjust pro- cedures, training and expectations accordingly. Fortunately, both individual pilots and or- ganizations can reduce the peril of multitasking. Pilots can treat interruptions, suspending tasks, deferring tasks or performing tasks out of normal sequence as red flags. When interrupted, they can reduce vulnerability by pausing momentarily to mentally note the point at which the procedure is interrupted and by reminding themselves to return to that place later, before addressing the interrup- tion. When suspending or deferring tasks, they can identify when and where they intend to perform © Associated Press the task; create salient reminder cues, such as
22 | flight safety foundation | AeroSafetyWorld | August 2009 HUMANFactors
putting an empty coffee cup over Immanuel Barshi is a research psychologist 9. Wallis, C. (2006, March 19). “The the throttles when they have deferred at the Human-System Integration Division at Multitasking Generation.” Time, March setting the flaps to their takeoff position; NASA Ames Research Center. 19, 2006.