LIBRARY
OF THE
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
MAY 14 1974
MASS. INST. TECil
m
1
1
74
DEWEY LIBRARY
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
A
DFSCRIPTIVF MnDFL
OF nFTAILFO PnnPLF^1-FIMnH'G BFHAVJnR
Ashok Mnlhotra
January ]37k
695-74
MASSACHUSE', ^
TUTE OF TECHNOK
50 MEMORIAL DRIVE
VIBRIDGE.
MASSACHUSETTS
02139
M/^SS. INST. TECH.
MAR
1
DEWEY
A
1
74
LIERAFv'Y
DESCRIPTIVF MOOFL
OF nFTAII.FD pnnBLFM-FIMnil'n RFHAVJOR
Asfiol<
January
1
Mnlhotra
695-74
97*1
This paper Is nreatly
improved as
P.n.VI.
It
pleasure
Keen.
is a
nui dance.
a
result of discussions v/Ith
to acknowledne
his help and
Inf ornfiat on Systems
This work was supported by Honeywell
200 Smith Street, Waltham, Mass., 02154
I
Inc.,
This
paper was produced with the assistance of the PDP-10
computer acnulred by Project MAC, M.I.T. with a nrant from the
Advanced Research Projects Anency of the Department of Defence.
\4b28
no.(o95-]4
t.1
RECEIVED
MAR 12 1974
M.
I.
T.
LIBKAKIES
ir.iTPnn'iCTinr'
Mananenpnt science vlev;s thf mananp'r a?
attempts
his performance by
to Improve
maklnn process.
adequate models of dec
model
detailed
of
i
si
shov/s
It
on-maker and
I
Detailed prob
of the
decl
to be Inextricably
si
1
en^- f
ndl nn can be
on-raki nn process
linked
v/I
a
nev/
of specific
"frame-oriented"
a
I
lack of
a
paper presents
Isolation
the
problems from qross symptoms, based on
be one component
si
asslstinn the declslon-
This
on-maki no.
probl em-f ndl nr,
of human thounht.
i
handicapped In this, however, by
Is
It
dec
a
theory
considered to
but the
model
th another component;
the
search for -^olutlons.
Consider,
example,
for
the
manaaer
discovers
v/ho
profits for the last auarter were lower than expected.
from such
described
a
find problems at
Detailed
thln!<lnn.
a
he will
v/ant
level
I
concerned
Is
Involves
dl a
means
deeper and
particular
a
with problems
of
aspiration Is
that
situations and we hope that the model
subset of managerial activity.
0720G2G
v/I
1
kind
of
are qenerally
Typically, the
very Important.
not
characteristics are, however, common to
laroe
to
more detailed level.
probl em-f ndl nn
It
usually
familiar but are neither routine nor monumental.
mananer's
Startinn
problem, which may be reconnized by any of the
by Pounds (1),
that
These
problem
a
wide ranee of
1
provide Inslnht Into
a
Further, the model, after
PAGE
dcvolopTTont
anri
validation,
desinn
l<nov.'l
edne- based
of
1
provido
1
ouidolinps
other dec!
r.I
the
for
on- r,upport- systemr.
also verify that mananers do not think in terms of
The model vill
eauations
v/i
and
3
rather
but
terms
in
"frames"
loose
of
and
that
models are useful because they assist in the creation
mananerial
and selection
of
appropriate
more
for the presentation
implications
frames.
could
This
iinve
of manacement science results
and the interaction between mananers and mananement scientists.
This paper
thinking
It
and
a
describes
a
"frame-oriented"
detailed probl em-
model of
also analy7es the prob em- f ndi
an
1
experimental
settinn
i
naper outlines
extend the
documented here.
^!FEn
v/ork
a
I
theory
this
of
human
of
based upon it.
ndi nn
behavior of
nr^
in terms
section of the
f
subjects
tv/o
in
The final
model.
proposed experimental study
to
Fnn detailed PRnnLEM-riNOIMG
More
specifically, the detailed probl em- fi ndi nn process can
be characterized as the analysis of a problem krov/n to be soluble
and for which the solution criteria are well
is
rarely
process
reauired.
is
hierarchical, i.e.
result from hinh
further
Further,
fi ndi
vi nn.
f
i
ndi no
analyzed to
analyzed
(There is considerable
and hierarchical
See,
1
overheads can be
detailed causes.
nn/probl em-sol
prob em-
profits are
low
overheads then hinh
for even more
evidence of problem decomposition
probl em-
if
Innovation
known.
detailed
the
processes
for instance,
I'ewel
1
in
and
PAGE
(2).) Thus,
•".iron
cre;^tes
stprtinn fror tbr
tr^e of problens.
n
Efich
r-rosr-
nor'r
synntons thr
nroco?-?
of thr troo I? n prolilrn
and nives rise to secondary nodes that represent the set of
courses
process
of action
stops
v/lth
alleviate the
to
the
Isolation
nross
problems
of
influenced directly by decision variables.
Thus,
symptoms, the
that
can
"
i
ntel
combined
1
i
into a
sinnle process and
process leaves the manaoer with
the situation but
the search
to
be
also
v/ith
a
plan
closely related
and,
in
for problems Is
The culmination of the
not only a better
Thus, probl em-f i ndi no
its resolution.
seen
Simon's
qence" and "desicin" phases are, therefore, seen to be
intermixed with the search for solutions.
of
be
the search for
detailed problems is influenced by available solutions.
