You are on page 1of 40

MGMT 3130 Judgment and Decision

Making in Organizations

UNIT 1.2
DECISIONS INVOLVING
MULTIPLE OBJECTIVES

1
Where are we now?

Decision Heuristics Choice and Applications


Analysis and Biases Preference
Two systems of
Problem thinking Performance
Prospect theory
definition appraisal
Availability
heuristic
Motivating
Decisions involving Escalation of
Representativeness employees by
multiple objectives commitment
heuristic money? Or…?

Anchoring and Why people


Decision-making adjustment Time and
become bad
under uncertainty preferences
apples?
Embodied
cognition
Collecting more Is more choice
Bounded Going green
information always better?
awareness
2
UNIT 1.2
DECISIONS INVOLVING MULTIPLE OBJECTIVES

How people usually deal with decisions that involve


multiple objectives?
• Lexicographic choice
• Semi-lexicographic choice
• Elimination by aspects
• Sequential decision making: Satisficing
Using decision analysis to deal with multiple objectives
and tradeoff:
• Simple Multi-Attribute Rating Technique (SMART)

Goodwin, P. & Wright, G. (2009). Decision Analysis for Management Judgment, 4th Edition. UK: John Wiley &
Sons. Chapters 2-3. 3
To evaluate alternatives, we start by
CONSTRUCTING THE CONSEQUENCES TABLE

Alternatives

Objectives

We need to specify how well each alternative meets each of our objectives.
• Some objectives involve natural measurement
• Maximize profit: $$$
• Maximize exposure of a TV ad: Number of people surveyed who recall
seeing the ad
• Some objectives involve subjective measurement
• Job candidate's ability to facilitate contributions of team members:
Excellent, Good, Average, or Poor
(see next slide for an example of measurement tool)
4
• For example: The Association of American Colleges and Universities
uses the following subjective criteria to measure students’ ability to
facilitate contribution of team members:
• Excellent: Engages team members in ways that facilitate their contributions
to meetings by both constructively building upon or synthesizing the
contributions of others as well as noticing when someone is not
participating and inviting them to engage.
• Good: Engages team members in ways that facilitate their contributions to
meetings by constructively building upon or synthesizing the contributions
of others.
• Average: Engages team members in ways that facilitate their contributions
to meetings by restating the views of other team members and/or asking
questions for clarification.
• Poor: Engages team members by taking turns and listening to others
without interrupting.
• The meaning of each category (e.g. excellent, poor) should be clearly
described.
5
• We collect information on the consequences (how well each
alternative meets each objective), and put this information into a
more useful format: The consequences table.
• Example: Choosing among job offers
• Salary – $/monthly
• Flexibility of work schedule – Low, Moderate, High
• Vacation time – # of days/year
• Enjoyment – Boring, Average, Great
Job A Job B Job C Job D Job E

Monthly salary ($) 16,000 19,200 15,200 11,400 11,400

Flexibility of work
Moderate Low Moderate High Low
schedule

Vacation (days) 14 12 15 10 14

Enjoyment Great Average Boring Average Boring


6
DOMINANCE AND TRADEOFFS

(Better)
A good starting place when you
finish your consequences table is to
Alternative A
look for dominated alternatives and
eliminate them.

Attribute Y
Alternative B1
- If Alternative A is better than
Alternative B2
Alternative B on some objectives
and no worse than B on all other
objectives, B is said to be
dominated by A. (Better)
Attribute X

7
(Better)

Alternative A
When no alternative dominates
the others.

Attribute Y
Alternative B
You need to deal with tradeoffs.

(Better)
Attribute X

8
PRS
What can you say about these 5 alternatives?
Job A Job B Job C Job D Job E

Monthly salary ($) 16,000 19,200 15,200 11,400 11,400

Flexibility of work
Moderate Low Moderate High Low
schedule

Vacation (days) 14 12 15 10 14

Enjoyment Great Average Boring Average Boring

1. Job A dominates Job B.


2. Job C dominates Job D.
3. Job D dominates Job E.
4. Job A dominates Job E.
9
WITHOUT EXPLICITLY HANDLING TRADEOFFS,
how people usually deal with decisions that involve multiple objectives?

