the cognitive metaphor · 2003. 1. 28. · the lucent #5ess as a cognitive system (using...
TRANSCRIPT
The
Cognitiv
e M
etap
hor
J Christ
opher
Ram
min
g
Rev
ised
1/2
7/0
3
IPTO
�s D
efin
itio
n:
A C
ognitiv
e Sys
tem
is
one
that
�
!ca
n r
easo
n,
usi
ng s
ubst
antial
am
ounts
of
appro
pri
atel
y re
pre
sente
d k
now
ledge
!ca
n lea
rnfr
om
its
exp
erie
nce
so t
hat
it
per
form
s bet
ter
tom
orr
ow
than
it
did
today
!ca
n e
xpla
initse
lf a
nd b
e to
ldw
hat
to d
o!
can b
e aw
are
of
its
ow
n c
apab
ilities
and r
efle
cton its
ow
n b
ehav
ior
!ca
n r
espond r
obust
lyto
surp
rise
Sys
tem
s th
at k
now
what
they
�re
doin
g
Ref
eren
ce:
htt
p:/
/ww
w.e
ecs.
ber
kele
y.ed
u/C
IS/B
rach
man
.ppt
Tw
o H
ypoth
eses
to E
xplo
re
!The
cognitiv
e m
etap
hor
in its
elf
is
gen
uin
ely
use
ful
!M
any
(may
be
all)
non-c
ognitiv
e sy
stem
s ca
n b
e re
-im
agin
ed a
s co
gnitiv
e sy
stem
s
IPTO
�s G
ener
ic C
ognitiv
e Sys
tem
Arc
hitec
ture
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Com
munic
atio
n(l
anguag
e,ges
ture
,im
age)
Pred
iction
,pla
nnin
g
Del
iber
ativ
e Pr
oce
sses
Ref
lect
ive
Proce
sses
Rea
ctiv
e Pr
oce
sses
Perc
eption
Act
ion
STM
Sen
sors
Eff
ecto
rs
Oth
er r
easo
nin
g
LTM
(know
ledge
bas
e)Con
cepts
Sen
tence
s
Cognitive Agent
IPTO
�s G
ener
ic C
ognitiv
e Sys
tem
Arc
hitec
ture
(An e
xposi
tory
sim
plif
ication)
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Com
puta
tion
Perc
eption
Act
ion
Sen
sors
Eff
ecto
rs
Cognitive Agent
Model
His
tory
�Non-C
ognitiv
e� V
ideo
Pro
ject
ion
Proje
ctor
Dis
pla
y Surf
ace
�Cognitiv
e� V
ideo
Pro
ject
ion
Ren
der
ing P
C(C
ognitiv
e Agen
t)
Proje
ctor
(Effec
tor)
Cam
era
(Sen
sor)
Dis
pla
y Surf
ace
The
Met
aver
sePr
oje
ct
!U
se c
alib
ration info
rmat
ion t
o p
redic
t ap
pea
rance
of
dis
pla
y in
all
cam
eras
"G
eom
etri
c"
Colo
rim
etri
c ca
libra
tion
!D
etec
tca
libra
tion e
rror
thro
ugh
com
par
ison o
f pre
dic
ted v
ersu
s ca
ptu
red im
ages
!Corr
ect
appea
rance
by
drivi
ng
reca
libra
tion s
tep
Than
ks t
o Ji
m G
riffio
en,
Univ
ersi
ty o
f Ken
tuck
y, for
mat
eria
ls s
uppor
ting
this
exa
mpl
e (f
rom
his
tal
k at
FDIS
2002)
htt
p:/
/pro
toco
ls.n
etla
b.u
ky.e
du/~
griff/t
alks
/met
aver
se/f
dis0
2/
The
Met
aver
seCooper
ative
, Sca
lable
Dis
pla
y Fr
amew
ork
Net
wor
k sw
itch
/Hub
serv
erR
ende
ring
Clie
nt P
Cs
Pro
ject
ors
Cam
eras
Dis
play
Sur
face
•In
expe
nsiv
e C
OT
S
•F
eedb
ack
Loo
p
•S
elf-
conf
igur
ing
Cal
ibra
tion (
rela
ting w
hat
is
to w
hat
should
be
usi
ng a
model
of th
e in
tended
im
age)
Cal
ibra
tion:
reac
tion t
o s
urp
rise
12
34S
urp
rise
!