(3)
sub-
Since the objective Is to decide
problems that contribute to ft.
upon
h
under standi no
of action ceared towards
and probl em- sol vi ne
fact,
to
form
a
are
sinnle
syndrome.
problem branch may also terminate due to the inability
A
of
the manaoer to find more detailed problems or obtain the required
In
neneral, the exact set of detailed problems at
tree will
terminate will depend not only on the situation
information.
which
a
but on the manarer's perception of It,
he possesses to analyze It
to him.
paper.
the mental models and data
and the decision variables
available
These factors will be explored in later sections of the
PAGE
5
THE STRUCTURE OF MEt'TAL MODELS
Followinn
ono
"llhon
nnd
oncountorr, n
rhanno In his
a
CO,
r'lns!<y
nnalyze situntlons
viov/ of
v/c
postulnte thnt poonlo use franns to
huild
montal
nev/
situation
reposentnt ons
(or makes
a
therr!.
substantlnl
remory
present nroblen) one selects from
a
substantial structure called
stereotype or "scenario" --
a
"Frane".
a
This Is
elaborate
a
remembered framework to be
to fit reality by channinn details as necessary.
of a Frame describe
of
I
features and relations
adapted
The top levels
believed tc be
most
Important for the proposed scenario, while lower levels are to be
filled in with the particular details of the present reality."
frame
A
Is
a
situation
—
birthday
party or a
structure
are
like
structured format
havinn an
decline In sales
several
Information Is about
accic'ent,
kinds
hov/
for represent! no
bei nn at
volume.
information.
of
a
known
Attached to the
Some
to use (or fill) the frame.
Is about v/hat can be expected next
a
children's
of
this
Some of It
and some of it Is about
what
to do If these expectations are not confirmed.
Each
frame has
Information.
a
number of terminals for attachin<-' specific
These can be thourht of as slots that can be filled
Each terminal has
by specific Instances or by data.
which
It.
data
conditions
must be obeyed by the data or the instance that Is to fill
Thus the process of fllllnn
values or Instances to each
fillinn terminals may,
hov/ever,
a
frame consists of
of Its terminals.
invoke frames
for the
assioninci
The act of
terminal
PAGE
situations
consider
Let us
chi
1
renuire further
and
assionment of
Minsky
an example.
quotes
6
terminal values.
franment
a
of
a
dren's story:
There once was a Wolf v/ho sav/ a Lamb
drinkinn at the river and wanted an excuse
to eat it.
For that purpose, even thounh he
himself was upstream, he accused the Lamh of
stirrino up the water and keepinn him from
(etc.)
drinkin<^.
If
I
v/f>re
processes
relate the
to
repesentation of this story,
a
mental
first
by v/hich
I
there are two frames associated with this
The other is
the Lamb directly.
some
The next sentence
to trick or
invalidates the real-life
story-book frame.
the
Mow,
I
try and fill
I
invoice a
realize
ruse/strateoem frame, perhaps
this is
a
children's story,
a
which
trap the Lamh,
confirms
frame and
the terminal
frame wliich asks for the ruse or stratenem employed.
I
One
kills and eats
the story-book frame in
strataoem is required
ruse or
In my
situation.
the real-life frame in which the Wolf catches,
is
form
would say that readi no the
invokes the situation "Wolf wants to eat Lamb",
sentence
mind,
I
of this
To do this
very simple one since
and
try and
fill
its
terminals with the specifics of the ruse.
Ultimately,
frames and
default)
I
terminal
created
and
have analyzed the
will
values (some
a
mental
"understandl no" of the story and
about the story.
will
fit
into
a
In fact,
of
which may
model.
I
story Into
This
be
a
set of
filled
model
is
by
my
can use it to ansv/er auestions
the frames and terminals of this model
more oeneral structure of frames and terminals in
PAGF
my
minrl
—
my model of the world
to ansv/er
cuestlons about
contained
In
sallvatinn?"
anc'
the story
Such
It.
—
a
I
v/I
1
be able to vise it
1
beyond the
that no
question
may
7
facts
the
"Was
be
V.'ol
f
receive an affirmative ansv/er because a
and it v/Ml
default terminal value of the desi re-to-eat frame Is
salivation.
An interest! no confirmation of frame-oriented thinkinn comes
from
an experiment described by Mosher
children were asked to play
a
and Hornsby (5)
In which
variation of "twenty ouestions" and
determine the cause of an accident -- "A man Is drivino down
roa<^
In his
car,
noes off
the car
the road and
hits
a
the
tree."
Asked to describe his "system" for nettina the answers en eleven-
year-old responded!
—
liell, to eliminate bin thinos quickly
like
was there anythlno wronr with the road
v/as there
was there
anythinn wronn with the v/eather
was there anythlno
anythinn wronn v/ith the car
wronn with the person
if there's somethino v/ronn
with the person, you start from the bottom and <-o
to the top.
oroup like all the thinns with weather,
I
breaklnn (sic), then I rroup them smaller and
smaller till I net to the point.
—
It
—
seems clear that the description of the accident causes
invoke an analysis frame that
the
—
—
road, the weather,
iiim
to
contains four potential causes
the car and
the person.
explore each of these until he finds the applicable
He intends to
one.
Then,
he plans to Investlqate it further.
The
top
level
dlanrammed as below.
has
frames for
of
his
It
analysis frame
may,
therefore, be
Is reasonable to postulate that he also
each of the
terminal causes which
allow him to
PAHF
carry
tlie
analysis further.
ACCIDENT
(Go! no off the road)
I
ROAn
PROBLEM
Thus,
I
I
WEATHER
PROBLEM
CAR
the potential
causes in the frame are
branches of the problem tree.
is built up
by invokine
PERSON
PROBLEM
PROBLEM
user! to
form the
The next level of the problem tree
and fillinn frames
for the
applicable
probl ems.
PROCESSES IN SITUATION ANALYSIS
A^'O
COMPREHENSION
The above examples demonstrate the follow! nn processes in
situation comprehension*
1.
Invocation of frames from named concepts.
2.
Selection betv/een competinn frames and frame validation.
3.
Assionment of terminals.
A
more complete model of situation comprehension can be
described in terms of these processes as follows*
A.
B.
Invoke frames.
Select between competinn frames and validate
the selected frame.
C.
Attempt to fill each terminal of selected frame.
This may require the invocation and fillinn of frames
I.o.
steps A. to F. This make? the process rer.iirslve.
termirals filled?
Are all
P.
so,
If
proceed to
Wave all
E.
If so,
not,
If
E.
use characteristics of unfilled terminals to A.
If not,
important facts been considered?
proceed to
F.
use unaccounted facts to A.
Fit frames tonether to create an internal
F.
representation of the problem.
implies the capability of Invokinn new frames
above model
The
from niven
that differ
characteristics.