Decision makers often use the following methods to make decisions


involving multiple objectives:
• Lexicographic choice
• Semi-lexicographic choice
• Elimination by aspects
• Sequential decision making: Satisficing

10
Lexicographic Choice
• Steps:
1. Pick the most important attribute.
2. Choose the alternative that has the best performance on that attribute.
3. If there is a tie, move on to the next most important attribute.
• Suppose your most important attribute is # of vacation days, 2nd
enjoyment, 3rd flexibility of work schedule, and 4th salary.
Using lexicographic choice, your choice will be: ______________

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
Flexibility of work
Moderate Low Moderate High
schedule
Vacation (days) 14 12 15 10

Enjoyment Great Average Boring Average


11
• Is lexicographic choice a good way to make decisions?
• It’s quick if there are few ties.
• It’s easy, it only requires you to rank your attributes (e.g. Is salary
more important than vacation days?).
• BUT lexicographic choice is non-compensatory:
• An alternative’s poor performance on one attribute cannot be
compensated for by good performance on other attributes.
• E.g. If the most important attribute is salary and vacation days comes
next, then $10k salary and 1 vacation day is still preferred to $9,999
salary and 100 vacation days.
Job X Job Y
Monthly Salary $ 10,000 $ 9,999

Vacation 1 day per year 100 days per year

12
Semi-lexicographic Choice

• This differs slightly from lexicographic choice in that, if the performance


of alternatives on an attribute is similar, the decision maker considers
them to be tied and moves on to the next attribute.
• Suppose your most important attribute is # of vacation days,
2nd enjoyment, 3rd flexibility of work schedule, and 4th salary. And, your
rule is: “if difference in # of vacation days is 2 or less, choose the job with
greater enjoyment.” Your choice: _______

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
Flexibility of work
Moderate Low Moderate High
schedule
Vacation (days) 14 12 15 10

Enjoyment Great Average Boring Average


13
Elimination by Aspects
• Steps:
1. Pick the most important attribute.
2. Decide on a cutoff point for that attribute.
3. Eliminate all alternatives that do not meet that cutoff.
4. Move on to the next most important attribute.
• Suppose your most important attribute is salary (cutoff: $15,000), 2nd
vacation (cutoff: 12 days), 3rd flexibility of work schedule (cutoff: moderate),
and the least important one is enjoyment (cutoff: average). Using
elimination by aspects, your choice will be: __________
Job A

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
Flexibility of work
Moderate Low Moderate High
schedule
Vacation (days) 14 12 15 10

Enjoyment Great Average Boring Average 14


• Is elimination by aspects a good way to make decisions?
• It’s quick and easy.
• It protects you from mistakenly selecting inferior alternatives
because alternatives that do not meet the cutoff are removed.
• It is also a good strategy to narrow down a huge number of
alternatives to a few ‘finalists’ that you can evaluate in more detail.
• But, it’s also non-compensatory:
• Only one attribute is considered at a time, not explicitly handling
tradeoffs among attributes.

15
Sequential Decision Making: Satisficing

• The previous methods are intended to describe how people make a


decision when they evaluate all alternatives simultaneously.
• In some situations, however, decision makers evaluate alternatives
sequentially.
• People satisfice, i.e. search for new alternatives until they find a
satisfactory alternative that will suffice.
• Thus the chosen option is just “good enough,” may not be the best
available option. The choice also depends on the order in which the
alternatives present themselves.

16
HANDLING TRADEOFFS:
SIMPLE MULTI-ATTRIBUTE RATING TECHNIQUE (SMART)

• Steps:
1. For each attribute, assign subjective values (utilities)
to reflect the performance of the alternatives on that attribute.
2. Determine a weight for each attribute. Importance
3. For each alternative, compute a weighted utility.
Then, make a provisional decision.
4. Perform sensitivity analysis.
• Let’s illustrate with an example.

