Chapter 3

Structuring Decisions

Slide1 of 49
 Introduction

Suppose elements of Decision Problem (DP) are

available, i.e.:
•Objectives that apply to the decision context
•Immediate decision and subsequent decision(s)
•Alternatives for each decision
•Uncertain elements (events)
•You know how to evaluate consequences

How does one proceed structuring the DP?

STEP 1: Filter & Operationalize the Objectives

•Classify objectives as means or fundamental objectives
•Classify how to measure fundamental objectives

Slide 2 of 49
 Introduction

STEP 2: Structure the elements in a logical framework

•Structure Logic and time sequence between decisions
•Structure Logic (dependence) between the uncertain events
• Structure time sequence of uncertain events related to the
sequence of decisions
•Represent Logic by using Influence Diagrams or Decision Trees

STEP 3: Fill in the Details, e.g.;

• Give precise (unambiguous) definitions of decisions &
uncertain events
• Specify probability distributions for the uncertain events through
a combination of data analysis & expert judgment.
• Specify precisely (unambiguous) how consequences are measured
and formalize trade off between objectives.

Slide 3 of 49
 Introduction

Comment with Respect to STEP 1:

Identifying objectives is Creative Process
1. Develop a wish list. What do you want? What should you want?
2. Identify alternatives. What is the perfect alternative, a terrible alternative,
some reasonable alternatives, what is good or bad about each?
3. Consider problems and shortcomings. What is wrong or right with
your organization? What needs fixing?
4. Predict consequences. What has occurred that was good or bad?
What might occur that you care about?
5. Identify goals, constraints, and guidelines. What are your aspirations?
What limitations are placed on you?
6. Consider different perspectives. What would your competitor or constituency
be concerned about? At some time in the future, what would concern you?
7. Determine strategic objectives. What are your ultimate objectives? What
are your values that are absolutely fundamental?
8. Determine generic objectives. What objectives do you have for your
customers, your employees, your shareholders, yourself? What
environmental, social, economic, or health and safety objectives are important?
Source: Keeny, R.L. (1994) " Creativity in Decision Making with Value-Focused Thinking,"
Sloan Management Review, Sumer, 33-41

Slide 4 of 49
 STEP 1: Filter & Operationalize Objectives
Classify objectives as means or fundamental objectives
Maximize Safety

Maximize Use of Minimize Accidents

Vehicle-Safety Features

Motivate Purchase of Maintain Maximize

Safety Features Vehicles Driving Quality
on Vehicles Properly

Require Educate Public Enforce Have Reasonable Minimize Driving under

Safety Features about Safety Traffic Laws Traffic Laws Influence of Alcohol

Example of Means Objective Network

Slide 5 of 49
 STEP 1: Filter & Operationalize Objectives

Classify objectives as means or fundamental objectives

Maximize Safety

Minimize Minimize Minimize

Loss of Life Serious Injuries Minor Injuries

Adults Children Adults Children Adults Children

Example of Fundamental Objective Hierarchy

Slide 6 of 49
 STEP 1: Filter & Operationalize Objectives

Questions to aid in classifying objectives

Fundamental Means
Objectives Objectives
To Move: Downward in Hierarchy: Away from Fundamental
Objectives:

Ask: "What do you mean by that?" "How could you achieve this"

To Move: Upward in Hierarchy: Toward Fundamental

Objectives:
Ask: "Of what more general objective is
this an aspect?" "Why is that important?"

Decision Problems are evaluated using

only the Fundamental Objective Hierarchy
Slide 7 of 49
 STEP 1: Filter & Operationalize Objectives
Getting the decision context right:
•Enlarging Decision Context may increase the number of
objectives and alternatives that are relevant.

•Decreasing the Decision Context may cause current relevant

objectives or alternatives to become irrelevant.

Think of Example: Travel from A to B Comfortably

Three questions need to be answered affirmatively:

•Are you addressing the right problem?

