key dismukes, phd chief scientist for aerospace human factors nasa ames research center 12 january...
TRANSCRIPT
Key Dismukes, PhD
Chief Scientist for Aerospace Human Factors
NASA Ames Research Center
12 January 2002
Lessons from Aviation: Memory, Skilled Human Performance,
and All-too-human Error
Road Map of Talk
A cognitive perspective on error vulnerability of skilled professionals
Examples from one domain of error: prospective memory– Lapses in everyday life
– Lapses in flight operations
Implications for professional training
Most Airline Accidents Attributed to Crew Error
Society: error = blame– Misrepresents nature of cognitive skill
– Undercuts safety
Research on human factors in aviation safety– Extrapolate to other professional domains?
Aircrew Performance
Trained to high level of performance; daily practice; annual recurrent training
Monitored periodically by check pilots
Highly motivated
NTSB accident reports reveal few examples of incompetence or malfeasance
Two Fallacies about Error
Fallacy: Error can be eliminated if pilots are sufficiently vigilant, conscientious, and proficient.
Truth: Vigilant, conscientious experts routinely make errors.
Fallacy: If an accident crew made errors in tasks that pilots routinely handle without difficulty, that accident crew must have been in some way deficient — either they lacked skill, or had a bad attitude, or just did not try hard enough.
Truth: The fallacy ignores sampling bias. No matter how many times an expert performs a procedure perfectly, the probability of error is greater than zero.
A Cognitive Perspective
Simply trying hard will not prevent errors
Unique human capabilities enabled by biological information-processing mechanisms– Vulnerable to error
Error is probabilistic, not deterministic
Illustrate with research on memory errors
Prospective Memory (PM)
Remembering to perform an action that must be delayed
Relatively new field of human memory research
Defining characteristics:– Delay between forming intention and opportunity to execute
(seconds to years)
– Delay filled with other tasks that occupy attention
– No explicit prompt telling us it is time to execute intention
So how do we ever remember to perform intentions?
– A theoretical perspective
Attributes of the Model
Mem
ory
More accessib
le
Activatedrepresentations
Activation
• Very limited capacity• Currently attended representations• Dynamic flow of contents• Representations compete to enter attention based on level of activation
External Stimuli
Less acce ssib
l e
Long-term memory• very large capacity
Focal Attention
Attributes: Activation and Retrieval of Memory Representations
Mem
oryM
ore accessible A
ctivation
“horse”
External Stimuli
Less acce ssib
l e
Long-term memory
Focal Attention
cowboy
saddle
Indian
racing
animal
“horse”
horse
Currently attended representation
provides activation to associated item
in memory Activation increases as function of
time in attention Activation decays as function of time
since last attended Activation is finite and divided among
associates Activation is divided according to
strength of links to associate
More accessib
le
Attributes: Goals are Memory Representations
Mem
oryM
ore accessible
Activation
prepare vuegraph
External Stimuli
Less acce ssib
l e
Long-term memory
Focal Attention
preparetalk
scheduleconference
room
flossdaily
win Nobel prize
Deferred intentions are a form of goal Goals are represented as condition/action
associates (If…then) Goals are associated in hierarchies of sub-
goals Sub-goal in focal attention helps maintain
activation of higher goal
More accessib
le
Einstein-McDaniel PM Paradigm
Instruct subjects to perform cover task (e.g., reading a paragraph or rating pleasantness of series of words)
Give additional (PM) task (e.g., “Press slash key when see a name of an animal”)
Subjects begin performing cover task
– Delay between starting cover task and trial with animal name
– Must remember to perform PM task without prompting
Prospective Memory as Competing Concurrent Task
Mem
oryM
ore accessible A
ctivation
lion
External Stimuli
Less acce ssib
l e
Long-term memory
Focal Attention
cub
lion
lion
ratepleasantness
Africa
catParticipate
in experiment
animal
presskey Prospective task
competes with pleasantness rating for retrieval
On-going task has inherent advantage
Outcome is probabilistic
Depends on multiple factors
More accessib
le
Implications for Real-World Performance
Cannot maintain delayed intentions in focal attention
– Must retrieve from memory when opportunity for execution arises
– Retrieval requires noticing some cue associated with intention
– Availability of cues and noticing cues is haphazard
Thus, memory lapses are commonplace
Strategies
Importance of delayed intention does not prevent memory lapses
Simply trying harder is ineffective
Strategies may help to some degree:– Habitual review: what do I need to do now?