17
Step 1: For each attribute, assign subjective values (utilities) to reflect
the alternatives’ performance on that attribute.
• For each attribute, identify the most preferred level and assign it a
utility of 1, then identify the least preferred level and assign it a
utility of 0.
• Examine each intermediate level and assign it a utility in between 0
and 1 so that the space between utilities represents the strength of
preference for one attribute level over another.

Least preferred Most preferred

Job A Job B Job C Job D


Flexibility of work Moderate Low Moderate High
schedule (?) (0.0) (?) (1.0)

Intermediate level
18
1
• E.g. If the improvement from
‘low’ to ‘moderate’ is seen by 0.75

Utility
you to be 50% as preferable as 0.5
the improvement from ‘low’ to
0.25
‘high,’ assign ‘moderate’ a utility
of 0.5. 0
Low Moderate High
Flexibility of work schedule

• E.g. If the improvement from 1


‘low’ to ‘moderate’ is seen by
0.75
you to be 75% as preferable as

Utility
the improvement from ‘low’ to 0.5

‘high,’ assign ‘moderate’ a utility 0.25


of 0.75. 0
Low Moderate High
• Remember that it is the interval (or Flexibility of work schedule
improvement) between the attribute
levels that we compare.
19
Job A Job B Job C Job D
16,000 19,200 15,200 11,400
Monthly salary ($)
(0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule (0.5) (0.0) (0.5) (1.0)
14 12 15 10
Vacation (days)
(0.6) (0.4) (1.0) (0.0)
Great Average Boring Average
Enjoyment
(1.0) (0.7) (0.0) (0.7)

20
Step 2: Determine a weight for each attribute.
Step 1 • Imagine two hypothetical alternatives: one with the worst
performance on all attributes, another with the best.
• For each attribute, consider the importance of the change from
the worst level to the best level.
• For the attribute that you would most like to move from
the worst level to the best level, assign a weight of 100.
• Assign weights to the rest of attributes based on their importance
relative to the most important change.
• Normalize the weights.

The weights reflect the importance of


changes (or swings) not the
importance of attributes per se
21
Job A Job B Job C Job D
16,000 19,200 15,200 11,400
Monthly salary ($)
(0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule (0.5) (0.0) (0.5) (1.0)
14 12 15 10
Vacation (days)
(0.6) (0.4) (1.0) (0.0)
Great Average Boring Average
Enjoyment
(1.0) (0.7) (0.0) (0.7) Step 1

Hypothetical Hypothetical Swing


worst best Weight
alternative alternative

Monthly salary ($) 11,400 19,200

Flexibility of work
Low High
schedule
Vacation (days) 10 15

Enjoyment Boring Great


22
• In this case, the swing from ‘Boring’ to ‘Great’ enjoyment is considered
to be 50% as important as the swing from $11,400 to $19,200 monthly
salary.
Hypothetical Hypothetical Swing Normalized
worst best Weight Weight
alternative alternative
100/180 =
Monthly salary ($) 11,400 19,200 100 0.556
Flexibility of work 20/180 =
Low High 20 0.111
schedule
10/180 =
Vacation (days) 10 15 10 0.056
50/180 =
Enjoyment Boring Great 50 0.278

Total 180
Just think of some random numbers
23
Why swing weights?
• Suppose that the choice is between just two jobs and that we are
evaluating these on only two attributes.
• If we consider relative importance of attributes, say, you consider
salary to be five times more important than vacation days …
Job X Job Y
Monthly Salary $ 10,002 $ 10,000
(5) (1.0) (0.0)
Vacation 10 days per year 20 days per year
(1) (0.0) (1.0)

Aggregated utility of Job X = 5, that of Job Y = 1; but sacrificing 10 days


of vacation to gain $2 sounds strange.
• If the alternatives perform very similarly on a particular attribute (i.e.
the range between worst and best is small), then this attribute is
unlikely to be important in the decision, even though the decision maker
may consider it to be an important attribute in general.
24
Step 3: For each alternative, compute a weighted utility. Then,
make a provisional decision.

