Jan 19, 1998 - a safety point of view. The technology changes on the flight deck also have ... to address if we are to maintain and improve aviation safety in a changing ...... S. Card,. J. Hochberg, and. B. Huey, editors,. Human. Performance.
COINITIV[
Learning
STSTEilS
IN|INIIIIMi
LAIOIATOIY
from Automation Surprises and "Going Sour" Accidents: Progress on Human-Centered Automation
/?i,.I_ L///,j David Cognitive
Engineering
Institute The
/ 7 c/__-
D. Woods
Systems
for
Ohio
Laboratory
Ergonomics
State
Nadine
_t- c /6
University
B. Sarter
Aviation Research Laboratory Institute of Aviation University
of Illinois
Final NASA Cognitive
Ames
Engineering
at Urbana-Champaign
Report Research
C/enter
in Aerosp/lce NCC 2-592 V
January
Applications
19, 1998
S'
/
1
Advances
in
technology
transports
has
an economic deck
and
also
system
had
have
and
many
had
different
new
sometimes
What
are these
from
them? Do
they
Or
Are
the
with
or near
is
other
accidents
arisen
or human
error?
third operators
and
aviation
shown 1).
should
we
-
the
flight
(Figure
what
possibility
the
experience,
have
as well
both
of the
Operational
jet
from on
aspects
and
of small
they
that
learn
represent
automation?
independent
small
a record
provide if we
many
effects
changes
automation,
a
a few
is otherwise
to address
positive
commercial
glitches
revealed
by
misses?
represent
problems
on
technology
have
cockpit
a series
been
occasionally
between
just
glitches
these
and
there
automation
performance.
problems
breakdowns
changing
areas
where
there
of outstanding
us with
evidence
are to maintain
and
are cracks
designs about
improve
and
to be
systems?
deeper
factors
aviation
safety
that in a
world? do
provide
the
reverberations
insight
technologies
into
and a
series
generic
of
(Sarter
by surveys,
operational
et al., 1995;
of
systems
automation
Degani
of human
on
over-automation
in what
we need
on
effects
problems
patched
Based
reverberating
perhaps
these
How
The
incidents,
accidents
Or do
of view.
represent
specific Do
point
problems
coordination
of have
surprising
instead
levels There
aspects
investigations,
new
effects.
a safety
research and
and
Introduction
and
technology
issues
about
(Winograd
and
investigations Woods,
1992;
experience
Eldredge
et al.,
change
1994; and
1991;
2
developing
Woods,
of
pilot 1995;
on
the
flight
deck
human-centered
1997)? interaction
1997a,
incident
data
Tenney
et al.,
1997
with
cockpit
b), supplemented
from
other
1995;
Wiener,
studies 1989),
(e.g., we
have
found
that
automation are
are
more
symptoms
that
human-machine the
Many
synthesized
new
al.,
1993;
of
the
results
summarizes
that
ability
interface
al.,
1996).
can
help
back
safety
a broad
and
concerning
In addition, studies
care
this
we
of
(e.g.,
Cook
kind
and
of
volumes,
find
physician
medicine
1996;
comprehensive
we were
range
This
the
research
for
and
safety
in other the
crews
crew-automation
groups.
future
roles
that
with
assess
flight
organizations,
changing
of
Moll
Woods,
research
Billings
player,
to
and
are (1996)
participation
able
in
improve
research team
safety.
results that
was
examined
the
implications
of observed
for
investments
to
in
effort
helped
us step
the
personnel.
aviation
managers, back
to maintain new
deck
problem,
to discuss
training
of operational
a FAA
investment
error"
and
this;
flight
pilots,
where
new
"human
of the
research,
modern
of stakeholders
areas
our
It uses
to
maintain
coordination
investments
point
implications
and
from
investigations.
perceived
line
related
emerged
arise the
the
has
accident
problems
of a team
project,
with
industry
implications
and
by our
between
difficulties
carrier
in critical
with
difficulties
phenomena
from
Woods,
pattern
us deal
more
In this
with
These
disaster.
and
two
why
tremendously
the
results
reports,
of
to step
facilitated
behind
and
the
incident
automation
The
towards
systems
in
and
paths
interaction
glitches.
patterns
Obradovich
discussed
understanding
make
patterns
crew
of individual
deeper and
et
research,
strategies
a series
surround
et al. (1994).
paper
related
that
computer-based
and
Woods
This
of
with
Charante
and
than indicate
kinds
interaction
1996).
problems
cooperation
same
van
the
and
systems
(Abbott
domain,
improve including
manufacturers, assess
and investments
the
enhance in
et
implications aviation automation
and of safety are
figure 1
Insert
Figure
1.
Human
The
Reverberations
here
of Technology
Change
on
the
Flightdeck
for
Performance.
2 Impact
One
about
way
of Technology
to recognize
is to listen
to the
interacted
with
•
directly
•
through
the
voices
many
Change
pattern we
in conversations their
that
heard
different
on Cognition
underlies
in our
judgments
the as
Collaboration
automation
and
investigations.
operational about
and
people
impact
human
In these
and
error
studies
we
organizations,
of automation,
expressed
in
surveys
about
in incidents
that
occurred
cockpit
automation, through
their
through
reported
their
behavior
performance
coordination
between
in
crew
simulator
and
studies
automated
on
that
systems
the
line,
examined
in
the
specific
flight
contexts. We
will
summarize
adopting
the
research
results
actual
point
Coordination
Pilots of
and many
automation
issues
made
2.1 Automation
results
of view
and
statements
the
of
the
of different in their
multiple stakeholders
words.
to us in different
converging
The
and
statements
studies
by
expressing
are
paraphrases
by the of
contexts.
Surprises: Breakdowns
instructors modern that
described cockpit
were
Between
strong
Crews
and
revealed
systems. but
and
sometimes
They
Automation
the
clumsiness
described silent
and
and aspects
difficult
complexity of
to direct
cockpit when
time
is short.
the
tools
The
We
that
users'
are
saw
and
supposed
perspective
expressed
by the
Wiener,
on
is it doing
•
What
will
•
How
•
next?
did
I get
into
Why
did
it do
this?
•
Stop
interrupting
•
I know
•
How
•
Unless
I stop you
computer
and
new
provide
generation
pose
in
challenges
imposed
"added
functionality."
of automated
describing
by
incidents
systems
is best
(extended
from
this
stare
user
Questions Woods
and (or
automated
taken)
by
the
to do.
The
sources,
for example,
changes
where on
its
environmental The
gap
the
a gap
are
system
p. 68).
these
point
they
inputs
such
pilot
being
the
doing, can
its
surprised
Sarter later
actions begin and
what
what
the
they
are
from
several
or indirect
mode
status
internal
1995;
of
arise
changes
Woods,
by
automation
and
errors
its
(Sarter,
surprises
as mode
inputs,
and
surprised
understanding are
program
_ul'prises
are
a mismatch
autonomously
crew
automation
between
what
human-
of it as a quirky,
Automation
user's
of
computer
think
crews
the
such
observer "A
must
system.
to do,
(Sarter the
to
where
between
for
as
user
1995,
of
conditions in
the
misassessments
erroneous
interpretation
results
that
trigger
one
agent
autoflight
set up
initial
in. why
term
i.e., situations
and to
systems
the
I want.
this?
illustrate
like
1997),
doing
can creep
(Lanir,
statements
lead
from
is so obscure ..."
busy. it to do what
statements
miscommunication which
to get
defined
Billings, not
I am
machine
and
person
and
while
at it, changes
interface
but powerful,
mode?
way
interaction
whose
a).
this
is some
questions
going
current
they
me
there do
These
users
them
face
now?
do
based
the
pilots
1989):
What
with
how
to serve
questions
•
taken
heard
and
behavior
logic
and
and
Woods,
when
the
sensed 1997
aircraft's
behavior does not match the crew's expectations. This is where questions like,
"Why
won't
It seems
that
displays
it do what
the
of data
on
behavior.
The
an
behavior,
they
is that
the
leave
a
The
greatest 1.
directions 2.
generally
strongly
in users'
act
mental
situations,
and
3.
feedback
about
relative
to the
Automation
state
crews
"funnel"
of evidence
observe
crews
simulations, linked
to the
and
find
design
sometimes
undesirable
aircraft
between
expected from
when of
the
and the
actual
situation.
aircraft
behaves
point
without
descent
altitude
without
actual
aircraft
behavior,
leveling
off.
it may
an
undesired
result
where
the
misunderstanding
not
occurs.
1996).
for
automation
surprises
is
direct
of how
without
their
activities
immediately
preceding
and
machine
future
partners
behavior
of
work
the
in
agent
world. one
automated about
own
partner,
the
are
interacting we
The in most
potential
their
models
of the
surprises
between
on
human
different weak
systems.
converge:
systems their
from
is reached,
top
before
arise.
misassessment
a target on
the
mode?"
of surprise
(cf., Billings,
that
this
automated
occurs
accidents
disaster
their
gap
the
through
into
to or recover
past
been
factors
from
gaps
detection
to avoid
automated
the
to respond
have
three
and
begin
I get
of the
point
detected
interval
shows
when
the
recovery
long
evidence
thus
is based
there too
and
of a problem
Unfortunately, persisted
activities
or flying
sufficient
notice
or
manner-flying
detection
not
does
has
can
this
descent,
did
of unexpected
crew
unexpected
initiating If the
the
"How
state
observations
Once
problem
in
the
is detected,
based
aircraft
generally
about
misassessment cases
crew
I want?"
kind
of
breakdown
systems. these with evidence
of automation
kinds
Our of
cockpit
automation
to the
the
coordination
investigations
coordination
of a variety and
in
of training
revealed
breakdowns. in performance users
a If we
full
mission problems
receive.
