slides - RAMAS Ecological and Environmental Software

Report
Framing, biases, heuristics, the
normative/descriptive gap, and
ecologically rational choice
W. Troy Tucker
Applied Biomathematics, Inc.
[email protected]
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
A cognitive illusion
From Pinker 2002
Hunting-gathering lifestyle
• Deep time
– 1st tools: 2.5 mya
– Paleolithic: 1.5 mya
– Anatomically modern: 500 kya
– Fully modern: 40 kya
Axioms of “rational” choice
• Transitivity, ordering – A>B, B>C, A>C
• Monotonicity – more is better.
• Invariance, stochastic dominance – expected value
determines preference.
• Reduction, substitutability – probability calculus
doesn’t affect decisions.
• Continuity – p can be anything between 0 and 1.
• Finiteness – no infinite values.
Biases and heuristics
•
•
•
•
•
•
•
•
Anchoring and adjustment heuristic
Availability heuristic
Representativeness heuristic
Simulation heuristic
Base rate fallacy
Conjunction fallacy
Framing
Status quo bias
Availability heuristic
• Ease of recall (bias), vividness (bias)
– Megan’s law
– Violent crime
• declined 10 straight years, as Fox news style and
capability spread, 3-strikes laws, mandatory
sentences, etc.
Representativeness heuristic
• Regression to the mean (bias)
– Good sticks and bad carrots?
• Peak-end rule (bias)
– Recall average not sum
• Perceived happiness
• Conjunction fallacy (bias)
• Base rate fallacy (bias)
Conjunction fallacy
• Linda problem:
Linda is 31 years old, single, outspoken, and very bright. She majored in
philosophy. As a student she was deeply concerned with issues of
discrimination and social justice, and also participated in anti-nuclear
demonstrations.
Which is more likely?
P( A)  P( A  B)  P( B)
A. Linda is a bank teller.
B. Linda is a bank teller and is active in the feminist movement.
Base rate fallacy
• Another Linda problem:
Linda is 31 years old, single, outspoken, and very bright. She majored in
philosophy. As a student she was deeply concerned with issues of
discrimination and social justice, and also participated in anti-nuclear
demonstrations.
Which is more likely?
A. Linda is a bank teller.
B. Linda is active in the feminist movement.
Base rate fallacy
• Green cabs:
A cab was involved in a hit and run accident. 85% of cabs are green. 15% of
cabs are blue. A witness says the cab was blue. In a reliability test, the
witness correctly identifies the cab color 80% of the time. The witness is
wrong 20% of the time. What is the probability the cab involved in the
accident was blue, as the witness stated?
In 100 accidents: 85 are green, 15 are blue.
Witness sees 17 of 85 green cabs as blue (20% error).
Witness sees 12 of 15 blue cabs as blue (80% correct).
17/29 blue cabs seen in 100 accidents are really green.
60% chance witness was wrong.
Just because there are so many green cabs.
Base rate fallacy
• HIV:
About 0.01 percent of men with no risk behavior are infected with HIV (base
rate). If such a man has the virus, there is a 99.9 percent chance that the test
result will be positive (sensitivity). If a man is not infected, there is a 99.99
percent chance that the test result will be negative (false positive rate). What
is the chance that a man who tests positive has the disease?
Imagine 10,000 men who are not in any known risk category
(blood donors). One is infected (base rate) and will test positive
with practical certainty (sensitivity). Of the 9,999 donors not
infected, another one will also test positive (false positive rate).
Thus, two will test positive, one of whom actually has the disease.
Anchoring/Adjustment heuristic
(Tversky and Kahneman)
• Anchoring effect (bias)
– Excimer laser engineer salary?
• $99,790 vs. $68,860
• Effect size of anchor is 45%
• 13 trials over 10 yrs.
– Last 4 digits of SSN and number of docs in NY
• r=0.4
Framing (1)
• Asian disease problem
Imagine that the U.S. is preparing for the outbreak of an unusual Asian
disease, which is expected to kill 600 people. Two alternative programs to
combat the disease have been proposed. Assume the exact scientific estimate
of the consequences of the programs are as follows:
Program A: "200 people will be saved"
Program B: "there is a one-third probability that 600 people will be saved, and
a two-thirds probability that no people will be saved"
From Tversky and Kahneman 1981
Framing (2)
• Asian disease problem
Imagine that the U.S. is preparing for the outbreak of an unusual Asian
disease, which is expected to kill 600 people. Two alternative programs to
combat the disease have been proposed. Assume the exact scientific estimate
of the consequences of the programs are as follows:
Program C: “400 people will die"
Program D: "there is a one-third probability that nobody will die, and a twothirds probability that 600 people will die"
From Tversky and Kahneman 1981
Framing (3)
• Asian disease problem
Program A: "200 people will be saved"
Program B: "there is a one-third probability that 600 people will be saved, and
a two-thirds probability that no people will be saved"
Program C: “400 people will die"
Program D: "there is a one-third probability that nobody will die, and a twothirds probability that 600 people will die"
From Tversky and Kahneman 1981
Framing (4)
• Framing effects are widely observed
–
–
–
–
Wason selection task – details of the story
Bayesian reasoning – existence of % or decimals
Positive and negative affect – prospect theory
Endowment – discount vs. surcharge – prospect theory
• Some theories I like
– Cosmides - social contract theory
– Gigerenzer - natural frequencies
– Wang - group size signals kith and kin
Framing (4)
• Framing effects are widely observed
–
–
–
–
Wason selection task
Bayesian reasoning
Positive and negative affect
Endowment – discount vs. surcharge
• Some theories I like
– Cosmides - social contract theory
– Gigerenzer - natural frequencies
– Wang - group size signals kith and kin
Bayesian reasoning (1)
If a test to detect a disease whose prevalence is 1/1000 has a
false positive rate of 5%, what is the chance that a person found
to have a positive result actually has the disease, assuming that
you know nothing about the person’s symptoms or signs? ___%
Casscells et al. 1978 replicated in Cosmides and Tooby 1996
12-18% correct Bayesian reasoning
Bayesian reasoning (2)
If a test to detect a disease whose prevalence is 1/1000 has a
false positive rate of 50/1000, what is the chance that a person
found to have a positive result actually has the disease, assuming
that you know nothing about the person’s symptoms or signs?