12
34
Shad
ow
Rem
ova
l: re
action t
o s
urp
rise
12
34
5
Surp
rise
!
�Cognitiv
e� V
ideo
Pro
ject
ion (
addin
g a
model
of
the
conte
nt)
Ren
der
ing P
C(C
ognitiv
e Agen
t)
Proje
ctor
(Effec
tor)
Cam
era
(Sen
sor)
Dis
pla
y Surf
ace
Use
r Eye
Posi
tion
(Sen
sor)
Imm
ersi
ve,
Multip
roje
ctor
Ren
der
ing (
inco
rpora
ting t
he
use
r�s
per
spec
tive
into
a m
odel
of th
e im
age
conte
nt)
Oth
er A
cces
sible
Exa
mple
s
!Ja
va d
ebuggin
g"
Eff
ective
fault d
etec
tion/i
sola
tion w
ith
auto
mate
d m
odel
const
ruct
ion (
DID
UCE
syst
em b
y M
onic
a La
m)
!�S
mar
t� s
hip
s "
30%
red
uct
ion in m
ain
tenance
sta
ff in
USS Y
ork
tow
n e
xper
ience
(1)
!D
ynam
ic a
udit s
ubsy
stem
s"
How
Luce
nt
achie
ves
6 n
ines
of
ava
ilabili
ty w
ith its
5ESS
So
urc
es:
(1
) h
ttp
://
nra
c.on
r.n
avy.m
il/
web
space
/exec_
sum
/d
am
ag
e.h
tml
The
Luce
nt
#5ESS a
s a
Cog
nitiv
e Sys
tem
(usi
ng c
onst
rain
ts t
o m
odel
ex
pec
tations)
!Tw
o c
lass
es o
f audit m
odel
"Pro
gre
ss p
roper
ties
"Consi
sten
cy p
roper
ties
!Ben
efits
"Corr
ect
erro
rs e
ven if
def
ect
cannot
be
dis
cove
red
"Com
ple
men
tary
to
redundan
cy-b
ased
FT
"Rule
bas
e gro
ws
and
impro
ves
ove
r tim
e!
Eff
ective
nes
s"
5ESS d
egra
dat
ion
slow
dow
n o
f 45:1
(1)
"Exp
erim
enta
l re
sults
catc
h
85%
of in
ject
ed e
rrors
(2)
"O
vera
ll 5ESS a
vaila
bili
ty is
six
nin
es (
3)
AU
DIT
SU
BSYSTEM
Perc
eption
Act
ion
Sen
sors
Eff
ecto
rs
Ref
lect
ive
Proc
esse
sPro
cess
es
#5ESS
Inte
rnal
5ESS D
ata
Str
uct
ure
s
So
urc
es:
(1
) C
orr
esp
ond
en
ce w
ith
Larr
y V
ott
a(2
) A
Fra
mew
ork
for
Data
base A
ud
it a
nd
Co
ntr
ol
Flo
w C
heck
ing
for
a W
irele
ss T
ele
pho
ne N
etw
ork
C
on
tro
ller,
Bag
chi
et
al
20
01
(3
) A
RM
IS 4
3-0
1 f
ilin
gs
Nex
t st
eps:
fro
m c
om
puta
tion
to r
easo
nin
g (u
sing �
core
cog
nitio
n�)
!Sep
arat
ing log
ic fro
m fac
ts"
�le
arnin
g v
ia a
dditio
n o
f new
fac
ts"
Exa
mple
s:
exper
t sy
stem
s (M
YC
IN),
theo
rem
pro
vers
(HO
L)!
Sep
arat
ing t
he
qual
itat
ive
from
the
quan
tita
tive
"
�le
arnin
g v
ia t
rain
ing
"Exa
mple
s: B
ayes
net
s!