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
[0.556] (0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule [0.111] (0.5) (0.0) (0.5) (1.0)
Vacation (days) 14 12 15 10
[0.056] (0.6) (0.4) (1.0) (0.0)
Enjoyment Great Average Boring Average
[0.278] (1.0) (0.7) (0.0) (0.7)


Weighted utility of
0.756 0.772 0.444 0.306
each alternative

25
Step 4: Perform sensitivity analysis.

• Even though our SMART approach is very systematic, it still depends


upon subjective judgments.
• How good are the intermediate levels of attribute performance
(0 £ utility £ 1)
• How important are the attributes in the decision
(0 £ weight £ 100, before normalizing)
• The purpose of sensitivity analysis is to determine how sensitive your
choice is to the judgments that you’ve made. In other words, if your
judgments were somewhat different, would your choice remain the
same?
• On judgments that you are not very confident about, please conduct
a sensitivity analysis .

26
Example:
• Previously, we assigned the enjoyment attribute a weight of 50.
• Will our choice remain the same if this attribute is weighted differently?
• Which alternative is the best when enjoyment has a weight of 0?
• Which alternative is the best when enjoyment has a weight of 100?

Hypothetical Hypothetical Swing


worst best Weight
alternative alternative

Monthly salary ($) 11,400 19,200 100

Flexibility of work
Low High 20
schedule

Vacation (days) 10 15 10

Enjoyment Boring Great 50


27
When enjoyment has a weight of 0:

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
[100/130 = 0.77] (0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule (0.5) (0.0) (0.5) (1.0)
[20/130 = 0.15]
Vacation (days) 14 12 15 10
[10/130 = 0.08] (0.6) (0.4) (1.0) (0.0)
Enjoyment Great Average Boring Average
[0/130 = 0.00] (1.0) (0.7) (0.0) (0.7)


Weighted utility of
0.662 0.800 0.615 0.154
each alternative

28
When enjoyment has a weight of 100:

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
[100/230 = 0.43] (0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule (0.5) (0.0) (0.5) (1.0)
[20/230 = 0.09]
Vacation (days) 14 12 15 10
[10/230 = 0.04] (0.6) (0.4) (1.0) (0.0)
Enjoyment Great Average Boring Average
[100/230 = 0.43] (1.0) (0.7) (0.0) (0.7)


Weighted utility of
0.809 0.757 0.348 0.391
each alternative

29
Our choice is sensitive to the weight placed on enjoyment.
It can be seen that Job B is the best alternative as long as the weight
placed on enjoyment is less than 60.
0.9
0.8 0.8 Job B 0.809
Weighted utility of Alternatives

0.757
0.7 Job A
0.662
0.6 0.615
0.5 Job C
0.4 0.391
0.348
0.3
Job D
0.2
0.154
0.1
0
0 10 20 30 40 50 60 70 80 90 100
Weight on Enjoyment

*You may refer to the Excel file on Canvas for more information on how to construct this graph. 30
• Carrying out sensitivity analysis might lead decision-makers to
reconsider some of the figures they supplied. But in many cases,
sensitive analysis shows that our choices are stable over a range of
those figures.
• Note that this sensitivity analysis only allows decision-makers to
investigate the effect of changing one variable (value or weight) at
a time.
• Mustajoki et al. (2006) have developed a modeling method to
simultaneously vary weights and values. Take a look only if you are
interested and good at mathematics.

Mustajoki, J., Hamalainen, R.P., & Lindstedt, M.R.K. (2006). Using intervals for global
sensitivity and worst-case analyses in multiattribute value trees. European Journal of
Operational Research, 174, 278-292.

31
In-class exercise:
• Previously, we assigned ‘moderate level of flexibility’ a utility of 0.5.
• Is our choice sensitive to the utility assigned to ‘moderate
flexibility’?
• Hints: Calculate the weighted utilities of the alternatives when
‘moderate flexibility’ is assigned 0 and 1.