The
problems observed are interface
design
error.
If we
breakdowns significant misses
crew
and
at operational
and
errors
human
there
are precursor
is a chain interface
away
occasionally,
from
these
circumstances,
same
where or
but,
in
also
have
these
most
of human
coordination
cases,
a small
with
no
number
of
breakdowns
to the
automated
these
and
between
sequence
systems
of events.
the
and
sometimes
performance
where
flight
negative
problems
sequence
of
crews
ways,
occur
events
but is
in
later
outcomes,
problems
with
forms
that
coordination
in predictable
bad
find
contributor
between
performance events
predictable
we
we
human-computer
where:
circumstances
redirected •
these
"classic"
to certain
are a significant
affect
of
occasionally,
where
of the
innocuous
lead
Unfortunately,
there
characteristics
result
experience
occur
or accidents
words,
the
that
look
automation
In other
•
characteristics
consequences.
near
•
sometimes
other
occur
in
the
contributors
context
present,
of
and
more
events
vulnerable
spiral
towards
disaster.
2.2
The
These
Going
for
accident Cook,
(originally
occurs
and
at
from
1991).
an But
misassessments
and
the
situation
situation
least
when
evolving through
crew
accident
results
McDonald,
triggers from.
the
on
of
In
or a set of circumstances
event
manages
between
kind
based
unproblematic,
recover
effect,
in coordination
a particular
Woods
event
This
Accident
breakdowns
potential
and
Sour
studying this
come
of
a serious
is managed
into
hazard.
risky
is,
in
commissions the
and
that
create "going
room
class
isolation
that
the
operating
general
in
miscommunications, into
-
together
viewed
series
automation
sequence
situation a
and
of
from
principle, and
incidents;
or even
an
to be minor hindsight. possible
to
omissions,
human-automation incident
sour"
accidents,
appear or
the
team accident.
In
Several recent accidents involving While
they
terrain,
are
some
terrain
of these
since
crew
is
classically
the
example
sour
incident
data
scenario
complexity
and
to be a side
concern
rather
see
that
new
increasing
the
the
aircraft
of
et al., 1997).
effect
of complexity.
technology,
for the
into
flight
into
the
flight
one
vivid
and
developed
is increasing
the
sour
flight
Research
when
going
signature.
and
overview
Sarter
way,
potential
as managed
a brief
human-centered
this
as controlled
handling
(for
terrain
seems
the
technology-driven
into
aviation
show
be described
are
automation
flight
surprises
in
better
systems
the
raise
to
may
automated
of managed
The going
referred
cases
supervising
automation
in
a
operational
signature
(Billings,
1996).
After-the-fact, who
going
have
1994).
sour
complete
Since
the
opportunities
look
knowledge
system
to
incidents
of the
is managed
break
the
reviewers
•
"How
could
they
have
missed
•
"How
could
they
have
misunderstood
•
"Why
didn't
In
instructions
fact,
one
whether as, "All
for
reviewers, of the
together
The
test
and
lesson
coordination
what
learned between
logic an
with
advantage
the data
was
the
of
affairs
of the
incident
chapter
lead
et
it is easy
et al., 1994, piece
al.,
to see
benefits
of
6),
of information?" to us?"
to Y, given
the
inputs,
system?"
is a going of hindsight,
available,
(Woods
The
critical
X would
to outsiders
disaster.
Y, it is so logical
that
whether
necessary
to see
internal
(Woods
X, it was
dreadful
in hindsight,
towards
to comment
understand
and
state
hazard,
progression
allow
past
actual
into
hindsight
they
mysterious
why
was
sour
scenario
make no
one
is to
ask
comments
such
able
it all
to put
it meant?"
from
recent
the
automation
accidonts and
involving the
flight
breakdowns crew
is:
in
the
•
the
going
•
there
sour
scenario
is a concern
residual
that
this
risk_
in aviation.
data
and
Only
future
risk.
Investments
training
is an important
events
will
in turning
crews
to better
circumstances
category
offs
this
automation
automated by
guarding
of accident
a significant
whether
cockpit
pay
kind
represents
reveal
manage
produce
general
category,
portion
is a growing
into
a team
resources
in
against
this
a
of the
part
of the
player wide
and
in
range
type
of
of
accident
scenario.
Luckily,
going
systems.
The
•
the
sour going
expertise
the
that
are
compensates
coordination
team
and
strategi_
a great
We
heard
expertise
tactics
cautionary
burn
you."
Some
for teamwork.
complex
of two
factors:
personnel
and of
allows
trigger factors
this or
kind
of
exceptional
are
in general
and
notes
how
ways tricks)
of the
9
training
(through
systems
contribute
policies, job
to restrict
in
difficult
automation
from
successfully Some
to "be careful,
recipes.
use
line
situations.
them
the
to
procedures,
done
for some
embodied
expertise
departments,
reminding
or in particularly
design
operators' that
to get the
to pilots
are ways
how
automation
about
workarounds
Complexity
about
of
automated
Many
the
deal
develop
of some
in
very
progression;
Technology-Induced
features
simply
deficiencies
and
the
are
words,
and
a collection
for
individual
of automation
in
because
systems
incident
heard
of these
suite
blocked
we
the
strategies
even
together.
despite
it can
is usually
when
individuals
clumsiness
rare
human
only
breakdowns.
organizations,
the
erode
Expertise
investigations
usually
can
relatively
operational
or stop
come
3 Human
In our
in
significant
circumstances
are
progression
to avoid
problems
scenario
sour embodied
practitioners •
accidents
Some
of portions
situations. a Human
are of the
In other Factors
point
of view
and
produce
adaptation
Overall, the
technology
and
transition
training
training).
But
•
1994,
people
this.
strategies
Crew
skills
and
"They're
chapter
their
training
are
done,
expertise
primary
glass
limits
When
their
developing
and
amount
we
on
of resources
aircraft
the
to manage
for
as a set
limit
receive.
are tools
cockpit
that
behavior
there
systems
to a new
can
their
but
of the
constraints
pilots
tailor
automated
move
of human
5).
job
is one
many
because
organizations
to get
as pilots
there
we
consequences
for managing
experiences
managers,
et al.,
as a resource
to do
training
bad
(Woods
operational
ability
so few
recurrent
and
talked
(e.g.,
range
to
training
for
training
of
heard:
building
a system
that
takes
more
time
than
we
have
people." •
"There
is more
system
in different
•
"The
most
•
"We
need
•
"Well,
•
"We've
•
"We
are
•
"We
teach
the Economic
is new
those
forced
to rely
on
them
knowledge
the
how
to work
the
shrink
footprint).
When
people
to take
lead
in less seem
time)
show
that
requirements
on the
and
how
to use
it."
more
than
anyone
a policy." much
of] basic
modes
produce the
great
in training,
the
likes." they
learn
10
and
system
types as their
match
(the training in
expertise.
levels
in the
(better
the a
in training,
productivity
investments
human
of new
for people
in
reduce
or
improvements
in quality
developing impact
are
greater
to
footprint
benefit
than that
pressure
training there
rather
to believe
expenditures
it off."
line."
factors
(e.g.,
People
consistently
on
to click
it works
training
number
competitive
training
lower
especially
or capabilities."
with
recipe
[a certain system
forces
but
is when how
features
problem
time).
promise
to explore
that
of proficiency
same
data
use
investment
same
level
don't
to learn
handled
competitor's these
chances
and
training
thing
more
of the
it works,
situations."
important
we
rest
to know--how
same in
the
automation
However,
the
of automation role
changes
to more
of a manager
of enhanced human
3.1
safety
anomaly
requires
handler
that
we
(e.g.,
expand,
Sarter
not
et al.,
shrink,
1997).
our
The
goal
investment
in
expertise.
Complexity
The
second
reason
signature
is
automation
are
•
the
significant
(training
together.
only
and
a few
accidents
with
the
in
coordination
between
when
a collection
of factors
going
sour
human or
and
exceptional
For example,
is eroded
coordination flight
see
only
performance
factors crew
we
breakdowns
come
human
•
why
that
circumstances •
and
practice
due
to
local
factors
(fatigue)
or
systemic
with
training
investments),
is weak,
circumstances
are unusual
and
not
well
matched
experiences, •
transfer
•
small,
of control seemingly
between recoverable
Because
there
are always
because
these
incidents
identify
a host
behavior Focusing
on
any
manageable bulletin
to
policy,
add
While
these
lessons
of
of these
changes
-- just
shift
remind
crews
of
some
remedial
changes this
mismanage
may
incident
a minor
it is a symptom contributors
over
where
to
occur,
and
in
the
display
or bumpy, interact
a series
can
slightly,
X works
of
in
and
any
to very
trouble. local
a checklist,
circumstance
to
or machine
from
modify
up.
it is easy
team,
lead
add
incident
stages,
away
isolation
and
sour
in human,
sequence
points
how
is late
to a going
change
the
and
issue
Y,
reinforce
miss
the
a a
training.
be constructive
occurrence
successful
time
a small
signature.
of overall
actions
contributors
re-directed
one
automation
erroneous
evolve
have
and
multiple
of places
could
crew
system flight
in small
When
people
or non-routine complexity. deck
11
ways, and
they
automation
situation
into
It is a symptom performance
-
larger
seem
to
larger
trouble,
that
all of the
design,
training,
operational
policies
and
procedures,
certification
-
need
to problems
in human-automation
to
be
better
coordinated.
