49
Decision Making and Risk Spring 2006 Session 9 ychology of Decision Makin

Decision Making and Risk Spring 2006 Session 9 Psychology of Decision Making

Embed Size (px)

Citation preview

Decision Making and RiskSpring 2006

Session 9

Psychology of Decision Making

Agenda

Issues in “Psychology of Decision Making”

Recall Typical Decision Tree

Where is the room for psychology in decision making?

In other words, why should it matter? Separate from human computational constraints.

Option A

Option B

Outcome A1

Outcome A2

Outcome B1

Outcome B2

Payoff Portfolio A1

Payoff Portfolio A2

Payoff Portfolio B1

Payoff Portfolio B2

p(A1)

p(A2)

p(B1)

p(B2)

Psychology of DM Matters Filling the decision tree

Biases in Generating probabilities Structuring and attaching value to elements of the payoff portfolio Identifying decision options

Role of reference points. Broader framing effects

Task Option Action Loss Aversion Risk Preference Super and Sub-additivity

Role of Emotion, Regret and Counterfactuals

Please note that we are not talking about errors (which reflect variability of a random nature), rather, we are talking about biases (systematic deviations from “rational” decision making)

Generating Probabilities: Case 1

DRAM Inc. Decision Research in Advertising and Marketing,

DRAM Inc., is a new product introduction consultant.

They advise clients on whether to go ahead with a new product introduction or not.

To simplify things, imagine that all they tell the client is one of two responses, “Introduce” and “Do not Introduce”.

The following is a compilation of their recommendations and results of whether the call was good or not.

Some of the following products are real, others or not. All results are for the purpose of this discussion only, and hence, are fictitious.

Store Brand Ketchup for AEB Inc.

DRAM Advice: “Do not Introduce”

Turned out to be a bad call. A competing store introduced its brand

shortly thereof, and succeed very well.

Store Brand Pet Food

DRAM Advice: “Do not Introduce”

Turned out to be a good call. A competing store introduced its brand

shortly thereof, and failed.

Touchless Trashcan

DRAM Advice: “Introduce”

Turned out to be a good call. Touchless trashcan was received very

well in the market.

Space Saver Vacuum Bags

DRAM Advice: “Introduce”

Turned out to be a good call. Received very well in the market.

R/C Hovercraft

DRAM Advice: “Introduce”

Turned out to be a bad call. Received poorly in the market.

Batteryless Flashlight

DRAM Advice: “Introduce”

Turned out to be a bad call. Received poorly in the market.

Residential Power-Ladder

DRAM Advice: “Introduce”

Turned out to be a good call. Received well in the market.

Water-out-of-thin Air

DRAM Advice: “Introduce”

Turned out to be a bad call. Received poorly in the market.

Satellite Auto-Tracking Device

DRAM Advice: “Do not Introduce”

Turned out to be a good call. Competitor’s product received poorly in

the market.

External USB Storage Devices

DRAM Advice: “Do not Introduce”

Turned out to be a bad call. Competitors’ products received well in

the market.

Overnight Contact Lenses

DRAM Advice: “Introduce”

Turned out to be a good call. Received well in the market.

Plasmatron Fuel Efficiency Booster*

DRAM Advice: “Introduce”

Turned out to be a good call. Received well in the market.

* Real new product idea, source Imagineering.com.

Ink Stripper

DRAM Advice: “Introduce”

Turned out to be a bad call. Received poorly in the market.

CD-ROM Business Card

DRAM Advice: “Introduce”

Turned out to be a bad call. Received poorly in the market.

New Product X

DRAM Advice: “Introduce”

What is the probability that DRAM’s call on New Product X will be correct? Note your answer in the response sheet provided.

Sources of Bias in Probability Estimates

Source of Bias: Anchoring and Adjustment

Biases in generating probabilities Question wording effect - Anchoring and Adjustment

The range of responses suggest a possible response, which is combined with one’s (soft) judgment.

Ideally, it should be independent of the range of responses.

Other question wording examples: How long/short was the movie? How likely/unlikely are we to succeed? How many products have you tried?

1 or 3 or 10 or other? 1 or 2 or 3 or other?

Case # 2: Market Research Recommendation

You are not sure if a particular new product will succeed in the market place.

You ask your market research firm to advice.

They sample of group of potential customers, and tell that the product will likely succeed.

What will you do? How will you proceed?

Case # 2b: Physician’s DilemmaPhysician is fairly certain that

patient does not have lung cancer, but orders x-ray anyway.

X-ray comes back positive.What should the physician do?What do you think the physicians

do?

Questions to Ask

What is the base-rate for cancer?What is the detection rate for the

test?What is the false-positive rate for

the test?Cancer Present Cancer Absent

Test Positive 0.8 = true positive (detection rate)

0.1 = false positive

Test Negative 0.2 = false negative 0.9 = true negative

Likelihood of Disease Possible ways of getting a “+”

The test says “+” and the person has the disease The test says “+” and the person does not have the disease.

Imagine 1000 people in the population, and 10 has the disease (1% base rate).

If you test all 1000 people: Of the 10 who have the disease, 8 will come out positive. Of the 990 who don’t have the disease, 99 will come out

positive. Therefore, 107 will test positive, although only 8

both test positive and have the disease. So, the chance that a positive test means

presence of the disease = 8/107, which less than 8%.

