Biases and Errors

1
Lecture 22

• Biases and Errors

2
Assignment 5

• 1. Real world population sampled Women admitted
  to participating hospitals between 1964 and
  1966.Population of interest Humans, women,
  women taking the pill - could all be defended I
  think.Causal hypothesis The pill causes blood
  clots.2. Sample data.26/58 45 with blood
  clots taking pill.10/116 10 without blood
  clots taking pill.3. Design of
  experimentProspective

3

• 4. Random samplingIs there a reason to think
  that those taking the pill might have some other
  factor that leads to blood clots?Usual problems
  for prospective studies - biases, faulty
  memories as well as the general worries with
  self-selection.5. Evaluating hypothesis45
  -1410 -10Non-overlapping. Evidence of
  correlation.
• Estimated strength 59,116. SummaryThere is
  moderate support for causation. The strong
  correlation is good evidence. This is weakened by
  it being a prospective study.

4
Biases

• Economists assumed that humans acted to maximize
  expected utility homo economicus.
• Kahneman and Tversky showed this was false, which
  led to the field of behavioural economics.

5
Cognitive Bias

• Why arent people maximizing utility?
• What are they doing and why?
• Which heuristics are they using?
• Under what conditions is the heuristic useful?
• Bounded rationality Rationality under
  computational constraints.

6
Gamblers Fallacy

• Is a Head more likely after 10 Tails than after
  10 Heads?
• A Heads has a 50 chance no matter what happened
  before it.
• The frequency of Heads goes to 0.5 in the long
  run, but not by the outcomes correcting
  themselves.
• Casinos display the previous numbers on the
  roulette wheel because people who keep score lose
  more money.

7
The Gamblers Fallacy only applies to gambling
machines

• In most normal contexts, the gamblers fallacy
  is not a fallacy.
• If a lion hasnt been seen near the village in a
  week, is it more likely to be seen?
• Yes, as it is likely to patrol its whole
  territory over time.

8
Hot-Hand

• Most basketball fans believe that players are
  more likely to score a basket if they have just
  scored one than if they missed.
• Vallone and Tversky (1985) found no evidence that
  this was the case.
• http//thehothand.blogspot.com/

9
Ellsbergs Paradox

• An urn contains 30 red balls and 60 other balls
  that are black or yellow.
• Decision theory says that if A is preferred to B,
  C is preferred to D.
• But people choose A over B and D over C.

10
Explanation

• Uncertainty can be avoided by picking A, or D.
• So people seem to be paying a price to avoid
  uncertainty.

11
Allais Paradox
Gamble A Gamble B
Choosing Risky A and Safe B violates the axioms
of decision theory.
Many choose to do so because they are risk averse.
12
Risk Aversion

• Expected value of 10 if Heads, 0 if Tails
  5.
• If you prefer a certain 5 to the bet, you are
  risk averse.
• Risk aversion is different from the diminishing
  marginal utility of money.
• If you prefer the bet, you are risk-loving.
• Risk-loving behaviour
• Casinos
• Risk-averse behaviour
• Insurance
• Some argue that risk-aversion is irrational. The
  following cases certainly are.

13
Framing

• 600 people are expected to be killed by a
  disease.
• Plan A saves 200
• Plan B has a 1/3 chance of saving all 600 and a
  2/3 chance of saving no-one.

14
Framing

• 600 people are expected to be killed by a
  disease.
• Plan A saves 200
• Plan B has a 1/3 chance of saving all 600 and a
  2/3 chance of saving no-one.
• 72 of subjects said A

15

• Plan C leads to 400 deaths
• Plan D has a 1/3 chance of no deaths and a 2/3
  chance of 600 deaths.

16

• Plan C leads to 400 deaths
• Plan D has a 1/3 chance of no deaths and a 2/3
  chance of 600 deaths.
• 78 preferred D.

17

• People are more risk averse when it comes to
  losses loss aversion.
• Compare 5 or a 50 chance of 10
• -5 or a 50 chance of -10
• People are more willing to take a gamble to avoid
  a sure loss than to take a gamble for a bigger
  gain.
• By framing the question in terms of losses,
  people become more risk-averse.

18
Loss aversion and business

• Losses are more bad than gains are good.
• A 5 surcharge matters more than a 5 discount.
• This is also why trial periods are offered.
• Goods owned are considered more valuable than
  identical goods not owned.
• If you dropped a dollar, youd do back to class
  for it.
• But would you walk back to class if you were paid
  a dollar? NH

19
Anchoring

• Trivia How much does the Statue of Liberty
  weigh? The first answer suggested has a big
  impact on the answer agreed on.
• What percentage of African nations are in the UN?
  More or less than 45? More or less than 65
• Those asked if it was more or less than 45 gave
  lower estimates.
• This could be rational.