1 out of ___.
51
___
Casscells et al. 1978 replicated in Cosmides and Tooby 1996
76-92% correct Bayesian reasoning
Framing (4)
• Framing effects are widely observed
–
–
–
–
Wason selection task
Bayesian reasoning
Positive and negative affect
Endowment – discount vs. surcharge
• Some theories I like
– Cosmides - social contract theory
– Gigerenzer - natural frequencies
– Wang - group size signals kith and kin
• Asian disease problem
Program A: "200 people will be saved"
Program B: "there is a one-third probability that 600 people will be saved, and
a two-thirds probability that no people will be saved"
Program C: “400 people will die"
Program D: "there is a one-third probability that nobody will die, and a twothirds probability that 600 people will die"
From Tversky and Kahneman 1981
Kith and Kin
(Wang 2007)
Framing (4)
• Framing effects are widely observed
–
–
–
–
Wason selection task – details of the story
Bayesian reasoning – existence of % or decimals
Positive and negative affect – prospect theory
Endowment – discount vs. surcharge – prospect theory
• Some theories I like
– Cosmides - social contract theory
– Gigerenzer - natural frequencies
– Wang - group size signals kith and kin
Prospect theory
Prospect theory
Prospect theory
Prospect theory
losses
Loss aversion
gains
Emotion and Reason
• Risk premium
• Ambiguity premium
• Brain processes them
differently
• When ambiguous,
assume worst-case
– Emotion and reason
A. Rustichini, Science 310, 1624 (2005)
Ellsberg paradox
Glimcher and Rustichini, Science 306, 447 (2004)
Ellsberg paradox
Glimcher and Rustichini, Science 306, 447 (2004)
Anticipatory skin conductance
Bechara et al., Science 275, 1293 (1997)
Ultimatum game
• Selfishness axiom
– utility maximization
• Homo economicus
– Definition of “rationality”
– What are the “rational” optimum strategies?
• Proposer: always offer the smallest possible amount
• Responder: always accept
Ultimatum game
• 15 years of research in 25 western societies
– More than 100 studies published
– Homo reciprocans:
• Cares about fairness and reciprocity
• is willing to change the distribution of material outcomes
among others at a personal cost to themselves
• Rewards those who act in a prosocial manner
• Punishes those who act selfishly, even when punishment costs
• Is this pattern a human universal?
– Designed by natural selection to engage in reciprocal
altruism
Ultimatum game
• Evolutionary anthropology
• 11 anthropologists and one economist
• Twelve countries on four continents and New Guinea
• Fifteen small-scale societies
–
–
–
–
Three foraging societies
Six slash-and-burn horticulturalists
Four nomadic herding groups
Two sedentary, small-scale agricultural societies
• All games played for real stakes
– Equal to one-day’s local wages
Ultimatum game
Ultimatum game
• Results
– The selfishness-based model always wrong
– Cross-cultural variability
• Economic organization explains much of this
– Market integration
– Payoff to cooperation
– Individual-level variables do not explain game behavior
– Experimental play reflects common interactional patterns
Ultimatum game
Neuroscience of risk perception
The brain has many domain-specific calculators
(Marr 1982; Barkow et al. 1992; Pinker 1997, 2002)
Information format triggers specific calculators
(e.g. Cosmides & Tooby 1996; Gigerenzer 1991)
Different calculators give contrasting solutions
Or calculate different components of total risk
(e.g. Glimcher & Rustichini 2004 and references therein)
List of mental calculators
(after Pinker 2002)
•
•
•
•
•
•
•
•
•
•
Language (grammar and memorized dictionary)
Practical physics (pre-Newtonian)
Intuitive biology (animate differs from inanimate)
Intuitive engineering (tools designed for a purpose)
Intuitive psychology (theory of mind, autism, deception)
Spatial sense (dead reckoner and mental maps)
Number sense (1, 2, 3, many)
Probability sense (frequentist Bayes)
Intuitive economics (reciprocity, trust, equity, fairness)
Mental database and logic (assertions linked with logical and
causal operators)
People are bad risk calculators
… or often said to be bad when
1.