This
mananer's
mental
frames.
may
exist
In
limitations,
incrementally
a
In
from
old
frames
reasons
to specified
mechanism that
feedback
cases,
mind
or
terminal
structure built from
a
certain
mananer's
the
hi
the
appropriate frames
within
and,
certain
Often, these are created
erarachl ca
1
1
from
y
more
frames.
are one type of mental construct that may arise from
concept such as "profit".
specific
In a
nev frames may be created.
powerful, neneral
Frames
used
Is
situation anai nst
matches the niven
not
frames accordinn
for low
relations that say "Profit
are frames that analyze
profit and
and hinh
situations.
There
In
addition,
probably others
there
are
for more
cause-effect
increases If revenues Increase."
"Profit decreases if costs Increase." etc.
the
and
Basically, frames are
PAGE 10
for
Mror'
r,i
tuition conprehensi rn;
"catonor
cnlls
(6)
zl nn"
complox form of
a
Once
,
representation created.
Internal
this Is
accomp
1
I
v.'bpt
Prunrr
and
shor'
an
dictates the Implications and
It
relations to he used in drawino conclusions and
caiis'^-ef feet
the
I
the search for action alternatives.
The master "profit" concept
= Pvcvenue
"Profit
difficult to
creation of
-
with
work
nev/
directly.
relationship
the hacknround
utility
Its
loose evaluative
and seems
lies
frames and cause-effect relations.
low-profit, have
such as
that contains the
exists In
Cost"
in
the
Some frames,
judnements, such
as
"bad", associated with them.
PnnCFSSFS in DETAILFD PRnBLFM-FI^'^If'^
From the neneral model described above
processes
neneral
that
be
used In
e. n.
description frame for
a
his hobbles,
his
v/e
can soeclalize the
detailed probl em-
situation comprehension frames can
structures
Job,
will
aae and his
take on
f
a
I
ndl nn.
In
variety
of
person may contain his
a
net worth as
terminals.
detailed probl em- f ndl nr the analysis frames consist of
I
s».ib-probl
evi
terminals each of v/hlch may
a
list of
or may not exist in the
niven situation.
A.
The mananer will
check whether
for the problem (sub-problem)
directly.
If
so,
he will
a
decision frame exists
I.e.
In
if
it can be solved
attempt to asslon terminals
to it. These may be Inputs to a decision rule.
n
PAGE
If
B.
the nrohlem branch
this Is successful
If a
v/|
1
1
torrri Pr-ito.
decision frare does not exist or cannot be filled
adeauately he will
Invoke one or n^ore analysis frames for
the problem (sub-problem).
C.
Valldatinr nuestlons
v/I
1
1
he asked to eliminate seme of
the analysis frames. Typically,
D.
these are yes/no cuestions,
Each analysis frame will contain potential sub-problems as
terminals. The mananer will attempt to fill these terminals
i.e.
test each sub-problem to see if it exists. Each
sub-problem that exists will be analyzed further startinn
at A.
Competing analysis frames may also be eliminated at
this stane.
E.
In rare cases
alternative branches of the problem tree will
be created and alternative analysis frames pursued further
and eliminated on the basis of information obtained at a
lower level.
Thus, each problem is either attacked directly by
a
decision
frame or analyzed into sub-problems usirn its analysis frame.
practice, there
problem
branches
done to nather
have
special
bear on
Is
—
another
Imnortant method
they may be
more Information
knowledne I.e.
the problem.
of
held in abeyance.
or to consult
brin'-'
Sub-problems
his,
more
may also
abandoned in favor of more promlsine branches.
the problem tree may end In a "wait" state.
dealinn
This may be
someone
refined,
be
In
with
v/ho
may
frames to
(temporarily)
Thus, branches of
Finally, because of
PAGE 12
connltfvf limitations,
sub-probl
Wp
may forrot or
rpannror
tino
i
certnin
nnoro
and/or terminals.
eiis
have
Simon's (3),
discussed
advocated by Here (7).
this model
hovy
fact,
In
it
closer
is
does
Gore's model has
accord with
not
heuristic
to the
model
much broader scope but
a
early phases of frame invocation and validation in the above
the
model
seem to correspond to
Set"
and the later
and
the
analysis of problems
phases of hierarchical
selection
decisions
of
Orientation
his "Development of the
development
the
to
of
the
"Evaluation Set".
The
remainder of the paper
fits
model
These
findinn.
nuestions
be
a
knov/1
protocols
v/ere
be used to
protocols,
analyze how
1
taken to
v/e
1
1
the above
detailed
net some
They also provide, however,
processes and we
I
support-systems.
1
problem-
Idea of the
edoe-based manaoement Information system
able to answer.
information on problem-f ndl no
can
v/I
protocols of
actual experimental
a
sliould
nreat deal of
hope that
this
of knov/1 edne-based decision-
Improve the deslon
however, that, unlike
We must remember,
not a one-shot process.
real-life probl em-f ndl rn is
I
Typically, the manaoer will eo throunh
a
these
preliminary session
and
repeat the process anain uslnr- more detailed frames some of
then
which may
have been
Iterative nature
This
exper ment.
I
created as
result
a
of probl em-
f
I
ndl nn
of earlier
Is
analyses.
underplayed In the
PAGE 13
VAi.inATinr' rr
Appondix
T^!^^
1
modfl
presents
sconnrlo
the
wore n^kec! to arrive at
Mnnrtners
for this problem situation v/Ith
a
of
problem
a
the (simulated) assistance of a
computer system that would ansv/er their questions.
panes
present
'essions.
and
Finures
analyze
1.
tv/o
and 2.
settlnr.
recommended course of action
protocols
of
The foil owl no
prob em1
f
I
ndi
n*-"
present the frame structures of
the subjects as evidenced by the nrotocols.
Subject
1
Subject
1
has an M.P.A.
from
a
presti pious school
years of manarement experience. Me starts off v/ith
and about five
a
clear idea
of how to analyze the situation, but runs into some trouble v/hen
he finds that the system does not have the data he needs to fill
the first terminal. He then proceeds to modify his frame to use
the data available. Once this is done, he is able to fill
terminals and analyze the problem nuite rapidly.