3.2
The
An
underlying
is the
Escalation
Principle
contributor
escalation
relationship higher
principle
where the
activities
the
tempo
of
required
demands
(Woods
greater
with
all
tend
beyond
of
normality
using intelligent times demands, the
per
these
se.
with
of his This
the
is the
attention
(situations
situations), are
new away
and
human-machine
an
associated
of clumsy
and with
autonomous
tasks, from
at or-
tempo
to be concentrated
afford
example,
information,
with
trap
For
unusualness
tend
or the
processing
of activities.
interacting
essential
process
information
burdens
least
or her
the
other
with
a fundamental
underlying
textbook
or
is
(including
beyond
can
the
control,
burdens
practitioner
There
or pace
members up
or
agent,
or diversions
interface
trouble
or
interface
the
in
greater
If workload
machine when
the
to go
situations.
a computer
trouble
team
communication)
of
1994).
attentional
among
margins
al.,
the
monitoring,
communication
criticality
the
operations,
to cope
for
et
coordination
the
at the new job
automation
or very
memory at hand
to
(Wiener,
1989)
4
Designer
Reactions
to Coordination
Erratic
Listen
to how
breakdown •
The "trouble
designers
in the
coordination
hardware/software free"
respond
Human
when
system
Behavior
they
between
are confronted
people "performed
aircraft).
12
Breakdowns:
and
with
evidence
of a
automation: as
designed"
(crashes
of
•
•
"Erratic"
behavior
behavior;
"brain
otherwise
skillful
The
too
(variations
burps," human
other
Those
just
phobic,
people
is,
system
people
computer
that
on
this
some
performance;
hardware/software
some
•
human
irrational
don't
understand
or too set
in their
are
"diabolic"
quasi-random
is "effective
or organizations
theme
degradations
human
in
in
behavior).
general
it" (e.g.,
human
and
those
logical
who
are
to too
us, old,
old ways).
or countries
"have
trouble
with
modern
technology." •
•
"We
only
them
out
pressure,
the
have
to go further
previous
Other
what
of it, but we
"I wanted the
•
provided
customer
for!"
(or
"we
tried
to
talk
to be customer-centered").
but ..." -
design,
asked
I was
supplier's
constrained
standard
by -
designs,
compatibility cost
with
control,
time
regulations. parts
capabilities
of
the
industry
of our systems
capabilities
and
recognize
(e.g.,
"haven't
kept
ATC
not
characteristics
what
is difficult
of
does the
to do with
up"
the
accommodate
newer
highly
with
aircraft
the or
automated
advanced advanced
ATC
aircraft
does under
not time
pressure). Some the
of these design
very
comments
world
complex
4.1
arrival
Escaping
Overall,
from
these
locked
into
a
failed.
going •
sour
human trouble
design
and
of
many
incidents
-"clear aircraft"
pressures
and
constraints
users,
economic
in
pressures,
procedures).
of Human
from
thinking
--either
Error
and
versus
developers that
this
reviewers
either
serious
multi-cultural
of comments
mindset
Too
free
for
and
departure
components
error
real
Attributions
kinds
independent box
(e.g.,
reflect
Over-Automation
show
technology
electronic stakeholders,
box
how and
failed
or
after-the-fact,
we
remain
people that
are
human attribute
to misuse
(La Burthe,
of
automation 1997)
13
or to
... contributed
to crashes
of
*
over-automation
-
"... statements
agains_automafion This
opl:n_ition
human
'per
this
by
demarcation
set up in order
Bootie,
p. 101).
The
1995,
primary
number
lesson
of industries
coordination cannot
be
thought to
expertise
or degrade
bottom
line
isolation Automation
and
(Hutchins,
the
ways
people
and
us smart"
research who
have
1996).
towards
constructive
to coordinate
the
human
Accident
analyses
contributor other
Woods,
industries
interpreted
that
70 or 75%
come
up
as a motivation
But some
view
these
problem,
and,
remedial
training,
The
lessons
progress
comes
machine
team
are
human 1993).
be considered Hutchins,
in
1995a).
a single
team
team's
coordination
of this
type
from -
that
(Norman,
system,
this
in
People
enhance
(e.g.,
as a joint
real
devices
can
it
in a large
1990).
cannot
adapt
(Kelly-
is
of scenario
developing
better
human-centered
design
1997).
suggest
to about
and
Norman,
technology
of
a breakdown
us dumb"
in
line
disasters
technological
Breakdowns
disaster.
arbitrary
represent
or "make
and
influence interaction
"an
and
artifacts
to coordinate
for
and
use
that
for malfunctions"
(e.g., the
is that
potential
(Winograd
technology
it, "make
people
blame
accidents
from
factors
as
of incidents
Technological
Billings,
path
the
them.
specialists
human-computer
interface
sour
separately
people
1995b;
an important and
the
term
analysis
going
of recent
from
on
to apportion
careful
about
assist
the
is that
between
supposed
The
from
of the
commentator
defining
Factors
1997).
misunderstanding
One
point
by ... Human
" (La Burthe,
is a profound
performance.
makes
se'
made
or dictate
of aviation
with
about
for
paying
statistics
as a result,
breakdowns
the
want
all pilot
human
mishaps. same
increased
superficially
they
in
as an
14
Similar
percentage. attention indication
to eliminate action
performance
through
the
human expanded
are
tabulations This
be
Factors.
of a human element,
in
should
to Human
a
error provide
procedures.
However, that
research
"human
learn
error"
about
people label
these
function, human
One
way
errors.
mode) are not
have
when
simply
just
human
error
or a machine
system
breakdown
track
system
configuration,
input
current
differently
Woods,
1995
increases
a; Woods
without
activities.
is gaps
system. organizational
this
of design-induced
Jones,
given
as
and
in
each
has
(ways
always
been
in which
1951
in
the
that
and
aviation
for
for
make design domain).
mode
error across
state
and
automated
monitoring
and
training.
The
for
models
concerned
and
interactions
limits
being
user
(Sarter
potential
of the
automated
technological, affected
with us
by
the
dumb) (e.g.,
-.
loses
interprets
operation
human,
system
15
who
of the
things
to improved
a user
opacity
and
errors
both
and
others,
(one
mode
system
mental
in a
are a kind
adequately
affecting
that
of
errors
a system of
case
configuration
about
in users'
design
Mode
it takes
resource
human
is the
feedback
complexity
especially
contributions Fitts,
the
the
an intention
Note
The
which
and
error
one
of modes
operators
interact,
error
1947;
5).
To
ways.
in
mode
chapter
to train
example
began
current
improve
misconceptions
factors
Factors
1994,
coupling,
systems
and
In
fundamental
to
it difficult
these
the
of a proliferation
resulting
makes
managing
and
et al.,
changes
The
system
result
on
as a consequence
modes
Human
depending
and
in
training
failure.
in that
1994).
influence
executes
were
al.,
found
to go behind
induced
operator
has
system
that
configuration.
of human-machine of the
of design
device
in a different
factors
in predictable
an
if the
need
we
risk et
that
organizational,
of a kind
occur
(Woods the
the
are
and
improve found
performance
be appropriate
it is, in fact,
issues
analyze
there
example
errors
to safety
of deeper
and
human
Mode
when
contribution
constructively
words,
classic
would
and
to identify
influence
and
human
researchers
In other
simple
that
issues
error
that
mode
the
is a symptom
these
performance. factors
on
and
the
others.
identification as one
Fitts,
However,
of
1946; it
its
Fitts is
a
profound
misunderstanding
a shift 'the
from-
'the
incident
was
if we
trade
pilot
There
are
always
jointly
sufficient.
of
incident
4.2
for
Design
human
designer
or
the
that
lead
such
as human
over-automation
are
We
each
factors
often for
-
no
progress
et
al.,
necessary are
to -
1994).
but
a part
progress
forms
implies
error"
make
only
of the
comes
managers
inappropriate
this
(Woods
to failure
or
that
operator
error
predictable
or through
or
error'.
designers
for Human-Centered
If diagnoses
error
potential
towards
to think
manager
organizational
again
factors
of technology
by pilot
contributors
performance
Strategies
caused
results
or designer
and
But
research
manager
multiple
the
use
by
error
understanding
clumsy
was
caused
contributors.
shape
of the
set
from
inadvertently
of error
to
through
organizational
the
pressures.
Design
error
(be
misleading
it operator,
and
designer
unproductive,
then
or
manager)
how
do
we
or make
progress?
A necessary and
design
three
step
(Billings,
basic
centered,
1.
first
1996).
attributes:
and
is to adopt
context-bound
research
A problem-driven
approach
the
difficulties
meet activity
2:
the
these
basis
for error
and
challenges
demands?
across
Human-centered
people
This
perspective
Human-centered
Human-centered
modeling
"human-centered" can
design
is
and
is problem-driven.
begins and
with
is the
in routine
and
and
can
in that arise?
nature
of
exceptional
design
of practice.
do people
collaborative situations?
is activity-centered.