Source of Bias: Confusion of the Inverse

When people are asked what is the probability of disease given positive, they incorrectly equate that to the question of what is the probability of positive given the disease.

Likewise, when the manager is trying to determine the chance of real success given that the market research says so, he/she incorrectly equates that to chance that market research will correctly identify a success.

p(disease/positive) ≠ p(positive/disease) p(true success/MR success) ≠ p(MR success/true

success)

Generating Probabilities: Case 3Heart Disease Risk

Public health goal: get people to engage in heart-healthy life style.

Adoption of behavior is a function of:perceived vulnerability to the disease

So, often, the goal is to enable people to correctly judge their risk of disease.

How are Probability Estimates Generated?

Retrieved from memory.

Computed and constructed based on available information – on demand.

Probability Estimates – Retrieval View

Probabilities represent a long range frequency.

They are computed using both long range frequency and individual characteristics and stored in memory.

Retrieved on demand.

Constructionist View of ProbabilityQuestion activates contents in

memory.

People look at:What is cued.How easily it is cued.

Perceived Vulnerability to Heart DiseaseFour groups:

Eight risk increasing factorsEight risk decreasing factorsThree risk increasing factorsThree risk decreasing factors

What is the likelihood of heart disease?

Pattern of Findings

0

10

20

30

40

50

60

70

Three Risk Factors Eight Risk Factors

Risk IncreasingRisk Decreasing

Source of Bias: Computation and Ease of Retrieval

Probabilities are constructed/computed on demand.They are not simply retrieved.What is retrieved matters.

People use “ease of retrieval” as a cue.If I can easily generate the facilitating

(inhibiting) factors, the probability must be higher (lower), than when I cannot easily generate them.

Case # 3: Columbia Tragedy

After Challenger, several changes were instituted in NASA.

Then, Columbia disaster occurred in February, 2003.

The first question was, ‘why did NASA continue flying the Shuttle with a known problem…?’

Is this a good or a bad question to start with?

Source of Bias: Hindsight Bias

Likelihood of event happening is “p”, say 50%.

It is judged to be “p”, i.e., 50%.

Event happens.

People unconsciously “up” their probability of the event happening, “I really thought the event was going to happen”, i.e., 100%.

This is hindsight bias, or “knew-it-all-along” effect.

Hindsight Bias

An inability to discern truth from fact due to repeated hindsight reconstruction of facts and relationships between presumed causes and effects.

Applications?Why is this a problem?When is this not a problem?

Case # 4: Linda Problem

• Linda is 31, single, outspoken, bright, concerned with discrimination, social justice and against nuclear weapons/energy.

• She is most likely • (a) a bank teller or • (b) a bank teller who is active in the

feminist movement.

Source of Bias: Conjunction FallacyEvents that tend to occur together

than separately are thought to be more probable that each of the component events.

Why is this a bias?

Conjunction Fallacy

Bank Teller

Feminist Movement

Bank Teller and Feminist Movement

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

BT BT + FM

The combination bank teller plus active in feminist movement seems more representative of Linda’s past than just bank teller alone.

Case # 5: Heart Attack Probability

A health survey conducted in British Columbia. A representative sample of all ages and occupations were selected. Mr. F was included in the list by chance. Which of the following statements are more probable?

Mr. F has had one or more heart attacks.Mr. F has had one or more heart attacks, and

he is over 55 years old.

Source of Bias: Causal Conjunction Fallacy

When a presumed cause of an event is included in the description of the event, and people are asked to judge the likelihood of the event, they actually judge the likelihood of the event given the cause.

This often makes it more probable.

Case #6a: What the Team Thinks

You are looking to fill your decision tree, and it turns out you do not know what the probability for one of the outcome states is.

The probability relates to the chance that the product will succeed if you introduced it now.

So, you ask the rest of your team members and, the estimates are as follows:

0.1, 0.3, 0.3, 0.1, and 0.2

What will you do?

Case #6b: Birth Rate Problem

Two hospitals, one with 15 babies on average and other with 45 babies on average per month.

Which of the hospitals would you find a greater number of months for which more than 60% of the births are boys?

Larger Smaller Neither

Source of Bias: Small Sample Fallacy

Small samples are representative.

Small and large samples are equally reliable.

Small samples have similar distributions as large samples

Relevance to real managerial decisions?

Case #7: Comparing to Others

Often one generates probability estimates by locating ourselves with regard to others.

What would the typical investor do in my shoes?

How likely am I to succeed relative to the typical peer?

What would the typical competitor do in this situation?

Probability Estimates Relative to Others

-60

-40

-20

0

20

40

60

% m

ore

or le

ss th

an a

vera

ge p

erso

n

Drinking Problem Own Home Improved Salary Heart Attack*

Source of Bias: Self-Positivity Bias

People believe that they are more likely to experience good events and less likely to experience bad events compared to the average person.

Self-presentation Knowledge differences Ability to encode positive and negative

information Ability to recall positive and negative information

at the time of generating risk judgments. What does “average” other person mean anyway?

Sources of Bias in Probability Estimates - Summary

Anchoring and Adjustment Confusion of Inverse Retrieval versus computation/construction

and ease of retrieval Hindsight Bias Conjunction fallacy

Causal conjunction fallacy

Self-positivity bias Small sample fallacy