•Can you make the decision? (Decision Ownership)

•Do you have the resources (Time & Money) to analyze

the DP in the current decision context?
Slide 8 of 49
Step 2: Structure Elements in a Logical Framework

Investment example:

Venture
Performance

Return on
Invest? Investment

Slide 9 of 49
Step 2: Structure Elements in a Logical Framework
Maximize Overall Satisfaction

Fundamental
Objective Hierarchy
Invest in Return on
Computer Industry Investment

Venture
Invest? Performance

Invest in
Return on
Computer
Investment Influence Diagram
Industry

Overall
Satisfaction

Slide 10 of 49
Step 2: Structure Elements in a Logical Framework
Influence Diagrams
1. Elements are represented by:

Decision Nodes: Invest?

Venture
Chance Nodes: Performance

Return on
Consequence Nodes: Investment

Slide 11 of 49
Step 2: Structure Elements in a Logical Framework
2. Logical relationships are represented by arrows:
Indicates Sequence; points only to Decision Nodes
Indicates Dependence; points only to Chance Nodes
& Consequence Nodes

A B C D

Decision A is made Outcome of C is known

before Decision B before Decision D is made

E F G H

Decision E is relevant Outcome of G is relevant

to assessing probabilities to assessing probabilities
of outcomes of F of outcomes of H

Slide 12 of 49
Step 2: Structure Elements in a Logical Framework

• It is important to note here that we deviate from

the convention in the book that does not distinguish
between sequence arcs and influence arcs.
• This distinction is made here only for teaching
purposes. Once one is comfortable with the differences
between these arcs and their interpretations one could
use solid arcs throughout the influence diagram
Comments on Influence Diagrams:
•Influence diagram captures current state of knowledge
•An influence diagram should NEVER contain cycles
•Interpreting an influence diagram is generally easy
•Creating influence diagrams is difficult

Slide13 of 49
 Step 2: Basic Influence Diagrams
Basic Risky Decision

• One should be able Outcomes

to identify basic Business Wild Success
influence diagrams Result Flop
and modify/combine
them to match
specific problems • Involves one decision
and one uncertain event

Investment Return on
Choice Investment

Choice Business Result Return

Alternatives Savings Wild Success 2200
Savings Flop 2200
Business Business Wild Success 5000
Flop 0
Slide14 of 49
 Step 2: Basic Influence Diagrams

Outcomes
Hurricane Hits Miami
Forecast
Path Misses Miami

Forecast Imperfect Information:

"Hits Miami"  Involves one decision and
"Misses Miami" two uncertain events at the
time of the Decision Analysis.
 One uncertain event is
known at the time that the
immediate decision is made.
Consequence  Solving the influence
Decision
diagram results in one optimal
decision for each possible
outcome of the information source.
Alternatives
Consequence
Evacuate
Choice Outcome Risk Cost
Stay Evacuate Hits Miami Low Risk High Cost
Misses Miami Low Risk High Cost
Stay Hits Miami High Risk High Cost
Misses Miami Low Risk Low Cost

Slide15 of 49
 Step 2: Basic Influence Diagrams
Sequential Decisions:

Hurricane
Forecast
Path

Wait for
Evacuate?
Forecast?

Consequence

Hurricane example

Slide16 of 49
 Step 2: Basic Influence Diagrams
Sequential Decisions:

Forecast Weather Forecast Weather Forecast Weather

Day 1 Day 1 Day 2 Day 2 Day n Day n

Protect Protect Protect

Day 1? Day 2? Day n?

Cash Flow Cash Flow Cash Flow

Day 1 Day 2 Day n

Net Present
Value
Farmer example

Slide17 of 49
 Step 2: Basic Influence Diagrams
Intermediate calculations for additional clarity:
Revenue Cost

Version 1
Introduce Profit
Product?

Units Fixed Variable

Price? Sold Cost Cost

Version 2

Introduce Profit
Product?

Slide18 of 49
 Step 2: Basic Influence Diagrams
Intermediate calculations for additional clarity:

Units Fixed Variable

Sold Cost Cost
Price?