0.
1*
I'd like to know the break-even ouantity for each unit
and are we manuf acturi nn sufficiently in excess of it?
A.
i:
I'e
hi=5ve
direct cost for each unit, overheads for plants
and head office.
Hov/
would you like
ire
to calculate
break-even quantities?
The subject seems to plunoe rioht into his lov/-profit frame and
PAGE
since there Is no
v/ay
to attack lov/-profIt directly.
Is no decision frame for
it,
U
there
I.e.
he asks a generalized ouestlon to
determine If the sub-problem that makes up the first terminal of
the analysis frame exists. The answer he receves makes him
suspect that the frame may le inappropriate and so he asks
a
nuestlon to validate It.
product made on
0.
2'
Is each
A.
2'
Continuous process.
a
continuous or
The ansv/er convinces him that his neneral
a
batch process?
frame Is correct but
the productlon-belov/-break-even-volume terminal which he was
attemptinn to test in question
Is
1.
Inappropriate. He has,
therefore, to use more neneral concepts on cost and production to
obtain
a
suitable replacement and check if production is laroe
enounh. He finds this In
a
low-prof t-marn| n terminal
he proceeds to validate and fill
I
a
for which
frame Immediately and
a
hlnh-
overhead-cost terminal which he tackles later.
Q.
3*
Can you o|ve me
A.
3*
All
a
plot of fixed cost vs. variable cost
for each product?
fixed costs are In plant overhead. All variable
costs are attributed to products.
0.
h'
Do you have marain of profit vs. product for most
recent history?
for the last five years,
A.
k'
Yes,
Q.
5*
How Is profit marqin calculated?
)
PAGE 15
A.
51
Contri
hiiti
on margin is crlculntr?fJ as list nricr
If^ss
direct cost if the plant Is specifier'. List price
less averane cost
0.
6:
If plant
not specified.
is
Can you compute profit marnin vs. quantity for each
product?
A.
6:
Contribution marcilns are constant with quantity.
(Explanation of contribution marrin and overhead
costi no.
Q.
7*
A.
7*
Can you tell me actual maroln with quoted prices
for each product?
do you want averane quotation prices?
Yes,
Questions
3,
't,
5, 6,
valldatinn the
1
and
are concerned with creatinn and
7
ow-prof t-mar nl
I
n
frame.
the distribution of sales vs. size of order?
8:
h'hat
A.
8:
Do you mean vs. nuotatlon otjantlty?
0.
9:
Yes
A.
9«
(A pareto
0.
is
distribution is presented with
customers
Question
8
Is
a
beln<-'
a
few
responsible for most of the sales.)
valldatinn nuestlon
.
Mote that the answer he
receives is in fact the answer to another question -- "VJhat Is
the distribution of sales by customer?" -- but since the shape of
the distribution Is as expected he validates his frame and roes
ahead!
6
PAGE
n.
10:
V.'hat
A.
10:
Unit
1
the mar'-'In iisln^ the price of niiotatlon?
Is
for the lareest five customers?
101
Marein
0.
11:
11
:
103
4.80
3.75
104
105
10.50
6.38
What Is the contribution marnin for each product
usi nn
A.
$5.20
102
1
I
st prl ces?
Unit
101
Maroln
$6.00
Ouestlons 10 and
11
102
103
5.50
4.00
104
105
12.00
7.00
attack another terminal of the low-profit
frame I.e. "Are ouotatlon prices too lov/?". Satisfied that this
problem Is not serious the subject nroceeds to the testinn of
another sub-problem terminal In question 12.
0.
12:
Do you have cost data on actual cost vs.
A.
12:
Yes.
Question 13 Is
a
budqet cost?
nenerallzed ouestlon, similar to question
deslnned to eliminate
a
1,
Unfortunately, the system
sub-problem.
cannot answer It.
Q.
13*
Have the maml ns been maintained?
A.
13!
I
don't know what you mean by "maintained".
Ouestlons 14 and 15 are specific questions deslnned to test the
hi
nh-manuf actur nn-cost and
i
1
ow-prof t-marql
I
n
sub-problems.
answers Indicate that these sub-problems are not operative.
0.
14:
Can you calculate the difference between actual
The
7
PAGE
anr'
A.
0,
1't:
151
hiidr-etpc'
cost for erch product?
102
103
Actual $31.00
28.15
Budnet $30.00
27.00
Unit
1
101
1
105
O't
h3.k^
2'^.00
'i7.25
'M.ro
23.00
't7.00
What Is the percentane deviation of budneted to
actual marnln?
A.
15:
Unit
101
Deviation
102
103
]0h
^i%
3%
h%
3%
—
Question 16 attacks another sub-problem
The ansv/er indicates that this Is
a
hi
105
0.5%
nh-overheac'-cost.
problem and the subject holds
temporarily in abeyance.
It
0.
16:
A.
16:
Can you provide data on actual plant overheads and
deviations from budeet?
Plant
1
1.0.
1
2
3
30%
27%
15%
'f
Actual
Deviation
23%
19.8%
The percentane deviations v/ere provided before the actual
overheads and the subject proceeded to the next question before
the latter were completed!
Question 17 eliminates
hi
a
sub-problem that
ah-manufactur no-cost frame. Since
i
been eliminated as
a
hi
Is
a
terminal of the
rh-manuf actur nn-cost has
I
problem by the ansv/er to question 14 there
was no need to ask this question.
In fact.
It
seems to have just
PAGE
occurerl to the subject thnt the cost of
sfnnTflcpnt factor in the prof tahi
i
1
T
1
lead must he p
ty of the company and he
asks the question without placinc the sub-problem in its proper
frame. Looked at another
v/ay,
problem seems to have elevated
late realization of this
the
to the hir-hest operative frame.
it
0.
17*
How have my lead costs fluctuated over the year?
A.
17*
Month
jan feb mar apr may jun Jul
$/Ton
.16 .17 .17 .18 .20 .19 .19 .20 .19
aun sep oct nov dec
.
1
<S
.17 .18
Questions 18 and 19 investir-ate sub-problem terminals of the
hi
nh-overhead-cost frame.
0.