2.
3.
4.
5.
6.
7.
8.
Presented with percentages, large numbers, or
single-event probabilities
Experts tell them the risk
Presented with incertitude (versus variability)
Risk is seen to be imposed
Risk is out of personal control
Rare events are observed - representativeness
When children are at risk
etc.
People are bad risk calculators
… or often said to be bad when
1.
2.
3.
4.
5.
6.
7.
8.
Presented with percentages, large numbers, or
single-event probabilities
Experts tell them the risk
Presented with incertitude (versus variability)
Risk is seen to be imposed
Risk is out of personal control
Rare events are observed - representativeness
When children are at risk
etc.
When risk is imposed
…People perceive more risk
Even when the risk is smaller than voluntary risks
Multiple mental risk calculators perceive risk
Some perceive risk of disease, death, economic cost
Some perceive risk of social contract violation
(e.g. Cosmides 1989, Guth 1995, Henrich et al. 2005)
Bilateral anterior insula: disgust (e.g. Sanfey et al. 2003)
The problem of “altruism”
• Kin-directed altruism
(Hamilton 1964, Haldane)
– Explains social insects, parenting, nepotism
• How can non-relatives cooperate?
Cooperation with non-relatives
Reciprocal altruism
(Trivers 1971, Axelrod & Hamilton 1981)
2 problems: how to evolve, how to maintain
• Evolve is still not well understood
Prisoner’s dilemma: Tragedy of the commons
• Maintaining cooperation requires 5 emotions (at least)
– Friendship, moralistic aggression, forgiveness, guilt/shame,
sympathy/gratitude
Risk of being cheated
• Wason selection task and logic
– Evidence of cheater detection module: patterned
violation of logical deduction
• (Cosmides & Tooby, Gigerenzer & Hug)
– Cheaters looked at longer, remembered better
• (Chiappe, Brown, Dow, Koontz, Rodriguez, & McCulloch
2004; Mealey, Daood, & Krage, 1996; Oda, 1997)
– Neuropsychology - Bilateral limbic system damage to
temporal pole and amygdala impairs detection
• (Stone, Cosmides, Tooby, Kroll, & Knight 2002)
Strong Reciprocity
• Ernst Fehr and Simon Gächter
– team earns money when all cooperate
– Punishers (moralistic aggression)
• Spend money to ensure freeloaders don’t prosper
• Note – this is “irrational”.
– People do pursue own self interest
• But, definition of “self interest” includes fairness,
equity, justice, prudence, generosity, etc.
Strong reciprocity (2)
• Human emotional constitution embraces
prosocial and altruistic notions of in-group
and out-group identification, and reciprocity
– A direct result of evolutionary history
• (Gintis 2005, Bowles and Gintis 2003)
– Moral principles are “evolved facts in the world”
• Evolved and transformed according to natural laws
Ecologically rational choice
• The normative/descriptive gap
– Panglossians vs. meliorists
• Stanovich and West 2000
– Evolutionary social sciences are panglossian
• Expect people are generally well designed.
• Deviation from a normative model signals deficiency of the
model or an ecological mismatch.
• Two categories of error, “pardonable errors by subjects and
unpardonable ones by psychologists” – Kahneman 1981.
• The real trick is to disentangle the two.
– Novel environmental factors
END
Anthropology of risk
6 Risks:
– Accidents
– Subsistence failure
– Disease
– Inter-group
competition (war)
– Cooperation failure
(free riders)
– Paternity
Pueblo Bonito in Chaco Canyon, NM
Economic Decisions and
Foraging Theory
• Harvesting resources from the environment
is a problems all organisms must solve
• Natural selection should lead to efficient
harvest and use of scarce resources
• Two early models
– Patch Choice
– Prey Choice
Prey choice
• Fine-grained environments
– Composed of randomly distributed prey
– Need to know which prey types to pursue
• Decision: pursue or not upon encounter
• Currency: long term net rate of energy capture
• Constraints:
– Search and handle mutually exclusive
– encounters sequential, random, and in proportion to
abundance
– No short-term effect of foraging on prey abundance
– Forager has complete information
Construction of foraging models
• Decisions - what we want to explain
– How long to stay in a patch, what subset of foods to pursue
• Currencies - what we assume to correlate with fitness
– This is what is maximized or minimized
– Time, net calories, calories per hour, protein per hour
• Constraints - what limits the forager
– Cognitive capability (rules of thumb)
– Physiological limitations - pigs can't fly
The Prey-choice model
ei = average net energy from prey type i upon
encounter
hi = averge handling time per encounter for prey type i
li = abundance of prey type i
pi = probability of pursuing prey type i upon encounter
Prey choice model test
• The Ache of Paraguay
– First contacted around 1978
– About 200 individuals
• 15-50 people in a foraging group
Data
• data gathered about Ache foraging
From Hawkes et al. 1982
Result
From Hawkes et al. 1982
Time discounting
• Equity premium puzzle
– 50:50 $50k or $100k = $51,209?!?
– Glimcher and Kable - hyperbolic function

similar documents