16
activity-
in understanding
field
How
of
1997).
an investment
expertise
that
Woods,
research
in terms
problem-driven,
and
design
to
be characterized
(Winograd
What
research
approaches
use and
and What
artifacts coordinated
are to
In
building
and
designers and
often
the
see
on
activity
that
with
the brings
field
New
possibilities
people
emerge
in a field
with
computers
goals
and
becomes and
in different
ways,
in
tools
activities
(b) how
in that
the
field
do
Human
collaboration,
cognition,
classic
example
is
reproduced
finding
influences
the
improving
or
other
words,
differently,
the
same
can lead
Another
of this
problem
is not
Rather,
this
context
in which
just
state
problem
so
it appears. related
de-emphasizing
the
to the
or will
in
the
of
interacting
the
practitioner's
The
artifacts
activities
involve
becomes
question
shape
of their
performance -
to
(e.g.,
the
so that
they
depend
on
then
cognitive
goals
and
_
task
function
as
context.
A
the
work data
amount what
worse, of data;
17
fundamental
data the the
and
a problem
solve
Zhang
a formal
is the
a
How
cognitive
pieces
operating
shifts
pursuit
effect
from
Even
try to make
1995).
artifacts
needed
because
we
pressures
domain.
Science.
much
design
the
is context-bound.
and
example
is hard
of other
design
performance
to different
focuses
in press).
work
problem
This
focus
adapt
Cognitive
degrading
them.
other
in
representation
in
of performance.
the
and
cognitive
the
and
(Woods,
research
between
do
task
practitioners
3" Human-centered
human
Dominguez,
the
underlying
of people
of activity
(the
of analysis
but
and
systems
isolation,
activities
researchers
separate
and
and
focus
do computer-based
coordinative and
the These
(a) how
context
Flach
of practice.
activities
in
constraints
1988;
when
interaction
use,
In human-centered
to the
(Ehn,
human
of two
technology
together.
sensitive
of activity
of
or the
them
for
in terms
aspects
people
technology
actual
technologies
the problem
computer)
attention
new
studying
is
that
Norman,
description,
when
therefore
overload of
data
either 1994).
different
be
is informative
levels
At the sifted
consists
context
also
of
includes
heart
through.
depends
context
In
represented
problem. to
represented
problem,
and
and
much
on
more the
the than state
of the problem solving process and acting
in that
in most
impact
on
cases,
human 1997;
developers thoughtful
et
their
work
the
promise Eventually,
The
to make
the
computer
undergo
their
operational
complexities
words,
road
intentions"
(see
come
is a gap
between
actual
those
on
broadest
strategies contribution
that
level,
users.
into
play,
if at all,
frontier
or
to
practice,
activity.
of
only
which
human-
be
later user-
results
deck.
with
These
at this to
flight
hold
engineering
intentions
is paved
all
to users.
usability
automated
which
and
Knowledge
designers'
systems
because
seems
technology
the
intentions,
around
that
and
to
In
in other
user-centered
et al., 1997).
5 Progress
At the
the testing
the
good
collaboration
technology-driven
like
lead
addressed
technological
a technology
and
motivated
will
they
activity the
to potential
usability
system
such
primary
connect
to technology-centered Sarter
new
are
technology-centered
usability
accessible
they
eventually
cognition,
some
and
Well-intentioned
is pushing
which
its hypothesized
press).
Despite
albeit
on
because
because
focus
human
there
and
"the
people
(Winograd
the
is the
system,
technology
However,
centered
of the
performance
in
fundamentally
primary
interaction
stage.
and
are built
typically
Woods,
predict
in itself
influence interfaces
interfaces
they
remains
technological
to
1997;
and
developed.
technology
is organized.
creating
expectations
based
is human-centered
system
usually the
al.,
performance,
the
development developing
and
is developed
collaboration,
because
in of
technology
Sarter
people,
improvements
else
new
cognition,
feel
usability
goals
situation.
Today,
Woods,
the
researchers
organizations
Depends
have can
on ...
identified
follow
to safety:
18
in
an
a few effort
basic to
human-centered
increase
the
human
•
increase
•
avoid
•
evalua_
specific
increase
To
the
Design,
operational,
together
to adopt
and
certification.
- to work but The
goal
factors
5.1
Avoiding person of
create
diversity"
to turn
of
that the
test
ability
complexity
or
will
of
not
is the
Trying
to
change
banishing
to
the people
steps
use
them
to
state,
Factors
into
practical
error
tolerance
issue
modes
Failures
to manage
source
accumulated
eliminate basic
vulnerabilities with
19
and
other
because
no
single
and
in the
pursuit
customers, modes,
this
created
from
is the
options
problem
behavior
failures.
detection.
in complexity
complexity
"erratic"
associated
of the
(valid
features,
features,
-
design
as additional
all of these
But the
of design
and
multiple
for this
work
community
But
increase
all
methods
eliminate
accommodate
needed.
as part
complex.
complex
are
must
center
error."
source
and
systems
circumstances.
as "human
view.
more
cost
try to manage
The
Neither
and The
of: operational
trying
several
is a difficult
to make in
get more
accumulate.
will
potential
observability
Human
results
to detect
operational decides
and
for effective
excess
training
of their
organizations
to the
research
Complexity
person.
point
a challenge
Operational
are categorized the
creates
safety
analysis
Excess
must
the
regulatory
errors.
gradually
who
and
predictable
options
improving
to
for error
improvements
systems
user
methods
or organization
local
and
research,
is to improve
Avoid
by
contribution
economical)
that
in terms
error,
detection
human
industry
resource
of human
training
expertise.
This
with
and
intentions,
in human
improve
complexity,
at error
and
to errors,
in technology kinds
skill
invest
tolerance
operational
changes
activities •
system's
excess
create •
the
across
complexity is not
inside
an
operational
through
remedial
by the Instead
complexity. human
error
a
is a symptom involve
coordination
system that
of systemic
approach
Mode
Not
operations
modes
"basic."
Which
for safe area
and
different
Making
flight
and area
5.2
where
This
industry.
that
will
coordinated
about
information
need
they
in
all
transition
the
factors
in
contribute
to
only
of
different
and
which
of the
purpose
difficulties
to variations
and
define
indication
same
due
the
training;
complexity
Another
and
available
carriers
excess
the
change
or carriers
are
different
achieve
for
all pilots
taught
operation?
that
mode
across
teamwork
is
have
a set modes
are
different
as
essential
disarray
requires
coordination
across
an
competitive
in
ways
but
simplification
modes
changes
been
pilots
transitions
is of
(indirect
have
between
making
Error
and
in simplifying
mode
for
fixes
in
this
names
on
many
international, needs
to
be
in others.
interaction
in
are
represent
industry
place
Indirect
by
Still
which
the
system
decks.
progress
collaborative
of the
used
modes
modes
modes
multi-party
are
taught,
efficient
is that
One
is
are
performance.
preferences. all
in
meaningful
is illustrative
Not
solutions
parties
with
human
all modes
and
complexity. "basic"
start
affect
simplification
involved.
The
of multiple
must
predictably
factors.
and
better
mode
very changes
identified
importance
or
as a major
automation.
fit pilot
great
mode factor
Simplifying
models
is another
is
the
reversions). in breakdowns
these very
transitions high
priority
improvement.
Detection
Research
has
machine
systems
providing
better
through
shown
that is effective
feedback,
Improved
a very error especially
Feedback
important detection. feedback
2O
aspect Error about
of high
reliability
human-
detection
is improved
the
behavior
future
by of the
aircraft, be
its systems
balanced
with
investment going
One of
area
of need
authority
evolve
picture in the
without discussed generally
behavior
the
is
technical
available
and
context
of
Observability presence
1992,
is of data
sufficient
and
change
increases
and to
system's
and
state,
but
to increase integrated these
may
of machine
agents
surprises.
has with
machines'
how
automation
surprises
behavior
an
authority
critical against
future
need
activities,
a
feedback?
emphasizing
agent
can
guarding
and
feedback
leads
is
current
miscommunications
shown the
rather
As
that
crews
automation only
from
when
aircraft
abnormal.
of
low
to the term
observability
cognitive
captures
observation
that
distinct
from
in some
form
to have
feedback
is a concomitant
automation
symptomatic
This
of
complexity
in better
the
autonomy
automated
data.
there
observability
their
refers
about
situation,
on
that
how
technological
forms
sufficiently
term
from
not
data
becomes
result
in
detect
about
to invest
Increasing
increase
do not
and
complexity,
of the current
earlier,
performance
As
new
future.
an
displays
and
through
d.vnamic
human
feedback
systems.
increasing
Improving
improving
is improved
In general,
feedback.
But where
automated
observability
This
for
scenarios.
autonomy,
automation.
improved
area
sour
the
or the
the is
data
needed
to extract
relationship
to
availability, location.
in
among
fundamental
in some
something
observability
work
where
front
of
your
data,
refers
human eyes
observer feedback.
to
the
mere
perception,
to see
the
meaning
effective
which For
is
it"
"it
is
(O'Regan,
p.475).
Observability results what support Billings,
from
refers the
interplay
information
at what
attentional 1997).
to processes
The
between point
guidance critical
involved
in extracting
a human in time
(see
and
Rasmussen,
test of observability
21
user
useful
knowing
a system 1985; is when
information. when
that
the
to look
structures
Sarter, display
It
data
for to
Woods
and
suite
helps
practitioners
notice
expecting
One
(Sarter
and
example
of
generation flight
display.