18:
How many days inventory do we have on hand?
A.
18:
33
0.
19:
Who pays transportation cost for lead and for batteries?
A.
19:
We do.
days.
After auestion 19 has been ansv/ered, the subject says he knov/s
enounh about the situation but immediately realizes that he has
missed
a
terminal of the lov/-profit framie. He asks
a
final
question and eliminates it.
Q.
20:
A.
20:
12
What percentane of plant capacity Is beinn utilized?
Plant
Utilization
90%
95%
3
4
98%
89%
PAGE 19
The sunnested nctlon pifn frllov/s directly,
thoLinlit.
thout nnprrent
He feels quotation prices nay be low and wants to
invest! eate whether they coulr) he raised.
fill
v/I
Me feels incompetent to
the low-puotation-pri ce frame and decides to seek expert
help. He wants to investinate this further. He muses about the
hi
qh-mananement-sal ar es terminal of
I
not investinate this terminal.
hi
eh-overhead-cost. He did
Perhaps he felt he knew the answer
or perhaps it did not occur to him earlier.
[^AGF
LOli-PRCFIT
T
I
I
LOWPRODUCT I ON'
I
HIGH-
LOWPRICE
MEG. -COST
I
I
I
NO
I
NO
I
LOWPROFITnARGIN
HIG'I-
LEAD-CnST
I
UP
I
I
NO
YES
I
HIGHnVERHEADCOST
YES
I
I
WAIT
I
I
EXCESSINVENTORY
I
InvestI oate
If prices
can be
ral sed
I
I
NO
I
I
HIGH-
TRANSPORTATION
COST
LOVi-
CAPACITY
UTILIZATION
I
I
YES
NO
I
WAIT
I
InvestI pate
further
COMMENTS
1.
2.
HIGH-LEAD-COST should he a ternlnal of
HIGH-MANUFACTIIRING-COST as dl sussed In the text.
LOW-CAPACITY-UTILIZATION occurs twice under
different nulsesi once as LOW-PRODUCTION and once under Its own
name. The hi nher terminal may be a vestlqe of the
BELOW-BREAK-EVEN terminal that wps re.p}acec\ due
to data not beinn available.
FIGURE
1
FRAME STRUCTURE FOR SUBJECT
1
20
PAGE
Subject
1
Subject
2
v.'as
also has an TI.B.A. and
experience. He asked
v/Ith.
tv/o
years of rnanacenent
neneral, validatinn questions to start
tv/o
Then he not dov/n to business and attacked the first
terminal of the low-profit frame
plannino time.
—
Insufficient production
This Is tested and eliminated In questions
3
to 7
as not belne significant.
are the five basic battery types?
0.
II
I'.'hat
A.
i:
They are different products vlth differences In
0.
2:
V.'ho
A.
2:
Sears.
0.
3-
Hov/
performance and reliability.
Is our maj'or
customer in the U.S.?
lonn in advance do customers place their orders v/Ith
us?
A.
3*
0,
h'
don't understand what you mean by "advance".
I
How lonn is the Interval betv/een the placinn of
a
ouotatlon and delivery to the customer?
A.
h'
VIould you
f
Q.
5:
21
T
rst del
like the averaqe time between quotation and
i
very?
Yes.
A.
5-
One month.
Q.
6:
How lonn Is this Interval for our lamest customer?
A.
6:
Three months.
Q.
7-
Is
this the averane Interval over the last five years?
PAGE 22
A.
7:
Yes,
Ouostlon
Is.
disposes of nnother ruh-problem terminnl. Anrv/nrs to
B
nLiestions
it
9
to
1
1
Tnrllcato t!iat hi nh-transpcrtatl on-cost may be a
problon. The subject seems to feel that the problem can be
attacked directly throunh
a
decision frame and does not Invoke an
analysis frame,
How often do branches order from
plant not closest
0.
8:
A.
S'
2% of orders no to a plant not tied to a branch.
0.
9*
lihat
A.
'1:
0.
in:
no you have hudoeted transportation cost?
A.
10:
Yes, we do.
0.
11-
Vlhat
A.
11:
a
to It?
percontaoe of product cost Is allocated to
transportati on?
21"'
Is
the variance of between budgeted and actual
transportation cost?
30% over budnet for 1973
Questions 12 to 14 Investlnate the relationship between
Inventories, sales and production and their ansv/ers seen to
Indicate that this
0.
12:
Is
not a problem.
Do you have a formula for Inventory and production
cycle rules?
A.
12:
We do not have that data.
PAGE 23
n.
13:
A.
13'
VJhat
has bron our customor service level
rjurlnn the
past year?
Sorry,
I
don't know what yon mean by "customer
service level".
0.
^h'
What percentane of customer orders have been shipped
from stock?
A.
14:
65%
Ouestion 15 investlnates another sub-problem terminal and its
answer indicates that overhead costs are
nuestlons that follow
I
nvesti
n?itf»
tlie
hi
a
problem. The five
nh-overhead-cost analy;
frame and assinn terminals.
n.
15:
Ry how much did overhead expenses increase over 1972?
A.
15:
35%
0.
16:
What percentane of sales comprised overhead expenses?
A.
16:
Sorry,
0.
17:
What was our contribution margin in 1973?
A.
17:
I
don't understand the question.
For which products would you like the contribution
marol n?
0.
18:
For the whole line.
A,
18:
Unit
101
Budneted Marnin $6.00
102
103
5.50
h,00
1
0^i
12.00
Q.
19:
What were the overhead expenses for each product?
A.
19:
I
Q.
20:
What were these expenses by plant?
have overheads by plant and head office.
105
7.00
PAGE 2h
A.
20:
Plant
II.
Overhead
0.
12
3
4
$
The subject did not wait for nuestion 20 to be answered but
proceeded to auestions
products have too low
This
I
21
a
and 22 to Investloate v/hether sono
turnover and may profitably be dronned.
nvesti nati on proved abortive.
'
0.
21
A.
21:
was the percentane of each product as
VJhat
a
of total
sales?
Unit
101
102
103
104
105
Sales
10/'
50%
20%
10%
10%
0.
22:
For 1972?
A.