20

• Subjects were asked to write the last two digits
  of their social security number.
• Then they were asked to make mock bids on various
  items.
• The half of the audience with higher two-digit
  numbers would submit bids that were between 60
  percent and 120 percent more.
• Definitely not rational!

21

• Do more words start with K or have K as a third
  letter?

22

• Do more words start with K or have K as a third
  letter?
• 3 times as many words have K as a third letter
• But we answer the question by thinking of words
  that start with K or have K as a third letter.
• Its easier to search for words by their starting
  letter.

23
Availability heuristic

• If you can think of it, it must be important.
• My granny smokes like a chimney and shes 96.
• A group that was asked to imagine Ford winning
  the 1976 election judged it more likely to happen.

24
Base rate fallacy

• A city has 100 terrorists and 1 million
  non-terrorists.
• A camera triggers an alarm when it sees a
  terrorists and is right 99 of the time.
• Given that the alarm has been triggered, what is
  the probability that the person is a terrorist?
• 1/100

25
Conjunction Fallacy

• Linda was a bright outspoken philosophy major was
  deeply concerned with issues of social justice.
• Is it more likely that Linda is
• 1. a bank teller
• 2. a bank teller and active in the feminist
  movement?
• 85 say 2. Why?

26
Representativeness heuristic

• Explains the base-rate and conjunction fallacy.
• What people respond to is how representative the
  option is
• A feminist bank teller is more representative of
  what we would expect from the description of
  Linda.

27

• One way of spelling this out is that 2 is more
  strongly confirmed by the evidence than 1.
• 2 starts with a lower prior probability. The
  description of Linda significantly increases this
  value above its prior.
• Degree of confirmation P(HE) - P(H)

28
Reversion to mean

• Unusual events tend to be followed by average
  events.
• Pilots who do poorly on one flight will usually
  do better the next.
• Pilots who do well on one flight will usually do
  worse on the next.
• Pilots who are rewarded after doing well then do
  worse.
• Pilots who are punished after doing badly then do
  better.

29
Wason selection task

• Which cards must be turned over to test the
  following hypothesis?
• If a card has an even number on one side, it has
  a primary colour on the other.

3
8
30

• Only 8 and pink are correct.
• Turning 3 does not test the hypothesis.
• Turning red does not test the hypothesis. (It
  tests If primary then even)
• About 20 of subjects get this right.

31
Logical Structure

• Test If P then Q
• You must turn over P and -Q.

-P
P
Q
-Q
32
Why do people do so badly in the Wason test?

• Failure to understand the conditional
• If vs. only if.
• If and only if a card has an even number on one
  side, it has a primary colour on the other.
• This is tested by turning all the cards.
• Only if a card has an even number on one side, it
  has a primary colour on the other.
• This is tested by turning 3 and pink.

33
Social contract Wason test

• If you are drinking alcohol then you must be over
  18.
• Easy. Why?
• Weve innate cheater-detectors?

17
22
Beer
Coke
34
Precautionary rule Wason test

• If you work in a lab, where rubber gloves.
• Easy. Why?
• Weve innate systems to manage hazards? No. Weve
  learnt.

Works in lab
Doesnt Work In lab
Wearing rubber gloves
Not wearing rubber gloves
35
Mind-control

• Coke is rated higher when consumed from a cup
  bearing the brand logo rather than from an
  unmarked cup (McClure et al., 2004)
• a slice of turkey is rated higher if thought to
  come from a popular brand rather than an
  unpopular one (Makens, 1965)
• Perrier is preferred to Old Fashioned Seltzer
  when the beverages are consumed with the labels
  showing, but not otherwise (Nevid, 1981)
• preference for ones favorite beer vanishes if
  the labels on the beers being compared are
  removed (Allison Uhl, 1964)
• describing the protein of nutrition bars as soy
  protein causes them to be rated as more grainy
  and less flavorful than when the word soy is
  not included (Wansink, Park, Sonka, Morganosky,
  2000)
• bitter coffee seems less so if consumers are
  repeatedly misinformed that it is not bitter
  (Olson Dover, 1978)
• strawberry yogurt and cheese spreads are liked
  more if labelled full-fat than if labelled
  low-fat (Wardle Solomons, 1994)
• and, intriguingly, people eat more vanilla ice
  cream if it is accurately labelled high fat
  than if it is labelled low fat (Bowen,
  Tomoyasu, Anderson, Carney, Kristal, 1992).

36
You loved this course