Revenue Cost

Introduce
Product?
Profit Version 3

Slide 19 of 49
 Step 2: Basic Influence Diagrams
Intermediate calculations reflecting Fundamental Objective Hierarchy:

Bomb-Detection
System Choice

Detection Time to Passenger

Cost
Effectiveness Implement Acceptance

Overall
Performance

Slide 20 of 49
 Step 2: Building an Influence Diagrams

1. List all the decisions.

2. Draw sequence arcs between decisions.
3. Identify the consequence node.
4. Breakdown the consequence node using the FOH.
5. Draw relevance arcs from decision nodes to the
intermediate calculation nodes.
6. List all the uncertainty nodes.
7. Draw the relevance arcs between uncertainty nodes.
8. Draw the sequence arcs from uncertainty nodes
to the decision nodes.
9. Draw the relevance arcs from the decision nodes to the
uncertainty nodes.
[Link] the relevance arcs from the uncertainty nodes to
the intermediate calculation nodes.
Slide 21 of 49
 Step 2: Building an Influence Diagram

All details (outcomes, choices, payoff) should

be developed in a tabular format for each
node in the influence diagram.

Common Mistakes:
•See influence diagrams as flow charts.
• Many chance nodes going into the immediate
decision node to reflect uncertainty.
•The inclusion of cycles.

Slide 22 of 49
 Building an Influence Diagram: Example
Toxic Chemicals and the EPA

The Environmental Protection Agency often must decide whether to

permit the use of an economically beneficial chemical that may be
carcinogenic (cancer-causing). Furthermore, the decision often must
be made without perfect information about either the long-term
benefits or health hazards. Alternative courses of action are to
permit the use of the chemical, restrict its use, or to ban it al together.
Tests can be run to learn something about the carcinogenic
potential, and survey data can give an indication of the extent to
which people are exposed when they do use the chemical. These
pieces of information are both important in making the decision. For
example, if the chemical is only mildly toxic and the exposure rate is
minimal, then restricted use may be reasonable. On the other hand, if
the chemical is only mildly toxic but the exposure rate is high, then
banning its use may be imperative.

Slide23 of 49
Building an Influence Diagram: Example
Toxic Chemicals and the EPA

STEPS 1 - 5

Economic
Usage
Value
Decision

Net
Value
Cancer
Cost

Slide 24 of 49
 Building an Influence Diagram: Example
Toxic Chemicals and the EPA

STEPS 6 - 7

Economic
Usage
Value
Decision
Survey Net
Value
Cancer
Laboratory Cost
Test Exposure
Rate

Carcinogenic
Potential

Slide 25 of 49
 Building an Influence Diagram: Example
Toxic Chemicals and the EPA

STEPS 9 - 10

Economic
Usage
Value
Decision
Survey Net
Value
Cancer
Laboratory Cost
Test Exposure
Rate

Carcinogenic
Potential

Slide 26 of 49
 Building an Influence Diagram: Example

Test Analyze and Fix (TAAF)

We are deciding on whether or not to release a new product. The
product reliability, that is, the probability that the product functions
satisfactorily, is a main driver in profits that may be derived from the
product. We may decide to release the product immediately based on
our best current assessment of the product reliability or test the product.
A test consists of selecting a single product and seeing if it performs its
required task. If it does not, the product is redesigned using the test
results and a fixed cost for redesign is incurred. This redesign may or may
not improve the product reliability. After the redesign we may decide to
release the redesigned product or test the redesigned product. If the
redesigned product does not perform its task, the redesigned product will
be redesigned using the latest test results, etc. This cycle continues until
the time has come that the final decision to release or not to release has
to be made. The testing and re-design takes one week (regardless of the
outcome of the test) and we have two weeks to make the final decision.
Assume that the product reliability and the cost of testing affect the profits.

Slide 27 of 49
 Building an Influence Diagram: Example
Test Analyze and Fix (TAAF)

Final
Release 1? Release 2?
Release?

STEPS 1 - 3 FINAL PROFITS

Slide 28 of 49
 Building an Influence Diagram: Example
Test Analyze and Fix (TAAF)

Final
Release 1? Release 2?
Release?

Profit if Cost of Profit if Cost of Profit if

released Test 1 & released Test 2 & released
after 1 Redesign after 2 Redisgn after final

STEPS 4 - 5 FINAL PROFITS

Slide 29 of 49
 Building an Influence Diagram: Example
Test Analyze and Fix (TAAF)
Current Reliability
Outcome Reliability Outcome
Reliability after Test 2
Test 1 after Test 1 Test 2

Final
Release 1? Release 2?
Release?