Job A Job B Job C Job D


Monthly salary ($) 16,000 19,200 15,200 11,400
[0.556] (0.7) (1.0) (0.6) (0.0)
Flexibility of work Moderate Low Moderate High
schedule [0.111] (0.5) (0.0) (0.5) (1.0)
Vacation (days) 14 12 15 10
[0.056] (0.6) (0.4) (1.0) (0.0)
Enjoyment Great Average Boring Average
[0.278] (1.0) (0.7) (0.0) (0.7)

32
Your answer here:

When ‘moderate flexibility’ has a utility of 0, the weighted utilities are:


0.700. 0.772 0.389. 0.306
Job A: ______ Job B: ______ Job C: ______ Job D: ______
Thus the choice should be ________.
Job B

When ‘moderate flexibility’ has a utility of 1, the weighted utilities are:


0.811. 0.772 0.500. 0.306
Job A: ______ Job B: ______ Job C: ______ Job D: ______
Thus the choice should be ________.
Job A

Sensitive
Our choice is (sensitive/insensitive) to the utility assigned to ‘moderate
flexibility.’

33
Conflicts between intuitive and analytic results

• Your intuitive choice may be different from that obtained through


SMART. This conflict should be examined.
• This conflict could be because the decision problem was too large
and complex for you to handle without using DA, so that your true
preferences were not reflected in your intuitive judgment.
Research has suggested that the correlation between “preference
derived from intuitive judgment” and “preference derived from
SMART” decreases as the number of attributes in the problem gets
larger.
• The conflict may also suggest that an important element of the
decision problem has not been captured by the analysis (e.g.
missing or overlapping objectives).

34
An Example of Decision Making Using SMART

HKUST HKU HKBU


Training on research Ranked 6 Ranked 77 Not in the top
(Proxy: Research ranking) 100
(1.0) (0.3) (0.0)
Training on teaching Poor Poor Good
(0.0) (0.0) (1.0)
Commute time and means 35 minutes by 5 minutes by 20 minutes by
taxi walk taxi
(0.0) (1.0) (0.2)

For each attribute,


• The most preferred level has a utility of 1.
Doing a PhD in • The least preferred level has a utility of 0.
Management
• Intermediate ones have a utility in between 0 and 1
(i.e. 0 £ utility £ 1).
35
Hypothetical worst Hypothetical best Swing Weight
alternative alternative
Training on research Not in the top 100 Ranked 6 100
(Proxy: Research ranking)

Training on teaching Poor Good 80

Commute time and means 35 minutes by taxi 5 minutes by walk 20

For each attribute, decide how important it is for you to


move from the worst outcome to the best outcome.
• Give a weight of 100 to the most important move.
• Other moves, give a weight in between 0 and 100
(i.e. 0 £ weight £ 100).
36
Normalize the weights.
HKUST HKU HKBU
Training on research Ranked 6 Ranked 77 Not in the top
(Proxy: Research ranking) 100
(0.5) (1.0) (0.3) (0.0)
Training on teaching Poor Poor Good

(0.4) (0.0) (0.0) (1.0)


Commute time and means 35 minutes by 5 minutes by 20 minutes by
taxi walk taxi
(0.1) (0.0) (1.0) (0.2)

Weighted utility

0.50 0.25 0.42

Pick the alternative with the HIGHEST weighted


utility as the provisional choice.

37
She is not so sure about the weight she gave to “teaching-related training,”
so she conducts a sensitivity analysis.
Sensitivity analysis shows that HKUST is the best alternative as long as the
weight placed on “Teaching-related training” is less than 96.
0.90
0.80
Weighted utility of alternatives

0.70
0.60
0.50
HKUST
0.40
HKU
0.30 HKBU
0.20
0.10
0.00
0 10 20 30 40 50 60 70 80 90 100
Weight on teaching-related training

38
Time for REFLECTION

Describe a decision problem that you’re trying to solve.


How would you collect information about the consequences?
What decision strategy would you use (lexicographic choice, elimination
by aspects, SMART, etc.)? Why?

39
INTENDED LEARNING OUTCOMES FOR UNIT 1.2

By now, you should be able to:


• Apply the following decision strategies and describe their
pros and cons
• Lexicographic choice
• Semi-lexicographic choice
• Elimination by aspects
• Sequential decision making: Satisficing
• Perform a SMART decision analysis (be sure that you
understand every step)

40

You might also like