"changes
injunctions
environment.
effective
Minor
new
concepts
•
transition-oriented
•
future-oriented
need
and •
reveal
what
pattern-based possible
For
example,
usable map
display
target--
an
situation
and
check support
vertical
better
you
on
one
or
the
pilot it."
indications
in
a
annunciations
adopt
to
Simple
in these
attention
and
primary
mentioned
at
in feedback.
current
contribute
stare
changes
changing is not
very
Researchers
and
fundamentally
new
activities.
navigation
display
trouble
navigation
future
rather
to make
profile
22
the
current
sequences
above
quickly have
a
big
up and
assessment.
comprehensible display.
picture
in a way
criteria
pick
to read
The
example
However, the
than
more
of vertical
provides
and
an overall
developments
on
only
transitions,
significant
at a glance
modes
detection. based
captures
is a tremendous
that
and
when,
to scan
of data
navigation
the
and
conditions
form
events
operationally
next
piece
some
about
generally
be able
vertical
especially
feedback
approach
should
horizontal
and
for
the
activities
mode
automation
happen
individual
integrated
reading
on
As
unless
test
is to highlight
to require for
annunciations
current
or abnormal
making
is likely
about
should
-- pilots
each
mode
redirect
develop,
current
goal
unexpected
integrate
to
looking
to be:
-- the the
observability
improvement
-- provide
configuration;
to
of the
crews
low
at or call out
ways
to cooperate
to inform
in
closely
significant
specifically
transitions.
sneak
tuning
any
need
approaches
The
not
were
of automation
mode
always to look
to provide
industry
indications
pilots
are
flight
tracking
can
for
generally
likely
with
very
is the
crude
they
a).
with
decks
These
what
1997
displays
problems
us,
than
Woods,
of flight
reported to
more
that
developing is much
of
and moving
of the
desired
the
current
supports
quick
displays more
difficult
to
because it needs
is inherently
to develop
of vertical
Going
begin
need
and
and
comprehensible
How
Let
us look
flight
and
In
1990).
new
industry
to support
economic
pilot
as a whole management
this
we
are
ensure
ready
better
a collaborative
awareness
to
demands
to support
researchers
that,
feedback
prudence
feedback
need
and
to aid
to
improved
pressures,
regulators,
this
that
is better
do
innovations
we
evidence
what
To
on
Feedback:
nears
they
take
over,
when
process
to prototype,
test
and
monitoring.
W e
the
next
provide
more
Transfer
of Control
on
This
window
of
observable
the
the
too
and
general
extra
trouble
developing
silently
trouble
until
to compensate.
or be unprepared
to handle
the
a bumpy
of control
and
transfer
problem
example
or
and
has of
been
this
a part
disturbance significant
of several
is asymmetric
lift
incident conditions
trouble.
unusual on
for
crews
capability
in
One
of the
of its authority
late
between
compensate
unaware
a well-coordinated
parameters notice
limits
breakdown
can
remain
resulting
or engine
in
comment
the
scenarios.
contrast,
of a coordination
can
over
excursions.
by icing
Bumpy
Automation
Crews
take
accident
could
on
automation.
may
relevant
progress
example
crew
caused
with
Better
at one
automation
control
conflict
up,
to Provide
(Norman,
once
concepts
The
automation.
deck
The
problem.
provide
carriers,
opens
5.3
accidents
forward
opportunity
display
recovery.
adopt
to move
the
the
manufacturers,
in context,
new
and
to make
detection
dimensional
automation.
Despite
we
among
test
incidents
is needed.
error
and
navigation
sour
that
a four
human
difficulty target. work
or
increasing
Or, in an or effort
team,
open
exerted
23
the
active
effort
needed
environment, by his
or
partner
her
would
to keep the
supervisor
partner
and
the
ask
about
the
safety
goals,
How an
difficulty,
can
we
open
of
extreme need
analogy
to a well
environment
and
set
the
the
visible
feedback the
use
investigate
better
ends
how
between
problems
that
of its envelope
we
when
authority,
to achieve
human
human
arise
or
or intervene
coordinated
to guide
coordination
feedback
problem,
team
can
provide
and
machine
working
more
automation
improved
overall
in
effective
partners?
For
is working displays
and
at the warnings
to indicate: when
the
automation
is
having
trouble
handling
the
situation
(e.g.,
turbulence); •
when
the
extreme
end
when
This
are
having
How
and
rather
trouble
when just
From
experience
these
questions.
and
gradual
escalation
signal trouble function,
one
thresholds
research
not
simply
we
often
or staged
may the any
but
crossing
know
the
to make
the
in the
very
function
not
level
constraints (simple or
or
or
action.
24
say
too
kind
automation well
yet
to define
what
way?
When
failing
to perform?
moving
indications late
How
is an
towards
a limit
alarm?
some
too
is how
sensitive
communicate
set
shift
whenever
performing
towards
surface.
intelligently?
a function,
crossing
for trim-in-motion) in
moving
question
in a context
a threshold
are
sounds
or
of a flight
design
this
effectively
Threshold
-
The
in performing
invoking
enough
that
target.
of authority
does
action
for control
to communicate
regions
than
warning
in competitior_
enough
"extreme"
extreme
of authority;
a performance
smart
agent
is taking
of its range
agents
specifies
system are
automation
too
extreme
on
the
alarms)
early.
answers are not
We
need
of feedback. is active much. action
(e.g.,
We to
want
to smart
a more
An
auditory
an
auditory
to indicate
accomplish
the
We
know
certain •
from
errors
can occur
nuisance
excessive
•
distracting
false
words,
silent,
through
multiple
up
feedback
more
systems
that
include:
that
talk
too
serious
tasks
are
being
that
comes
on
at
that
thing!").
-- "silence feedback
too
can
little,
into
perceptual
salience;
existing
attention?
prototypes
occur
too
talk
late
displays?
in other
in terms
or
Should
too as
much,
much
in
handled
a high
too
the
(e.g.,
noise
soon
automation
through this
Should
words,
Working
perceptual
visually
indications?
integrated
a
level
or it can
moves
be
towards
out
how
these
the
auditory
channel
be
a separate
new
the
indication
be
strongly
design
should
decisions
or
indication of
the
very
signal
requires
or high
capture
developing
of:
salience
relative
to the
larger
context
of other
possible
events
relative
to
signals,
along
a
priority along level
temporal
dimension
of other
issues
a strength
adjusting
the different
these
relative attributes
effective new
signals kinds
(when
or activities
dimension
of abstraction
Developing
and
alerts
similar
limits.
the
about
with
These
as voice
or a warning
misdesigned
Should
and
when
situation
speaking
authority
•
feedback.
such
warning
a difficult
In other
•
and
alarms,
trim
during
and
in designing
indications
constant
•
domains
situations,
•
pilot
in other
communication
wrong
too
experiences
feedback
going
(how
to ongoing based
about
on
communicate on
much
the
then),
or how
little
to say
and
at what
activities) data
on crew
automation
or indications of situations.
to
in One
the cannot
25
performance.
activities context improve
of
requires other
thinking
possible
feedback
or
signals increase
observability
by adding
time
arise.
as they
symbolic
but
crew's Instead, point
Our
will
need
to look
to the
need
for
performance bounds
balanced
in
identified
effective
design,
and
explored
in
detail
illustrates
the
complexity
5.4
Mechanisms
Giving
users
one
side
also
giving
their
the
reasoning
position. but
manual
potential
legwork focus this
more
the
those
decisions.
solution
agents
users
the are
say
computational
in working critical
machine
high-level users
need This
to
through
means
and
testing, that
that
will
set
must
be
need
The
large
given giving
26
in
to be
example
processes
team
players.
agent a
with way
of machine from
much
problems, However, the them
as
a resourc_
thus
other
authority control
machine's
than
agents
through provides
the
and over
them
to make
use
capabilities the
a
mundane
allowing
in order
in
improved the
of
is only Without
significantly
wrong
it in any
users
process.
into
what's
power free
reasoning
not
decisions. to be
human-machine
testing.
machine
to influence
i.e.,
all
Resources
agent's
to
which solution.
design
factors
the
situation.
tradeoffs
user
machine
able
relevant
be
for observability.
to direct
the
advantage,
potential,
make
ability
be
The
involved on
the
powerless
takeover.
great
the
might
remain
and
Automated
processes,
They
solution,
prototypes
in making
users
of the
will
it challenges
an integrated
some
at a more
data
of problems
for
one
displays,
in a particular
the
identified
case
More
because
scenarios
some
more
subsets
identified
mentioned
into
and
each
alarms.
to devise
of designing
visibility coin
sets
solutions,
through
more
is relevant
relevant
to Manage
of the
what
has
generate
feedback
feedback
example
on
sounds,
at coherent
to address
will
effective
digest
improved
one
targets,
be
on and
we
of this
more
not
or alarm
approach
displays,
to focus
analysis
some
on this
ability
indication
A piecemeal
codings
available,
a new
problem
to of to
A
commonly
automated
proposed agent
users
determine
Thus,
the
system The
and
machine
agent
productive
automated
agent
drawbacks.