Profit if Cost of Profit if Cost of Profit if

released Test 1 & released Test 2 & released
after 1 Redesign after 2 Redisgn after final

STEPS 6 - 7 FINAL PROFITS

Slide 30 of 49
 Building an Influence Diagram: Example
Test Analyze and Fix (TAAF)
Current Reliability
Outcome Reliability Outcome
Reliability after Test 2
Test 1 after Test 1 Test 2

Final
Release 1? Release 2?
Release?

Profit if Cost of Profit if Cost of Profit if

released Test 1 & released Test 2 & released
after 1 Redesign after 2 Redisgn after final

STEP 8 FINAL PROFITS

Slide 31 of 49
 Building an Influence Diagram: Example
Test Analyze and Fix (TAAF)
Current Reliability
Outcome Reliability Outcome
Reliability after Test 2
Test 1 after Test 1 Test 2

Final
Release 1? Release 2?
Release?

Profit if Cost of Profit if Cost of Profit if

released Test 1 & released Test 2 & released
after 1 Redesign after 2 Redisgn after final

STEPS 9 - 10 FINAL PROFITS

Slide 32 of 49
Step 2: Structure Elements in a Logical Framework

Investment example:

Return on Investment
Do not Invest
Interest @ 5%

Invest? Failure
Funds Lost

Invest Venture
Performance
Success Large Return

Slide 33 of 49
Step 2: Structure Elements in a Logical Framework

Decision trees and fundamentals-objective hierarchy:

Overall
Performance

Detection Time to Passenger

Cost
Effectiveness Implement Acceptance
System 1

System 2
System 3
System 4

System 5

Decision Trees reveal more detail

than Influence Diagrams

Slide 34 of 49
Step 2: Structure Elements in a Logical Framework
Do not Invest

Elements are represented by:

Decision Nodes: Invest?

Invest

Failure

Venture
Chance Nodes: Performance
Success
Maximize Overall
Satisfaction

Invest in Return
Computer on
Do not Invest Industry Investment

Consequences: Invest?
Invest

Slide 35 of 49
Step 2: Structure Elements in a Logical Framework

1. Decision Trees are evaluated from left to right

2. Only one alternative can be chosen after each
decision node
3. Outcome from a chance event need to be
complete, i.e. not more than one outcome can
happen at the same time and one outcome will
happen.
4. Decision Trees represent all possible future
scenarios
5. Think of nodes as occurring in time sequence
6. If for chance nodes the order is not important
then use the easiest interpretation.

Slide 36 of 49
Step 2: Basic Decision Trees

The Basic Risk Decision

Increase Power
Run for Reelection
U.S. Representative
(Intermediate)

Win
U.S. Senator
(Best)
Run for Senate

Lose Lawyer
(Worst)

Slide 37 of 49
Step 2: Basic Decision Trees

The Double Risk Dilemma

Increase Power

Win U.S. Representative

(Second-Best)
Run for Reelection

Lose Small-Time Lawyer

(Worst)
Win
U.S. Senator
(Best)
Run for Senate

Lose Big-Time Lawyer

(Okay)

Slide 38 of 49
 Step 2: Basic Decision Trees

Rang-of-Risk Dilemma

Maximize Court Outcome

Accept Settlement
Known Amount

Settle? Highest

Reject
Settlement
Amount of Court Award

Lowest

Slide 39 of 49
 Step 2: Basic Decision Trees
Maximize Overall
Imperfect Information Satisfaction

Risk Cost
Evacuate
Forecast: Safe High
“Hurricane Hits Hurricane Hits
Miami” Miami
Danger Low

Stay
Hurricane Misses
Miami Safe Low

Evacuate
Forecast:
Safe High
“Hurricane Misses
Miami” Hurricane Hits
Miami
Danger Low

Stay
Hurricane Misses
Miami Safe Low
Slide 40 of 49
Step 2: Basic Decision Trees
Maximize Overall
Satisfaction
Sequential Decisions
Risk Cost
Evacuate Now
Safe High

Evacuate
Forecast: Safe High
“Hurricane Hits Hurricane Hits
Miami” Miami
Low
Danger
Stay
Hurricane Misses
Miami Safe Low

Evacuate
Forecast: Safe High
“Hurricane Misses
Miami” Hurricane Hits
Miami Danger Low

