Value Focused Thinking 101
The Mechanics
Mike Bailey
asst. by Matt Aylward, Greg Parnell (by extension)
Keeney, Ralph L. 1992. Value-Focused Thinking, A Path to
Creative Decisionmaking. Harvard University Press,
Cambridge, Mass.
Value-Focused Thinking
• A.K.A. Multi-Objective Decision Analysis (MODA)
• Descendant of Saul Gaas’s Analytical Hierarchical
Processes (AHP), and prior methods
• Amalgamate and Balance incomparable Desired
Properties
• I want a car that’s…fast, fun, cheap, cool, roomy, easy
to take care of, red, with a bike rack, long-lasting…
– How do you score cars to get the right balance?
– What if the right balance is being determined by a group
(possibly experts)?
• Discussion on application in a group of experts…Keeney
MECHANICS
• Skip all of the VFT Theology
• Focus on how to EXECUTE
– Construction of the Value Tree
– Calculation of Value Weights
– Evaluation of an Alternative
• Motivated by a simple example: Selecting
a College
BUILDING THE VALUE TREE
• Structure important issues into a two-layer
tree…
– Values (Functions)
– Characteristics (Tasks)
Pick the
Best College
V1. Price
V2. Selectivity
V3. Livability
THE VALUE TREE
Pick the
Best College
V2. Selectivity
V1. Price
V3. Livability
THE VALUE TREE
Pick the
Best College
V2.
Selectivity
FUNCTIONS
V1. Price
TASKS
V3. Livability
BUILDING THE VALUE
FUNCTION
• Each measure’s importance is weighted
for its contribution
• Each alternative is measured against each
measure
VFT Practice: Weights done before measurements. Try to keep the
participants away from the alternatives for as long as possible.
WEIGHTS
•
•
Have each participant individually rank order
the measures
For our example
1.
2.
3.
4.
5.
6.
7.
8.
9.
Tuition
Housing
Peterson
Proximity
SAT avg
BBall Rank
Cool
Living Expenses
Bars
PAIRWISE PREFERENCE
MATRIX
1.1
1.2
1.3
2.1
2.2
3.1
3.2
3.3
3.4
1.1
0
1
1
1
1
1
1
1
1
1.2
0
0
1
1
1
1
1
1
1
1.3
0
0
0
0
0
0
1
0
0
2.1
0
0
1
0
1
1
1
1
1
2.2
0
0
1
0
0
0
1
0
1
3.1
0
0
1
0
1
0
1
0
1
3.2
0
0
0
0
0
0
0
0
0
3.3
0
0
1
0
1
1
1
0
1
3.4
0
0
1
0
0
0
1
0
0
BUILDING UP WEIGHTS
• P = sum of all preference matrix elements
– Something close to n2/2
• Si = number of times option i preferred
– ith row sum of matrix P
• Rank the measures by Si
• Build clumps (3-6 clumps)
COLLEGE EXAMPLE
Tuition
19
Housing
14
Peterson
12
Prox to Home
12
SAT Score
11
BBall Team
9
Cool
8
Living Expenses
8
Bars
2
VARIABILITY
•
•
•
•
Judgment call
Made by Analyst, not Participants
High/Medium/Low
How much variability in the measure is
present in the options being considered?
– Tuition: $8,500 to $38,000 [HIGH]
– Proximity: 1hr to 6hr [LOW]
– SAT’s: 1050 to 1200 [LOW]
WEIGHT MATRIX
Importance
Variance
High
Med
Low
High
100
97
95
Med
95
92
85
Low
70
60
30
>> Pre-assign numerical weights to each cell. Bin the metrics
according to Importance and Variability.
>> Enforce Monotonicity.
MEASUREMENT
• Now we consider each alternative
• We will build a utility curve for each metric
– Translates a measurable (X) onto a [0,1] utility
value (Y) over the range of the alternatives
• We will measure each alternative’s utility
value
Key Concept: Families of Curves
• Linear
• Concave
– Each unit of X
returns one unit of
Y
Y
– Initially, increments
of X return less
than one unit of Y
MPH
Y
Payload
X
• Convex
– Initially, increments
of X return multiple
units of Y
Concealment
Y
X
• S-Curve
X
– Combines convex
and concave
H2O Prod.
Y
X
UTILITY CURVE: TUITION
COST UTILITY CURVE
0.9
0.8
0.7
UTILITY
0.6
0.5
0.4
0.3
0.2
0.1
0
8
13
18
23
28
33
COST
This can get fancy, but a line through a few points is AOK.
ROANOKE COLLEGE
BIN
WEIGHT
SCORE
TOTAL
Tuition
19
H/H
100
0.3
30
Housing
14
L/M
60
0.5
30
Peterson
12
M/M
92
0.4
36.8
Prox to Home
12
L/M
60
0.2
12
SAT Score
11
L/M
60
0.5
30
BBall Team
9
H/L
90
0.2
18
Cool
8
M/L
85
0.9
76.5
Living Expenses
8
M/L
85
0.9
76.5
Bars
2
?/VL
0
0.9
0
SCORE
309.8
FINAL LAP
•
•
•
•
Score each alternative on each measure
Take the weighted sum
That’s the alternative’s score
FINE’ !
SOME ISSUES TO
WATCH OUT FOR
• Interdependent measures
– Peterson guide and SAT Avg
– Guidance: Work to indentify aggregated but objective measures
that are independent
• Value’s influence
– More component measures leads to unintentionally over/under
emphasis of a specific value
– Guidance: Establish a number (3) of measures per Value
• Arbitrary weights
– Weight matrix can be unbalanced or overbalanced
– Do sensitivity analysis in the open
• Filtering
– Delete (not just weight = 0) infeasible alternatives
– E.g. a tuition you just can’t/won’t pay