– Encode environmental cues likely to be present in window of opportunity for execution
– Create salient cues that must be processed during ongoing task
Cockpit operations are highly proceduralized
Taxi, climb out, descent, and arrival are sometimes quite busy
Each pilot is responsible for multiple concurrent tasks (e.g., searching for traffic and flying the airplane)
Interruptions, distractions, or preoccupation with one task to the detriment of another found in nearly half of NTSB accident reports
Issue may be management of attention rather than overload
Flight Crews Manage Multiple Tasks Concurrently
Lapses in monitoring (2/3 of reports)
– Aircraft position: taxi, altitude capture, navigation
– Aircraft status: systems, automation
– Actions of other crew member
Lapses in prospective memory (1/4 of reports)
– Complete interrupted procedure on checklist
– Retract speed brakes when leveling-off descent
– Reset flaps and bleed air after de-icing
– Stop fuel transfer
– Set take-off flaps when deferred
– Defer lowering landing gear in minimum fuel situations
Examples of Neglected Tasks
Aviation Incident Study
Competing Tasks that Distracted or Preoccupied Pilots
50% involved communication (e.g., conversation with other crewmembers/flight attendants, radio communication, ATIS, PAX announcements).
16% involved “head-down” tasks (e.g., paperwork, FMS entry, reviewing charts)
14% involved abnormals
8% involved searching for/responding to traffic
12% miscellaneous (e.g., decision-making, unstabilized approach)
Aviation Incident Study
Lapses in monitoring (2/3 of reports)
– Aircraft position: taxi, altitude capture, navigation
– Aircraft status: systems, automation
– Actions of other crew member
Lapses in prospective memory (1/4 of reports)
– Complete interrupted procedure on checklist
– Retract speed brakes when leveling-off descent
– Reset flaps and bleed air after de-icing
– Stop fuel transfer
– Set take-off flaps when deferred
– Defer lowering landing gear in minimum fuel situations
Examples of Neglected Tasks
Aviation Incident Study
Omitting a Procedural Step
Highly practiced procedures vulnerable to omission when:– Interrupted
– Performed outside normal sequence or context
Highly practiced procedures become largely automatic– Allows fast, smooth execution
– Requires minimal conscious supervision
– Execution largely under control of environmental cues
Most common error in maintenance: omitting a step in re-assembly– Example: finish tightening spark plugs
Why So Easy to Forget a Procedural Step?
With highly practiced procedures, retrieval of each step triggered by:– Current state of environment
– Execution of immediately preceding step
Interruption breaks chain of preceding steps -- no trigger
Environment may seem to indicate uncompleted step has been performed
No episodic memory trace for habitual actions
Omitting a Procedural Step
Prospective Memory Countermeasures
Recognize the threat– Vulnerable even when tasks are important and delays are short– Especially vulnerable: Interruptions and performing habitual procedures
out of normal sequence
Explicitly note interruption and identify when/where intention will be executed– In team situation, say aloud
Create salient reminder cues*
Avoid rushing procedures
Periodically review status and ask if anything is missing
Use checklists** Not always practical
Training Implications
Line Oriented Flight Training (LOFT)
Full-mission simulation
Normal operations and challenging situations (e.g., weather diversions, equipment failures)
Should include realistic:– Concurrent task demands
– Interruptions
– Distractions
– Delays
Training Implications
Error Management Training
Change in historic attitude
Errors are inevitable
Training should address:
– Recognizing potential threats
– Detecting errors
– Managing error outcome
Training Implications
Error Data from Routine Operations
What errors occur, circumstances, and how professionals respond
Accident/incident reports useful but are a limited and biased sample
LOSA: Line Operational Safety Audits– Large sample of daily operations
– Data on threats, errors, and how crews manage
– Comprehensive and realistic picture of challenges
LOSA + laboratory research powerful new approaches to training
Training Implications