Stay
Hurricane Misses
Miami Safe Low

Slide 41 of 49
 Step 2: Basic Decision Trees

Sequential Decisions

Forecast: Decision: Weather: Forecast: Decision: Weather:

Day 1 Day 1 Day 1 Day 2 Day 2 Day 2

Net
... Present
Value

Slide 42 of 49
Decision Trees and Influence Diagrams Compared
• When DP is complex decision trees, may get
too large for presentation.
• For presentation a DP influence diagram are superior.
• Decision trees show more detail, hence are more
useful for in-depth understanding.
• Influence diagrams are better in the structuring phase.
• For sensitivity analysis decision trees may be better.
• Influence diagram present the relevance between
uncertainty nodes, decision trees do not.
• Influence diagrams and decision trees are isomorphic.
• Each technique has its strength and weakness.
Using both may work complementary.
• Both must pass the clarity test. No misunderstanding
should be possible about the basic key elements in
the decision problem.
Slide 43 of 49
 Step 3: Fill in the Details
1. Define the elements in the decision model clearly
Consider EPA Example:

Objective: Minimize the social cost of cancer. Definition?:

•Incremental Lives Lost ?
•Incremental cases of cancer ?
•Incremental cases of treatable cancer ?

Uncertain Event: Rate of exposure. Definition?:

•Number of people exposed to the chemical per day ?
•Ingesting a critical quantity ?
•Skin contact ?

Slide 44 of 49
 Step 3: Fill in the Details
1. Define the elements in the decision model clearly
Consider EPA Example:

Objective: Minimize the social cost of cancer. Definition?:

•Incremental Lives Lost ?
•Incremental cases of cancer ?
•Incremental cases of treatable cancer ?

Uncertain Event: Rate of exposure. Definition?:

•Number of people exposed to the chemical per day ?
•Ingesting a critical quantity ?
•Skin contact ?

Slide 45 of 49
 Step 3: Fill in the Details

2. Asses Cash Flows and Probabilities

Sell Licence TRUE 70.00%

$25,000,000 $23,000,000

Patent Awarded 70.0% Decision

0 23000000

High 25.0% 0.00%

$55,000,000 $43,000,000

Develop in-house FALSE Demand

-$10,000,000 22900000

Medium 55.0% 0.00%

$33,000,000 $21,000,000

Low 20.0% 0.00%

$15,000,000 $3,000,000

Continue Development TRUE Chance

-$2,000,000 15500000

No Patent 30.0% 30.00%

0 -$2,000,000
Decision
15500000

Stop Development FALSE 0.00%

$0.00 $0

Slide 46 of 49
 Step 3: Fill in the Details
3. Defining Measurement Scales for Fundamental
Objectives
Objectives are measured in attributes, e.g: Dollars, Hours, Percentage
1. Objectives with natural attribute scale
OBJECTIVE ATTRIBUTE
Maximize profit Money ( for example dollars)
Maximize Revenue Money ( for example dollars)
Maximize Savings Money ( for example dollars)
Minimize Cost Money ( for example dollars)
Maximize Market Share Percentage
Maximize Rate of Return Percentage
Maximize proximity Miles, minutes
Maximize Fuel Efficiency Miles per gallon
Maximize time with friends Days, Hours
Minimize hypertension Inches Hg (Blood pressure)

Slide 47 of 49
 Step 3: Fill in the Details

2. Objectives with no natural attribute scale

Example: Max Quality. How does on measure it?

Solution: Define an attribute scale:

BEST, BETTER, SATISFACTORY, WORSE, WORST.

Each category needs to pass the Clarity Test:

• An attribute Scale passes the Clarity Test if an
outside candidate using your definition can allocate
specimens to the identified categories.

Slide 48 of 49
 Step 3: Fill in the Details
Example: Measuring Survey Quality:
Best Survey Quality:

State-of-the-art Survey. No apparent crucial issues

left unaddressed. Has characteristics of the best
survey projects presented at professional
conferences.
Worst Survey Quality:
Many issues left unanswered in designing survey.
Members of the staff are aware of advances in
survey design that could have been incorporated
but were not. Not a presentable project.
See additional category definitions on Page 82.

Slide 49 of 49