22:
Unit
101
102
103
104
105
Sales
15%
35%
10%
25%
2
The followinn four Questions, similarly,
percentane
5%
investlnate whether
Inventories are too hinh. This too proves not to be the case.
0.
23:
Do you have any Information on the balance sheet?
A.
23:
^'o,
0.
24:
Vihat
A.
24:
30%
0.
25:
Can you break this up by product?
A.
25:
don't have the balance sheet.
I
percentane of my assets are Inventory?
Unit
Assets
101
102
103
104
1%
5%
10%
8%
105
6%
PAGF ?5
Tho stibjpct snys he's
niipstlon thnt
Sf^r-nis
finl''>hnfl,
to fill
the
hut noos on to rsk
Inr.t
p
final
suh-problem terminnl of the
lov/-profit frame.
0.
26:
What was the turnover hy product?
A.
26:
Unit
101
102
Turnover
10.3
12.7
103
^.3
1
0'l
8.4
105
5.2
The subject conclurles, havlnn run out of frames and terrlnals.
Aealn, his policy recommendations follov/ instantaneously
—
he
would like to control transportation costs and overhead, perhaps
also channe the product mix and control Inventories.
The problem
analysis Is based upon sub-problems selected from amonn those
supplied by the analysis frames v/hich are percleved to exist In
this particular situation and for whom decision frames exist and
can be fill ed.
PAGE
LPlJ-PrnFIT
I
I
26
Pf.nE
?7
COMMFNT? ON THF PROTnCOI.S
1.
The structure of the frames sneri to be very
the
tv/o
frame.
Subject
2
has a much
it
1
In particular,
Subject
1.
contribution to
hi
terminal.
a
ffprert for
subjects. This Is strlhlnnly true of the lov-proflt
terminals. Subj'ect
Note,
di
v/i
der frame with specific
has fewer, more oeneral
terminals.
the position of the excess- nvrntory
i
analyzes excess- nventorl es as
i
nh-overlieaH-cost while Subject feels
contributes directly to low-profit. (He may be an
inventory control expert
!)
This illustrates
a
fundamental
difference in their underlyinn conceptual structures.
2.
Certain terminal values seem to be filled by default and
subjects do not seem to bother to ask questions to assl'-n
them.
For example, they know that manaoement salaries no up
every year.
3.
As the subj'ects were asked to assume they were askinn
cuestions of
a
computer system they may have tended to ask
more specific questions then they
v/ou
1
d a
human belne. Also,
some of their neneral questions could not be answered so they
were forced to ask specific questions Instead.
4.
The protocols Indicate that mananers ask neneral questions
(uslnn v/ords
eliminate
a
like "sufficient" and "maintained")
sub-problem.
If
It
to try and
cannot be eliminated they
ask more specific questions to asslen terminal values.
PAGE
5.
Most of the questions thnt atterrpt to olimlnnto
sub-problem terminal
asl;
for
r\
piece of data In relation to
a
norm, plan, standard or history. Often the re atl
1
onslil p
is a
loose one (expressed by words like "sufficient" and
"maintained") such as v/hether the
tv/o
numbers are of about
the same mannitude or whether one is nroater or lesser than
the other. This corresponds to the fact that the terminals
are named "low-profit", "hi eh- nventor es", etc.
1
1
In fact,
subject
1
disregarded part of the answer to question 16 and
subject
2
disrenarded the ansv/er to question 20 since these
were expected to provide absolute numbers. The few sinnle
numbers that seem to be
a
6.
si
nnif leant,
such as sales, invoke
maqnitude frame, such as "small-company".
We notice that the processes of validation,
frame creation
and terminal assienment do not take place in "lonical"
sequence.
level
In fact,
terminals are left hanel ne as hieher
terminals are filled, frames created, etc.
This suoqests that all elements of
retrieved at the same time, but
a
seem,
frame are not
to occur in some order
of importance. Further, the subjects seem to have special
processes that brine in the next terminal and check if all
terminals have been filled and if all the qlven facts are
accounted for.
7.
Since this
v/as
28
the first pass at the problem
a
number of
branches tended to be left in the "wait" state due to
a
PAGE 29
insufficient Information or bocruse the subject felt
to be less important relative to other ''ranclies.
thein
expect
V.'e
that this will be found to be typical of early
probl e!"-f
I
nr!I
nn attempts.
COMCLUSIOMS AMD OTHER OBSERVATinMS
Thus,
v/e
find
that the
described above fits the protocols
preliminary
however
test
and
detailed prob em-
of
POcJel
1
must
model
development and testinn before it can be useful
I
ndi ne
This is only
aulte v/ell.
the
f
a
further
av/ait
In the c'esirn
cf
decision-support systems.
Val
i
da 1 no
1
conclusion does
Quest ons One
I
emerne, hov'ever.
Most yes/no ouestions are asked to validate frames.
In
they are followed
the affirmative
data to
analysis
eliminate
a
stib-problem
Thus,
frame.
v/henever
ansv/ered In the affirmative,
supply some
analysis
of
marnin by
a
the Information
If
answered
askir'-'
terminals
asslnn
an
in
nuestlon has
yes/no
for
to be
that
relavent
may be
to
the
In some cases this can be
For example, the question "Do you have profit
product?"
for all
question
a
the support system should attempt to
frame that will be invoked.
done nulte easily.
marelns
or
by
answered
should be
products
by
a
durinn the last year
listlnr
of
the
rather than by
a
mere "Yes, we do".
General ty of
the Model
I
detailed
lie
maintain
probl em-f ndl no Is very similar
i
that the
process
of
from person to person.
PAGE 30
ni
fforonces
frames
structure of
f ferences
f
I
nrll
the
for
situation
at
determine the importance
will
dictate
hov/
orientation
The
a
process
capacities
rieht to the heart of the matter".
will
for
naive
users
may,
and v/ould,
therefore,
be
This may account for the expert's ability to "net
much nulcker.
alv/ays
problem due to
I
be worse;
Thus,
he may
nadenuate frames
anc'
If
the performance of the
be unable
to analyze the
them
he tries to clevelop
may strain his connltlve capacity and deorade his performance
he
other tasks.