Either
the
practitioner's
the
machine
in
several
and
the
not
do
entire
over
the
need
tools.
obvious from
brittleness
of or
Previous
work
aviation)
systems,
shown
manner
the
of a deteriorating
that
to be able
in a cooperative
the
(expert
has
the
having
many
a
more in
troubleshooting,
agents
or
benefiting
despite
of cognitive
algorithms)
has
middle
manual.
in
like
without
and
electronic
users
it,"
job
in the
machine
Instead,
problem,
joint
agent
is
support
planning
agents
agent
specifying
to
the
course
may of
the
possibilities
is difficult
interactions
between
may
challenges
be meaningful
very what
users
main
will
take
exactly
as a resource
of the
of interaction want
the
path
a human
which
takes
of
either
method, I'll
the
fully
This
operations,
types
or
cooperating
knowledge,
the
and
agents
or
where
cockpit
this
is a poor
to continue
by taking
to work
control
of
the
agents.
one
to
(space
architecture.
observability,
others
user
machine,
machine
it
the
adequately.
and
does
without
different
machine
may
or the
the
interrupt
in situations
automatic
that
with
do
and
to
a problem
the
- fully sense
deal
users
solving
of critiquing
you
machine
is not
in the
human
allow
in its entirety
modes
to
information
the
modes
two
"either
the
flight
role
problem.
say
automated
automated
Using
the
domains
cooperative
the
asked
is to
agent
only
the
situation
automation,
the
of
agents;
with
with
be
solving
challenging
into
this
problem
machine
system
can
sense of
the
in essentially
is a joint
for
over
the
is cast
operates
process
take
that
human
system
remedy
detailed
want
only
solution and user
are
goals,
determine
users. of
very
situational
27
things.
In
some
and
very
progress.
requires
is to
made
to make in
various
to
control
decisions
mean
in
general,
what some
portion what
of
factors,
level
the
users
a
problem, while
analysis the
in
corrections all
iterative and
and
sequence,
high
of
levels
cases,
Accommodating careful
In terms
nature
of
these of
the
of
the
machine perform
5.5
agent.
However,
this
in the
broadest
effectively
Enhancing
The
last
Human
area
contribution industry
for to
possible
investment
in
the
is human
to be reducing
myths
the
as investment
in
expertise.
automation
creates
In our
investigations,
of
automated
the
how
the
about
makes
models heard the
from
sources mode long
or level
instead their didn't
you
tried
to get
knowledge where
tangled
web
incidents
get
this
path turn
it to do what and
different
skill kinds
requirements.
personnel
as a resource
pilots
need
and
modes
Response: I had
demands of factors
having
For
programmed are most push
28
in
beyond
Many
persisted
their
too
intentions
to accomplish may
it was
ask,
supposed
heard
relatively the
W e
a particular
they
means
We
mental
to manage
getting
someone
do what
relevant
flight
situations.
out
direct
heard
logics.
how
where
it to do."
events
crews
trouble
example,
"It didn't
We
transition
to carry
or a more
about
automated
flight
successfully,
means goals.
and
to teach
automation
complexity
or erroneous
differing
were
the
knowledge
the
in
increased
function. of
is
is needed
that
new
complexity
in
to work
say
reflects
performance
skill
struggling
management
This
human
and
modes
of
it points
how
of automation
mode
when
shown
the
pilots
time
aviation
have
oversimplified
to another
it off?"
on
to develop
where
very
the
sources
that
how
departments
of switching flight
show
that
to accidents.
knowledge
means
of automation to
is to
human
investment
subsystems
for pilots
systems
offered
trying
that
ironic
the
less
operational
automated
training
automated
system
increases,
different
deck
is
of automation
many
heard
flight
it easy
of the
and
we
investigations
deck
impact
fact,
new
different
joint
improving
contributor
automation
In
of
at the
one
about
interest
investment
this
as a dominant
human
the
of scenarios.
It
performance
that,
range
if
expertise.
to human of the
is crucial
Expertise
safety
seems
process
routine
how rare
"Why to,
the
so I new
situations
- just
those
circumstances of
that
are
misassessments
practice
For
demands
managers
and
must
be fit into
of new with
heard
mismatch. basic
For
example,
set of modes
can
create
the
managing most
complicated
pressures
one
the
parts
shrinking
and
skills
that
that
for individuals
easiest
to learn
later
The
levels
cope
of
with
training on on
to learn,
of training
bind.
to
to be learned
to
footprint.
double
focuses
need
of new
transition
training
are the
training
developed
is to focus
the
deal
as a result
been
a progression
increases
a training
remainder
situation
through
is a great
and
have
tactic
systems
result
creates
that
leaving
ironic
automated
a small
tactics
and
the
knowledge
economic
many
sour This
departments,
roles,
about
to going
miscommunications.
of situations.
that
automation
vulnerable
kinds
combination
We
and
those
training
most
on
the
on their
just
line.
those
while
this a
This
parts
of
deferring
own.
the
This
tactic
works: •
•
if the
basics
provide
parts
or for
coordinating
if
there
is
continued
Another It
is
overwhelmed and
recipes
of
need
training economic
as much for
settings scenarios, and
acknowledge as their
pilots
to
through although
competitive
practice line the forces
limits
and
difficult
circumstances,
oriented scope squeezing
bind
and
checks
students
of this
they simulation of
is to teach
systems.
resources
what
29
more
supports
prevent
automated
the time
difficult
double
helps
of the
the
requirements.
training
and
complexity
managers
beyond
operational
tactic
learning
encourages,
minimum
this
aids
in more
that
with
efficient
that
automation
beyond
to cope
by the
training
the
used
time
the
base
environment
learning
tactic a
an
a coherent
this training
from Still,
approach
limits have
training time.
line is We
being
instructors and
allowed. learned
and
recipes.
try All
in oriented limited saw
to go spoke
realistic flight by evidence
the
of
an industry
time
and
limited
As
the
the
highest
do
so
under
program
or
priority
same
goal
squeeze help
more
and
rather
yield the
transition
practice
from
the
FAA
will
be
quality
a shrinking
training
in on
freedom
taken
in productivity effective
reduce
pressure
means
(reaching
training).
in human that
label
the
Trying
expertise
we
to
qualification
inadvertently
Economic
accidents
focus
advanced
further.
(more
and
and
increased
is, it can
investment
of incidents
has
(the
as this
even
in
is to identify
US industry
laudable
than
kinds
of limited
response
The
of improvements
faster)
prevent
one
with
However,
training
utilization
needs.
programs
AQP).
benefits
shrinks,
training
new
better
checks.
footprint
for
the
to get
recurrent
training
resources that
struggling
to
will
human
not error
after-the-fact.
Escaping
from
limits
of
this
minimum
oriented
towards
produce
an
training
should
growth
of
initial increase
can
the
full
management
knowledge
platform
Developing
becomes
practice
a pilot's
and
career? this
skills
range
we
resources In
approach. as evidenced
many Pilots,
expand
culture should
deck.
support
This
continued
improvement
of
beyond
systems
we
situations
useful
mental
of possible
that
see
pilots
models
conditions
an
that
depends
situations.
the
across
opportunities
a wide
ways,
the
in
general,
by pilot-created
3o
flight
that
range and
a
the
training
automated
in line-oriented
can
produce
automated
the
accurate
a wide
how
the
highly
and
is to recognize
or transition
continuous
with
requirements
of automated
to adopt
on
step
should
for mechanisms
because
mission
then
for managing
across
we
Initial
emphasis
is needed
effectively or
question
prepared
An
to master.
be applied
throughout
as the
knowledge
on part-task
The
serve
A first
Instead,
learning.
proficiency
expertise.
be able
is essential.
requirements.
initial
in
bind
continuous
proficiency
must
double
variety
aviation want guides
to practice of situations
industry to improve to the
is well their
automation
that
we
noticed
heavily less
in
in several
line-oriented
expensive
but
high
training
centers.
The
training.
New
training
fidelity,
part-task
industry
already
has
technology
training
in
devices
invested
the
form
of
is being
utilized
from
research
more.
5.6 Coordination
These
comments
on Human of
Among
also
Factors
changes
and
the
areas •
•
pressures
the
best
evidence
job
trainers
may
to use
it within
designers
•
the
with
of
improved
may
limited
encourage the
large
of circular
need
for
of them,
others
time
etc.).
Each
group
look
are
under
minimize
knows on for
the
so that
system
levels
industry
footprint;
placed
they
that
the
they
are
them.
So
when
solutions
in
other
we
can
train
people
window, to
get
ATC
to provide
to
better
accommodate
training
features
regulations
on items
is But
reaction
of lower
wrong
there
in
to enable
designed
across
considered
detail
that
alone,
has
31
all
reading
our
people
as a result
of
forced
of
these
to these
have
procedures,
improved
a great
to spend
aircraft.
areas
can
messages.
is symptomatic
traditionally
operational
are not
for glass
--
of glitches
areas design,
so they priority
is a deeper
to evidence
-- training, when
training
set of interconnected
solutions
coordination
this
of
high
capabilities,
for modified
in isolation.
kind
segment
for very
demands,
lobby
these
demands
for example:
efforts and
emerges
performance,
resource
advocate
training
autonomously
deck
with
that
re-designing
this
market
trainers
None
to flight
that
constraints
it is natural
advocate
may
precious
one
given
idiosyncrasies
designers
multiple
it all into
possible arises,
theme
each
deck,
may
to cope
(fit
of
flight
contribute
automation's •
in industries
previous
of glitches that
a general
Representatives
from
doing
illustrate
problems
performance.
constraints
Stakeholders
This
of a deeper
functioned
mostly
certification. deal,
be
and
Each this
has
created
the
However, the
In
generally the
risk
interaction
fact,
these
of failure
increasing
specify
the
For example,
inadequacy
training
of
a
this). in
have
become
more
control
(ATC)
While
of coupling
entity,
and
under
intense
economic
room
to make
and
limits
change
in
complex paper, the
the
it may
significant
precisely effects
advanced
procedure exact -
and
a price erode
because for other
aircraft,
seem
change parts
interact yet
ATC
one
system.