Usi
ng b
io-i
nsp
ired
alg
orithm
s"
�le
arnin
g v
ia a
dapta
tion,
evolu
tion
"G
enet
ic a
lgor
ithm
s, N
eura
l net
wor
ks (
Psy
napse
),
syst
ems
with e
mer
gen
t beh
avio
rKey
s: s
epar
atio
ns
of co
nce
rn;
dom
ain-
indep
enden
t al
gori
thm
s; lea
rnin
g o
ver
tim
e
Exa
mple
: B
ayes
Bel
ief
Net
The
Role
of M
odel
s an
d D
om
ain-
Spec
ific
Lan
guag
es (E
xam
ple
: F
irm
ato)
From
�Fi
rmat
o:
A N
ove
l Fi
rew
all
Managem
ent
Toolk
it�,
by
Yair
Bart
al, A
lain
Maye
r, K
obbiN
issi
m, Avi
shaiW
ool
Abst
ract
LAN
M
odel
Gen
erat
ive
DSL
Fram
ework
Spec
ific
LAN
Topolo
gy
Clo
sing t
he
Loop (
good d
ecla
rative
model
s ca
n found in D
SLs
can p
ote
ntially
be
re-u
sed a
s th
e co
re o
f a
cognitiv
e sy
stem
)
For
exam
ple
:
Perh
ap
s th
e s
am
e m
od
el
that
defi
nes
fire
wall
ru
les
can
be u
sed
to f
lag
un
exp
ect
ed
syst
em
beh
avio
r, e
.g.
att
ack
s
Res
earc
h c
hal
lenge:
what
to d
o
about
ove
rlappin
g m
odel
s?
Topolo
gy
as
expre
ssed
in c
onte
xt o
f fire
wal
l m
anag
emen
t
Topolo
gy
as
expre
ssed
in
conte
xt o
f VPR
Ns
VPR
N e
xam
ple
from
Ran
dy B
ush
and T
im G
riffin
,htt
p:/
/ww
w.r
esea
rch.a
tt.c
om/~
griffin
/info
com
2003.p
df
Rev
iew
: A R
ecip
e fo
r usi
ng t
he
Cognitiv
e M
etap
hor
1.
Giv
en a
n a
pplic
atio
n*,
iden
tify
th
e obje
ct o
f co
gnitio
n
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
*Choosi
ng t
he
applic
atio
n is
a dom
ain-s
pec
ific
iss
ue.
The
cognitiv
e m
etap
hor
cannot
hel
p w
ith t
hat
tas
k.
2.
Build
a m
odel
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Model
3.
Add s
enso
rs a
nd e
ffec
tors
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Perc
eption
Act
ion
Sen
sors
Eff
ecto
rs
Model
4.
Unify
with C
om
puta
tion,
His
tory
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Com
puta
tion
Perc
eption
Act
ion
Sen
sors
Eff
ecto
rs
Cognitive Agent
Model
His
tory
5.
Str
uct
ure
the
com
puta
tions
to
enable
lea
rnin
g
Ext
ernal
Envi
ronm
ent
Ext
ernal
Envi
ronm
ent
Com
munic
atio
n(l
anguag
e,ges
ture
,im
age)
Pred
iction
,pla
nnin
g
Del
iber
ativ
e Pr
oce
sses
Ref
lect
ive
Proce
sses
Rea
ctiv
e Pr
oce
sses
Perc
eption
Act
ion
STM
Sen
sors
Eff
ecto
rs
Oth
er r
easo
nin
g
LTM
(know
ledge
bas
e)Con
cepts
Sen
tence
s
Cognitive Agent
Deg
rees
of
�cognitio
n� Ad
ditio
nal m
appi
ng to
hum
an
men
tal m
odel
s an
d te
rms
Stan
dard
(but
goo
d) H
CI?
Expl
aini
ng/B
eing
Tol
dH
andl
ing
abst
ract
ions
of
erro
rs/a
ctiv
ities
; ha
ndlin
g en
tire
cate
gorie
sSt
anda
rd e
rror h
andl
ing?
Res
pond
ing
Rob
ustly
to s
urpr
ise
Com
puta
tion
base
d on
a m
odel
or
abs
tract
ion
of in
tent
ions
. D
omai
n-in
depe
nden
t alg
orith
ms
Any
com
puta
tion?
Rea
soni
ng
Abst
ract
ions
of h
isto
ry,
evol
utio
n of
mod
els,
kno
wle
dge
base
s, a
nd in
fere
nce
engi
nes
Any
use
of h
isto
ry?
Lear
ning
Mod
el u
sed
in c
onju
nctio
ns w
ith
feed
back
from
sel
f and
en
viro
nmen
tAn
y re
activ
e sy
stem
that
co
mpr
ises
a m
odel
and
use
s it?
Aw
aren
ess/
Ref
lect
ion
Sign
ifica
ntB
asic
/ Va
cuou
s
Ques
tions
and c
om
men
ts