Yes -llo Quest ons
I
seem
to
be
and ScannI nn
associated
with what
stratenles I.e., the testine of
In
Tee
considerable amourt of frame creation v/hile
reqiiire only validation
an expert may
n
connltlve
individual
dere<-!.
5?omeone with less experience may have only fifteen
Involve
therefore.
I
he consi
1
twenty.
novice
ftructurrl
many frames and terminals will be "over oolced"
situation.
or
hand.
<3nci
expert may have 1000 detailed frames for every aspect of
An
a
Further,
above.
1.
In the number
of various factors In
the analysis and the order in which they will
comment
pvd
stylo
nr'/probl em- sol vl nn
can ho attrl'^utor' to differences
of fecti vonnss
f'i
probl om-
In
mind.
the buildine
StratepI es Yes- no
Bruner
of frames
from
called scanninn
hypotheses that the subject
The converse, focussine, would
up
(6)
ouestlons
observed
has
seem to correspond to
properties.
Eruner
recoonlzes that "the task of search Imposed upon the user of ...
focussinn may become rather severe" and this Is even more true In
PAGE
sitiintion analysis
than Tn
nritnor
provldo us with
crncort
r.
pttnlnmorl
t-asl;r-.
alternate hypotheses to he scanned
Tims,
framf^s
and
seen to create frames only v/hen absolutely necessary.
v/e
simplest level,
Prohl etn-Fi ndi nn Sty! e Pn the
decide
I
to
fill
all
nvesti nati nn the next
sub-probl
branches.
a_
"^ome
i
(6)
a
or he
a
problem
maraner
may investigate
I
of these decisions
v/I
nvesti eates
found
that
by
the
1
be taken accordlnn to his
1
the importance
personality.
different
subjects
his
each
promlslrn
more
of various
sub-nroblems
others may be functions cf personality variables.
employed
may
frame before
Me may decide to leave some brrnches
while
lie
evaluation of
pr ori
while
state
"v/alt"
of
level
lov/er
as It arises.
ei-'
the
In
terminals
the
31
v/ere
concept
attainment
consistent
Cruner
stratenies
features
This may also be true of probl em-
f
i
of their
ndl nr/prob er1
solvlnn stratenies.
FURTHER EXPERIMENTAL INVESTIGATION
An
experimental
the model and
I
study is beinn
planned to further validate
nvesti nate some properties of the frame- structures
and the probl em-f ndi nn styles of different mananers.
I
of
subjects
v/Ill
be asked to
analyze
a
problem
A
number
scenarl-"!
similar
to the one presented In the Apnendlx under similar circumstances.
In addition,
each subject may be asked to take some psychol onl cal
tests, such as Kelly's Role
personal nuestlons.
Repertory Test (8), and answer
some
PAGE 32
The protocols voulri thon bo .inrlyzrri to Hetrrnl no tho frnmo-
and sono InHlcntorr. of prohl om-
structuror.
Val legation of tho modol
suhjoct.
knowl pdro-basod
of
ftinction
1
n<-li
i
si
no stylo for onch
unclorscoro tho utility of
of frame-validation and
support the processes
The
1
I
answor ri'estlons or
systems that
dec
v'l
f
on-sunport- systems
in othor vays
term! nal-
f
i
1
1
I
no.
problem analysis
In
could then be defined as the provldinn rf information tc
support
described in this paper as
the
comnonent processes of the model
v/ell
as the relief of strain on conrltlve capacity and short-term
memory.
also lead
The study may
under standi ne
Potter
process
doslon of
of
v/ould,
knov/1
of
cotJrse,
to
tl-e
elaboration
the
proh om- sol
lead
to improved
edoe-based decl
1
vi
model.
of the
pn/prob om1
fl
ndi no
ouldellnos for the
si on-support- rystems.
Studies of a larne number of frame- structures may show
manaoers
have only
profit".
incorporated Into
exploratory
a
this
If
the
few frames for
is
true,
l(nov/l
the
edoe-based
common nroblems like "low-
frame-structures
system and,
after
can
a
be
few
auestlons, the system will be able to anticipate the
manaoer's probl em- f ndi no process and thereby be of much
I
assistance.
that
It
will,
frames or terminals.
oreater
for example,
not allow him to overlook any
In addition,
by rnticipatlno reaiiests
the
system may be able to Increase Its own efficiency.
The
and
experimental data
probl em-f ndi no
I
styles
may also
are
shov/
that frame- structures
correlated
v/ith
personality
PAGE 33
detrrmi rantsbacI<oroiind.
bo
This
thoy
v/ou1c'
tailored to Individual
concoptiin]
constructr.
allow
I
rjocl s
or
pc'iicnt
on-siinport- systems
i
to
onr
1
hp.
styles and to different types of rnananers.
PAGE 3^
APPFf'nix
A pRnBirr-' scrpAPin rnn
rioho Mnlon is an rstnM
with
I
head offices locator! in
where the actual manuf acturi no
out over the continental
itsttmg the proposed mgdel
battery types for
tho ri d-v;est.
has four nlpnts
It
carried out.
is
leaH hattorlrs
These are '-nread
United States.
fifteen variations
various nurposes.
battery is identified by
Globe
rrrntifncttiror of
sliori
Union manufactures
Globe
1
Each
of five basic
distinct variety
bulk to twenty Tiajor customers
Union sells mairly in
located all over the United States.
Customers place lone
"nuotatlons"
specified
certain
Globe
v/ith
Union for
Globe
number.
unit
Union
quantities
supplies
anainst
Quotations on the receipt of orders from customer branches.
branch is expected to order from
closest to
it.
branches in
a
In neneral,
is
heavy
of
of a
these
Each
certain plant, usually the one
a
riven plant
customer
supplies
facilities
to
thie
types of units it sunnlies.
transportation can
and
proportion of product cost.
lack
a
ranr'e
set of states surround! no it.
plant manufactures all
Each
The product
of
unit nunber.
a
make up
a
lar<-e
Only In rare cases of shortages and
manufacture
a
specialized
tini t
v/I
1 1
batteries be supplied from other than the closest nlant.