32
of
the
no
degree.
of the
as
an
has
undergoing well. at least member
Coordination
aviation
the
capabilities
as
team
are
becomes
the
is a system
air
multiple
longer
with
-
system
learnability
a clumsy
areas
area of
throughput,
to some part
each
pressures
of managing
in any of the
for
to increase
margins
parts
training
performance
in terms safety
all
aircraft.
other
integration
that
demands
and
may
in
Design
ATC
economic
desirable
that
means
deficiencies.
designers.
of
fact
multiple
automation
economics.
the
and
to try to
new
many
deck
on
design
these
and
demands
This
design
safety
comment
on
how
in part,
means of
flight
and
to the
pressure.
for
face
general
develop
technology
increased
possible
is due
of managing
automation
needed
this
on
departure but
in
part,
up
still
people between
(and
between
designers,
many
to show
aircraft,
coupling
integration
be extended with
the
linkage
advanced
involves
areas.
heard
a closer
industry.
scenario
automation
operator
the
advanced
are
In
constraint
of
can
progress
another
be
inter-related
perspectives.
the
to
aviation
sour
increases
have
and
the
going
how
We
in
individual
recognize
because
improvements
isolated
many
needs
part,
example
by the
automation
manufacturer
This
traffic
of
levels
'throw-it-over-the-wall'
There
perspectives,
safety
these
philosophy.
or between
reduce
between
level
an operational
high
exemplified
or coupling
areas.
the
extremely
system
A on -is has
6 Conclusion
Overall, there are accidents. and
We
the
need
into
symptom
of in
Second,
larger
we
-
breakdowns
and
in
a
tame
the
•
through
better
feedback
•
through
more
practice
of particular
kind
a minor
going
sour
coordination
symptom
of
people
or non-routine This
people
system
and
where
occurrence
between
(Woods,
incidents
of incident
scenario.
overall
levels
needed
details
against
the
organizational
can
the
to mismanage
trouble
are
behind
guard
seem
turn
operational
patterns
to better
automation
situation
which
broad
scenario and
is a
machines
complexity,
at
both
1996).
complexity to operational
personnel,
at managing
automated
function
as a team
resources
in
a wide
range
of circumstances, •
by making
•
by creating through will
the
"intuitive" better
produce
In general,
coordination
the
mechanisms
can
for
excess
complexity,
between
coupled
challenges
involved.
Since
the
they
should
should.
Since
that
industry low), the
it is easy status
quo
valuing
because benefits
to ignore
not
is sufficient.
be
where
the
quickly,
automation
design
problems.
growth
simplicity
learned
in complexity of
operation,
through increasing
areas.
sour
any
scenarios
change
pay
the
threat This
but
safety
level
is
going
particularly
33
a coordinated
exact
level,
costs,
of the
in
will
are at a system
the aggregate the
can
performance
to limit
of going
difficult
that
or predict
of human
trying
player,
designs
to detect kinds
act by first
extremely
claim
automation
predictable
we
checking
Meeting
automation
that
costs
it is easy some
is very sour
on
the
for each other
high
because
and
is
parties party
part
(actuarial
scenario easy
manner
argue going
to
of
the
risk
is
that sour
incidents
by definition
looks
like
being
human
climate
a unique
to
before";
minimum
progress, such
was
as
the
going
based
enough
before")
sour
continuing
progress
collaborative
has
and
been
to point
where
progress
out
to
how
constructive
a
directions
this
halt.
Yet
question
for
build
the
to
areas that
This
difficulties
forward
specific
a
risk.
The
is
factor
produced
if observed
addressed.
case
approved
crawls
then
enable
general
("you
progress
be
have
competitive
practices
operators can
and
Each
common
climate
is demanded to
more
environment
past
factors.
dominant
legal
and
are
that
research
on
scenario and
the
to financial
what
environment of our
and
is exactly
manufacturers,
collaborative goal
standards
contributing
with
certification exposure
its pains,
local
of events
creates
safe
despite
regulators,
The
The
change
"it
several
combination
error.
where
leads
involve
movement.
for
may
constructive help
create
a
is possible.
Acknowledgments
The
research
Ames
on
Research Dr.
Research
Center
our us
in many
FAA
team
that
flight
deck
systems
our
reflections
and ways.
examined
on
the
the
Dr.
1-209;
pilots,
investigations different
and
NCC
many
is based
Cooperative
Palmer
(Grant the
assessment
(under
Everett
to thank specific
this
Center
Monitors
wish
which
We the
are
implications
Kevin
Technical
and
indebted
to
between
and
debates
of the
research
References
34
2-592;
and
NASA
Dr.
designers
their
views our
Langley
who
contributed
W e
experiences members
crews
sparked
results.
Technical
Abbott).
fellow
they
NASA
Kathy
and
flight
by
NCC
Corker) Monitor
shared
interface
discussions
sponsored
Agreement
instructors who
was
and were
to with
on
the
modern critical
to
Abbott,
K., S10tte,
Lyddane,
G., Amalberti,
H.,
N.,
Sarter,
Interface
Between
Aviation
Administration
Billings,
Cook,
Approach.
R. I. and
Woods,
aviation
for the
Clinical
Anesthesia,
Cook,
R. I., Woods,
Report,
April
Degani,
T.,
and
Woods,
D.
Flight
Deck
Systems.
Lawrence
R., R.,
(1996).
The
Federal
18.
The
Search
For
Erlbaum
A
Human-
Associates.
of automation
intravenous
Lalley,
T., Pearson,
Implications
of total
anesthesia
surprises
(TIVA).
in
Journal
of
8:29s-37s.
McDonald,
Corpus
Shafto,
cockpits:
of
prepared
Some
M., initial
International
Systems
J.S.
Cases. for
Sweden:
(1991).
Human
Cognitive
Performance
Systems
Anesthesia
Reporting
Kirlik,
A.
observations.
Patient
in
Engineering
Safety
(1995).
Mode
usage
In T. B. Sheridan of
Conference.
Work-Oriented
Foundation,
Automatic Boston,
Design
of
in
(Ed.),
Proceeding
Control;
MA:
automated of
Man-Machine
IFAC.
Computer
Artifacts.
Stockholm,
Arbetslivscentrum.
D., Dodd,
Flight
and
Federation
(IFAC-MMS)
P. (1988).
Eldredge,
N.J.:
Imrich,
F., Newman,
D. C. June
Hillsdale,
(1996),
S.,
1991.
A.,
the
Modern
Automation:
D. D. and
Laboratory
and
R.,
Aviation
future
A
R., Fabre,
: Washington
D.D.
Anesthesia:
E., Hecht,
Helmreich,
Flightcrews
C. E. (1996).
Centered
D., Bollin,
G., Thiel,
Tigchelaar,
Ehn,
S., Stimson,
R.S., and
Management System.
Mangold,
System (Battelle
S.J. (1991).
incidents Report,
35
A review
reported
to
prepared
for
and the the
discussion
Aviation Department
of Safety of
Transportation).
Columbus,
OH:
Volpe
National
Transportation
Systems
Center. Fitts,
P. M.
(1946).
Journal
Fitts,
Psychological
of Aviation
P. M.
(1951).
Stevens
requirements
Medicine,
17(3),
Engineering
(ed.),
in aviation
of
design.
270-275.
psychology
Handbook
equipment
and
equipment
Experimental
design.
Psychology,
In
1287-1340.
S. S. NY,
Wiley.
FiRs,
P. M.
and
Jones,
"pilot-error" Report
experiences
Hutchins,
operating
Dayton,
Dominguez,
Instrument,
E.
Science,
of
factors
contributing
aircraft
controls.
Aero
Medical
OH:
C. O. (1995).
and
(1995
to
460
Memorandum Laboratory,
Air
Goal.
a).
How
Use-Centered
Integrating
the
Ergonomics
a
cockpit
Design:
in Design,
remembers
July.
its
speeds.
Cognitive
19, 265--288.
Hutchins,
E. (1995
Hutchins,
E.
Report,
Analysis
Command.
J. M. and User,
in
TSEAA-694-12.
Materiel
Flach,
R. E. (1947).
b).
(1996).
Cognition
The
in the
Integrated
Department
of Cognitive
Wild.
Mode
Cambridge
Management
Science,
University
MA,
MYF Press.
Interface.
Technical
of California,
San
Diego.
KeUy-Bootle, Press,
S.
(1995).
Cambridge
The
Computer
Contradictionary
MA.
36
(2nd
edition).
MIT
La
Burthe,
C.
(1997).
Presentation in the
Lanir,
Moll
13-16,
E., Cook,
automated
Stassen,
R. I., Woods,
Systems
Norman,
1992,
Pergamon
D. A. (1990). interaction,
Royal
D. A. (1993).
Design
Industrie.
Safety
and
Security
66-72.
Yue,
L. and
context: in
and
Howie,
Physician
the
heart
Evaluation
interaction
room. of
M.B.
In
H.G.
Man-Machine
p. 263-274.
'problem'
of automation:
'over-automation.'
of London,
D.D., in
Airbus
DC.
