Neuroeconomics of Games and Decisions* Colin Camerer, Caltech

1
Neuroeconomics of Games and Decisions Colin
Camerer, Caltech  

• Neuroeconomics
• Grounding micro-economics in details of neural
  activity
• Part of behavioral economics (using psychology to
  inform theories of rationality limits)
• Part of experimental economics (new techniques)
• Part of neuroscience (higher order cognition)
• Neuroeconomics Camerer, Loewenstein, Prelec J
  EconLit (85 pp), Scan J Econ (25 pp), Why
  economics needs brains

2
Collaborators

• Caltech Meghana Bhatt, Ming Hsu, Ralph Adolphs,
  Cedric Anen, Steve Quartz
• Iowa Dan Tranel
• Baylor Brooks King-Casas, Damon Tomlin, Read
  Montague

3
Three directions in neuroeconomics

• I. Support for rational-choice models
• Belief neurons
• Expected-value neurons
• Monkey shopping satisfies GARP
• II Support for behavioral alternatives
• Loss-aversion in monkey shopping
• Learning in trust games
• Ambiguity vs. risk (Knight, Ellsberg)
• III New concepts
• Equilibrium as a state of mind
• Neural correlates of strategic IQ
• Biological basis of demand

4
Economically-important regions of the human brain
5
Cingulate (yellow), orbitofrontal (pink),
amygdala (orange), somatosensory
(green), insula (purple)
6
Important facts about the brain

• Functional modularity
• but plastic esp. in childhood
• Behavior depends on circuits
• Human brain is primate brain neocortex
• Language, social organization (institutions)
• Infants, fraternity parties show similarity
• Many biological functions are automated
  conscious attention is scarce (flicker paradigm)

7
I Rational choice in the brainMidbrain neurons
anticipate reward (L), encode value function V(.)
learning (R) (Schultz, Dayan, Montague Sci 97)
8
Neuron firing rates (y axis) encode expected
value (x-axis) (Glimcher)
9
Monkeys play mixed equilibrium as humans do
(Dorris-Glimcher Neuron 04)
10
Capuchin monkeys respond to prices (Keith Chen et
al 05)
11
II Behavioral economics in the brain

• Monkey choices are sensitive to reference points
• Reference point (initial food reward endowment)
• 1 2 1 2
• Outcome 1 1 (1,2)
  (1,2)
• Choice 79 21 71
  29

12
Design goal Link stimuli with unobserved
parametric processes/variables with circuitry
0-step thinking 1-step thinking Equilibrium
Cbr(B) w(red)-P(red)
13
Overview of fMRI
14
Data transformations
Statistical parametric map (SPM)
Image time-series
Design matrix
Kernel
General linear model
Realignment
Smoothing
Statistical inference
Normalisation
p lt0.05
Template
Parameter estimates
15
Ambiguity Aversion (with Ming Hsu et al)

• This material is in review and cannot be publicly
  circulated at this time.

16
III New ideas

• Limited planning in bargaining?
• limited steps of thinking
• Equilibrium as a state of mind
• Biological bases of demand

17
Rubinstein-Stahl alternating offershrinking-pie
bargaining

• 1 offers division of 5 ------------? accept
• ?
• 2 offers division of 2.50 -----------?
  accept
• ?
• 1 offers division of 1.25 ------------?
  accept
• ?
• (0,0)

18
Limited planning in bargaining (Science, 03)
19
Cognitive hierarchy thinking in games (Camerer,
Ho, Chong, QJE 04)

• Step 0 players choose randomly
• Step k players have beliefs gk(h)
• Step k players choose si(k) argmax s Sh
  gk(h)pi(s,s(h))
• One-step-below gk(k-1)1
• Nagel (1995), Stahl-Wilson (1995),
  Costa-gomes-Crawford-Broseta (2001)
• Nornalized overconfidence gk(h) gk(h)/Shk-1
  gk(h)
• gk(h) 0 for hgtk
• Link to hierarchical QRE (Palfrey-Rogers-Camerer,
  on this computer)

20
Limited equilibrationBeauty contest game

• N players choose numbers xi in 0,100
• Compute target (2/3)(? xi /N)
• Closest to target wins 20

21
(No Transcript)
22
Neural correlates of iterated belief
(Bhatt-Camerer GEB in press)

• 8 dominance-solvable games.
• C, B, 2B in random order for each game
• Paid for choice (x.30) or accuracy B, 2B (15)
  against live opponent outside the scanner.
  (Enables measure of scanner on behavior.)
• N16 Caltech community students

23
Example 1-step (easy) game B dominates A. L-R
payoff separation allows eye tracking
24
Conformity to equilibrium There are many
nonequilibrium trialsNote C matches 2B more
often than B matches 2B
25
Equilibrium is a state of mindExpected
reward theory of mind
(in equilibrium ?) (?
out-of-equilibrium)
26
Subject complaining after an experiment (Zamir,
2000)
27
Ultimatum games This is your brain on
unfairness(Sanfey, Rilling et al, Sci 13 March
03)
28
Consistent 2nd-order beliefs (c2b) vs
inconsistent (false) (c?2b)

• 2nd-order belief consistency differential
  activates dorsolateral prefrontal cortex (DLPFC)
• DLPFC also seen in ultimatum games after low
  offers (intentions matter)
• DLPFC is part of 2nd-order belief circuitry?

29
Deactivation in insula and high strategic IQ

• Strategic IQ (x-axis) How much you earn from
  choices beliefs
• Correlated (-) with activity in L insula in
  choice task
• ? Are overly self-focussed people poor strategic
  thinkers?

30
Insula and low strategic IQ

• Strategic IQ (x-axis) How much you earn from
  choices beliefs
• Correlated (-) with activity in L insula in
  choice task
• ? Are overly self-focussed people poor strategic
  thinkers?

31
Correlates of higher strategic IQ
32
New ideas Biological basis for demand

• Economics takes demand as given. But
• Influence of advertising
• Familiarity and habit formation (tight playlist
  radio stations)
• Imitation of movie stars/TV shows
• LA Law? boom in law school applications
• Sense-making drive? demand for closure ?
  lawsuits
• Media If it bleeds, it leads, NASCAR races
• Does the amygdala control the TV remote?
• Addiction Is golf or shoe-shopping like heroin?
• Labor market discrimination (Phelps et al)?
• Unfamiliar black faces activate white student
  amygdalae

33
Conclusions

• I Rational choice processes in the brain
• Monkey belief neurons, games, shopping
• II Behavioral economics in the brain
• Monkey loss-aversion
• Ambiguity activates amygdala-OFC, risk striatum
• Lesion patients with OFC are rationalfor the
  wrong reason?
• III New ideas from neureconomics
• Limited strategic thinking? equilibrium as a
  state of mind
• Skill (strategic IQ) correlated with precuneus,
  caudate,
• correlated - with insula
• Biological basis of demand

34
Activation in cingulate cortex spindle cell
density