Plants
manufacture
accordinn
to
certain
inventory
and
PAGi
production
Thoy
cycle riilos.
Per f ormnnco
riirrct costs ano' ovorhrnr's.
as
service are
customer
iMK'nofr on
t^nrlnot
pn.-iln'-.t
nair criteria
the
iiio^t
.t-.
vol)
plant mananer
-for
Plants are not run as profit centers because prices
evaluation.
on
rrr oxpoctcc' to
are nenotiated
nuotations
hy
the head
even thoueli
office
standard price lists exist.
It
a
February 1974 and as President of Globe Union you
is
little concerned
Despite
received.
at the results
a
you have just
for 1973 that
increase In sales
20%
are
over 1972,
profits
decreased by 1%
You
combination of
decrease
In
decrease in
that the
feel
causes*
three
contribution
price and direct
cost) or
increase
marel ns
a
profit cou
In
(difference
,
profitable units.
You v/ould like to Investiqato
decreased
usinn
profit
strict
control on
list prices
ouotatlon
which
the pricinn
purpose of
As
cl
Is
to
1
1
take
a
to
a
serve
cost
to a
expenses,
botv/een sell Inn
mix tovard
th<-
less
cause of the
Information
System.
decision tr
enforce
of quotations,
or Introduce
this exercise
appropriate under the
vi
supposed
are
pricinn,
Union
Globe
the
be due
r!
overhead
channe in product
Dependinn on what you find, you
1
revlev/ and reset
as
control
determine which
nuidelines
for
nronram.
The
decisions
are
rcurstances.
sales erowth has been very
healthy, you are Inclined to
disrenard competitive actions In your analysis.
You also assume
that the cost and other data contained In the system Is accurate.
PAGE 36
Tlio
follovMnn two panes contain
Union Information System.
a
description of the
G
1
oho
PAGI
Tho
fo11ov/Tnn two pnnos contain
Union Information System.
n
description of tho Hloho
PAGE 37
APPFMDIX
(Contlnupfl)
1
GLOBE UN inn inrnRMATinr' system
The Globe Union Informntion
prices and
costs,
durinn the last five years.
posed to It In simple Ennl
functions
and
Systorn contains
other Indicators of
I
It
Is
sh about
price for
unit 103".
answerinn
ouestlons about Itself,
Items
contains,
It
capable of answerinn riuestlons
the contents of the
contents such
of these
In
as "profit"
addition, the
explain
the
snins,
on
clatci
Globe Union's operations
I.e.
system Is
database
or "averare
capable
of
enumerate the data
It can
procedures
embedded
In
tf.e
functions, etc.
The system can be nuerled much as one
to
ouestlons,
ansv/er
appropriate responses
cannot reply to.
O:
V.'hat
A*
I
prepare reports,
to
A typical
requests
dialon
It
v/I
v.'ou
1
d
etc.
does
use an assistant
It
not
v/Ill
provide
understand
th the system nay be*
data do you have reeardi nr unit costs?
have actual and budneted costs for each unit at each
pi
ant.
O:
What was the cost of unit 103 in plant 4?
a:
$78.23
or
PAGF 38
was the
pricr for nnl
o:
l.'hnt
a:
$81.00
O:
Do yon have a model
a:
Yes.
o:
llov/
a:
It
1
I
r.t
for cont ri
t
»^ut
103?
i
on marol n?
does it work?
comnutos the list (standard) price minus actual cost
for the niven unit.
n:
What was the contribution for unit 113 at plant 2?
a:
$9.20
Was the contribution for unit 81?
O:
I'hat
a:
$0.30
o:
What was the avreane cost of unit 81?
A*
Sorry,
O:
W'hat
a:
$78.67
O:
What
a:
$76.00
I
don't
knov/
the word "avreane".
was the averac^e cost for unit 81?
v/as
the averare budneted cost for unit 81?
PAGF
PIBLirnRAPHY
1.
PoupHs,
F.
VJ.
"The Procoss of Problem Find! no,"
,
InHustrial Mannnement Rev lev. Vol
2.
Newell, A., and Simon,
H.
Simiilates Human Thoimht,"
A.,
11,
1969
Mo 1,
"GPS, a Pronram That
In Flenenhaum,
Felrman, J., "Computerr, and Thounht," McGravz-MI
3.
SI
nop, H.
"The
A.,
f'ev/
and
A,,
E.
1
1
,
1963
Science of Management Decision,"
In "The Shape of Automation for Men and Machines,"
New York, 1968
I'arper and Row,
k,
5.
MInsky, M.
"Frame-Systems," Preliminary Draft,
,
Artificial
Intelllnence Laboratory, M.I.T., Oct. 30 1973
Mosher,
A.,
F.
In Bruner,
J.
and Hornsby, J. P., "On askinn Questions,"
S.
et al,
"Studies In Connltlve Growth,"
John Wiley, New York, 1966
6.
Bruner, J.
7.
Gore,
S,
et al,
"A Study of Thlnklno,"
John Wiley, New York, 1956
A
8.
J.,
Vi.
"Administrative Heel
Heurlrtic Model,"
Kelly, G.
Vol.
1,
A.,
s
i
on-MakI no
John Wiley, New York,
"The Psychol ory of Personal
"A Theory of Personality," Norton,
•
19 64
Constructs,"
New York,
1955
19,
j\'
^'""''^
MINI
llill
6,'?/
^73
lllllillliHill!l|lllllill!lill|ii:,J
TDflO
3
DD3 TDb OMM
iiniiiiiiiiii
TOaO 003 7Tb
3
->
l,^
3fi7
-IH
T06D DD3
3
---
mi
T-^lb
lilllll
lllillllllilillillL,
3
TOflD
QQ3
fl27
2fll
3
TDflD
003
flE7
fiSl
^^f-7H
liffniii
^^fe3
TDfiD
003
fl26
fi27
^-7'/
^'^^-'^
illllliilllliillllllllll
3
TOflO
003
fle7
77fl
--JIlllllJIIIHK)
yc,^
3 TDflo
003
I
3'5'g
6.^'7V{
lilllflPllf'flf'l
3
TOfiD
DQ3
fl^-p
g'
7?