July,
controllers
Press,
The not
Society
Norman,
interaction
Analysis,
Aviation
Interactions,
intravenous
editor,
on
at
Washington
of alienation.
Human-computer
with
perspective
Conference
January
Agents
Charante,
(1993).
and
Century.
(1995).
van
Factors
at International
21st
J.
Human
Inappropriate
Philosophical
feedback
Transactions
of the
B 327:585--593.
Things
that
Make
us Smart.
Addison-Wesley:
Reading,
MA.
Obradovich, cope
J. H. and with
Factors,
O'Regan, world
poor
38(4),
J. K.
Woods,
HCI
D.D.
design
(1996).
Users
in computer-based
as designers: medical
How
devices.
people H um an
574-592.
(1992).
Solving
as an outside
the
memory.
"real"
mysteries
Canadian
of visual
Journal
perception:
of Psychology,
The 46, 461-
488.
Rasmussen,
J. (1985).
Trends
in human
reliability
1185-1196.
37
analysis.
Ergonomics,
28 (8),
Salter,
N.
B.
(1996).
Multiple
Agents:
Parasuraman
N.
B.
and
the
B.
and
N.
Flight
B. and
Woods,
Factors,
37: 5-19.
G. Salvendy, edition,
Wiley,
B. and
New
Surprises
B. and
Commission: A
the
to
In
R.
and
Human
with
Cockpit
Flight
Psychology,
Pilot
Study
System.
Full
D. D.
(1995).
and
Management
2:303-321.
Interaction
of Pilot's
with
Model
International
and
Cockpit
Awareness
Journal
of
Aviation
in the
world
in
supervisory
C. Billings,
C.
(1997).
of
Human
did
we
get into
control.
Automation
that
Human
Surprises.
FactorsErgonomics,
second
York.
Woods, Observed Mission
"How
awareness
Handbook
Woods,
Automation
In
Automation
Interaction
with
(1994).
Experimental
editor,
Agent:
N.
Pilot
of Aviation
D.D.
D. D. and
Independent
Sarter,
(1992).
Journal
error
N. B., Woods,
N.
Cockpit
Pilot
4:1-28.
Mode
Sarter,
Individual
Automation
Experiences
Management
mode?"
Sarter,
on
From
editors,
D.D.
Woods, II: An
Psychology,
Sarter,
Woods,
International
Automation of
Research
Mouloua,
I: Operational
System.
N.
M.
in
to Quality,
Erlbaum.
Automation
Sarter,
Quantity
Trends and
Performance,
Sarter,
From
D.D. A on
the
(1997
a).
Corpus
of
Airbus
D.D.
Teamplay
(1997
Breakdowns
in
Simulation
Study.
publication.
38
Operational
A-320.
b).
with
Human
Mode
Errors
Pilot-Automation Manuscript
a Powerful Experiences
Factors,
of
and and
in press.
Omission
and
Coordination submitted
for
Tenney,
Y.J.,
Level
Rogers,
Flight
Design
Deck
E. L. (1989).
Human
factors
Winograd,
T. and
Design.
In
Science
Woods,
D.
D.
(1996). In
Technology
and
D.D.
Cognition
and
J. and
Cognitive
T. Huang, DC,
and
Performance,
Designs
are
No.
on
High
Development
4669).
of
Hampton,
("glass
177528).
Challenges
P. ]ones, July,
a
VA:
cockpit")
Moffett
Field,
and
Analysis
(1994).
M.
39
Interactivity,
S. Kasif,
Apparent Mouloua,
(Eds.),
and National
Simplicity,
editors,
Real
Automation
Erlbaum.
about
How
Artifacts
Shape
in press.
Sarter, and
Center,
Representations
18:87--122
Human-Centered
1997.
R. I. and
Computers
for
Information,
Ergonomics,
Systems,
Science,
Report
Hypotheses
L., Cook,
D.A.
#
Opinions
technology
Automation:
Collaboration.
Norman,
the
Systems:
Human
Information
Toward Report
(1997).
R. Parasuraman
Cognitive
Ergonomic
D.
Decomposing
D. D., Johannesen,
Error:
D.
Washington
(1997).
Pilot
Center.
J. Flanagan,
Foundation,
(1995).
of advanced
Contractor
Research
Complexity,
tasks.
(NASA
Human-Centered
Intelligence.
Zhang,
Contractor
Center.
NASA-Ames
Woods,
(NASA
aircraft.
R.W. Issues:
Research
CA:
Woods,
Pew,
Langley
transport
Woods,
and
Automation
Philosophy
NASA
Wiener,
W.H.,
N.
B.
Hindsight. WPAFB,
Behind
Hu man
Crew Dayton
in distributed
Systems
OH,
1994.
cognitive
Complete
List
NASA
of Publications
Ames
Research
Cooperative
Agreement
Engineering
Book
Chapters
D.D. Sour
Woods and Accidents.
in
Domain,
N. Sarter,
Woods
D.D.
editor,
Wiley,
New
D.D.
Woods.
Erlbaum, and
Human
From
in press.
Automation
Surprises.
In G.
FactorsErgonomics,
Automation: and
Apparent
M.
Performance:
Quantitiy
Woods.
Cognitive
Management: editor, Human D.D.
second
Theory
to Quality,
Simplicity,
Mouloula,
From
Cockpit and
editors,
and
edition,
Real
Automation
Applications,
Individual
Pilot
Erlbaum,
p.
to Multiple
Automation In R. Parasuraman Human Performance, Erlbaum,
Woods
and
N. Sarter.
and
Activities
Evaluating
Cooperation.
editors,
Verification
Issues,
Springer-Verlag, Woods.
Demands
and
Validation Berlin,
Process
the
In J. Wise,
p.
and 267-
Tracing
Jersey,
D.D.
Woods.
Hochberg,
1993,
Engineering,
of New
of Complex
Systems:
Technology and
on
P. Stager,
Human
Factors
Methods
for the
Study
of Cognition
In G. A. Klein, in Action: Models
Outside
J. Orasanu and and Methods,
of R. Ablex,
p. 228-251.
Modeling and
Impact
V. D. Hopkin,
Fault In N. Stanton, London, 1994.
1993.
the Experimental Psychology Laboratory. Calderwood, editors, Decision Making New
in Dynamic
Abduction and Disturbance Management. Factors of Alarm Design, Taylor & Francis,
Human-Machine
Aided
NJ,
and Going Engineering
1996.
D.D.
D.D.
HiUsdale
Surprises Cognitive
1997.
Agents: Trends in Research on M. Mouloua, editors, Automation 280,
Applications
of Human
Decomposing and
Sarter.
from
2-592
C. Billings.
In R. Parasuraman
Technology 3-17, 1996. N.
NCC
Aerospace
Handbook
York,
Complexity,
support
N. Sarter. Learning from Automation In N. Sarter and R. Amalberti (Eds.),
in the Aviation
Salvendy,
on
Center
Research
Cognitive
based
and
B. Huey, Academic
predicting editors, Press,
human
Human New
4o
error.
Performance York,
1990.
In J. Elkind, Models
S. Card, for
J.
Computer-
Journal N.
Papem-
Salter
anti
Observed
and
Manuscript
D.D.
Agent: A Corpus the Airbus A-320. R.I. Cook and room. Human D.D. Woods. management. N. Sarter
Errors
and
Woods.
Teamplay
Adapting 593-613,
D.D.
and
D.D.
Experimental
Woods.
Study
Management 1994.
for publication, with
in the
Pilot
of Pilot's
System.
and
Model
and
in the
get
into
Human
Cockpit
that
Factors,
of the
on
operating
fault
mode?" 37: 5-19,
Automation
of Aviation
in intelligent
Surprises
in dynamic
we
Awareness
Journal
of flexibility
did
control.
with
Independent
Automation
technology
world
Interaction
International
D.D. Woods. The price Systems, 6:1-8, 1993.
and
Mission
1997.
a Powerful
to new 1996.
in supervisory
Woods.
Commission: in a Full
and directed attention 2371-2393, 1995.
"How
awareness
and
Coordination
of Operational Experiences Human Factors, in press.
D.D. Woods. Factors, 38(4),
and
of Omission
submitted
The alarm problem Ergonomics, 38(11),
error
N. Sarter
Mode
in Pilot-Automation
Study.
Sarter
Mode 1995.
Woods.
Breakdowns
Simulation N.
I).D.
II: An
Flight
Psychology,
interfaces.
4:1-28,
Knowledge-Based
N. Sarter and D.D. Woods. Pilot Interaction with Cockpit Automation I: Operational Experiences with the Flight Management System. International Journal N.
of Aviation
Sarter
and
D.D.
phenomenon.
Related
2:303--321,
1992.
Woods.
Situation
Awareness:
A critical
but
International
Journal
of Aviation
Psychology,
ill-defined
1(1):43-55,
1991.
ReportsT
K. Abbott, Lyddane,
Psychology,
S. Slotte, G. Thiel,
Tigchelaar, Flightcrews Administration,
N. and
D. Stimson,
E. BoMn,
R. Amalberti,
Sarter,
F. Fabre,
R. Helmreich
Modem Washington
Flight
and Deck
S. Hecht,
T. Newman, D. Woods.
Systems.
D. C., June
41
T. Imrich,
R. Pearson, The
Federal
18, 1996.
R. Lalley,
Interfaces Aviation
G.
H. Between