Neuroeconomics: The Neurobiology of Decision-Making
Ifat Levy Section of Comparative Medicine Department of Neurobiology Interdepartmental Neuroscience Program Yale School of Medicine
Harnessing eHealth and Behavioral Economics for HIV Prevention and Treatment April 2012 Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions
Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions
Neuroeconomics
NeuroscienceNeuronal MentalPsychology states “asEconomics if” models architecture
Abstraction Neuroeconomics
Behavioral Economics
Neuroscience Psychology Economics
Abstraction Neuroscience functional MRI
VISUAL STIMULUS functional MRI: Blood Oxygenation Level Dependent signals
Neural Changes in oxygen Change in activity consumption, blood flow concentration of and blood volume deoxyhemoglobin
Change in Signal from each measured point in space at signal each point in time t = 1 t = 2 t = 3 t = 4 t = 5 t = 6
dorsal
anterior posterior lateral
ventral dorsal
anterior medial posterior
ventral
Anterior The cortex Cingulate Cortex (ACC)
Medial Prefrontal Cortex (MPFC) Posterior Cingulate Cortex (PCC) Ventromedial Prefrontal Cortex (vMPFC)
Orbitofrontal Cortex (OFC) Sub-cortical structures fMRI signal
• Spatial resolution: ~3x3x3mm3 Low • Temporal resolution: ~1-2s Low • Number of voxels: ~150,000 High • Typical signal change: 0.2%-2% Low • Typical noise: more than the signal… High But…
• Intact human brain • Behaving human • Whole brain • Non-invasive Neuroeconomics
Cognitive Neuroscience Behavioral Economics
Neuroscience Psychology Economics
Abstraction New challenge: how do you make sense of such huge amounts of data?? Neuroeconomics
Mechanistic constraints of the human brain
Cognitive Behavioral Neuroscience Economics
Economic models as normative theory Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions certainty risk ambiguity
100% 50% unknown partial full full relief remission remission Choose one:
A B
Risky OR Ambiguous
Most people choose A, implying that B has fewer red than blue chips: red < blue A B
Risky OR Ambiguous
Most people choose A, implying that B has fewer blue than red chips: blue < red The Ellsberg paradox: a bag cannot have fewer red chips and fewer blue chips at the same time Non-certain outcomes
• Risk – probabilities of different outcomes are known
• Ambiguity – probabilities of different outcomes are not known
• Partial ambiguity – partial information
known probability risk aversion high probability low low probability high reward reward
$40
unknown probability ambiguity aversion known probability unknown probability low reward high reward
$110$0 $0$110 Value of risk and ambiguity
Risk and ambiguity affect the subjective value of an option in very different ways Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions
Research Question neural representation of value
multiple systems single system
reward reward punishment punishment immediate delayed cognitive cognitive delayed immediate emotional . . emotional .
Research Question neural representation of value
multiple systems single system
ambiguity vs. risk ambiguity & risk
x,y,0.5,5 Experimental design Experimental design
Amount Probability
Ambiguity level OR: - Winning color - $5 Parametric design Real bags One trial played for real money Gain-risk trials
p = 0.75 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p = 0.50 p (chose (chose risky) p
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p = 0.50
p (chose (chose risky) p p = 0.38
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p = 0.50
p (chose (chose risky) p p = 0.38 p = 0.25
Amount [$] $5
subject 1 Gain-risk trials
p = 0.75 p = 0.50
p (chose (chose risky) p p = 0.38 p = 0.25 p = 0.13
Amount [$] $5
subject 1 Gain ambiguity trials
A = 0.25 p (chose (chose ambiguous) p
Amount [$] $5
subject 1 Gain ambiguity trials
A = 0.25 p (chose (chose ambiguous) p
Amount [$] $5
subject 1 Gain ambiguity trials
A = 0.25 A = 0.50 p (chose (chose ambiguous) p
Amount [$] $5
subject 1 Gain ambiguity trials
A = 0.25 A = 0.50 A = 0.75 p (chose (chose ambiguous) p
Amount [$] $5
subject 1 Behavioral model MaxMin, Gilboa and Schmeidler 1989
subjective value
ambiguity level − β A α ( p ) · V risk 2 preference amount probability ambiguity aversion
stochastic choice model p (chose lottery) α
= 0.55, S1: gains Amount [$] β = 0.89
α
= 0.58, S2: gains Amount [$] β = - 0.03
0.75 A = 0.50 A = 0.25 A = 0.13 = p 0.25 = p 0.38 = p 0.50 = p 0.75 = p S1: gains S1: losses
α = 0.55, β = 0.89 α = 0.58, β = -0.04
p = 0.75
p = 0.50 p = 0.38 p = 0.25 p = 0.13 A = 0.25 A = 0.50 p (chose p (chose lottery) A = 0.75
Amount [$] Amount [$] Losses Ambiguity aversion Risk aversionRisk Risk aversion Gains Gains
Ambiguity aversion Risk aversion Losses Gains
A ( p − β )⋅V α 2
under risk subjectivevalue under ambiguity … … time Subjective value under ambiguity
ACC / MPFC
R L
caudate posterior amygdala cingulate
19 subjects, random effect analysis P<0.002 P<0.0001 Subjective value under risk
ACC / MPFC
R L
caudate posterior amygdala cingulate
19 subjects, random effect analysis P<0.01 P<0.001 Unique areas for SV under ambiguity?
PCC amygdala
No… %signal change %signal change
risk ambiguity risk ambiguity Research Question neural representation of value
multiple systems single system
ambiguity vs. risk ambiguity & risk
x,y,0.5,5 Uncertainty Summary 1
• High variability in risk and ambiguity attitudes across individuals • Areas in MPFC and striatum represent subjective value under both risk and ambiguity
Can attitudes towards risk and ambiguity explain phenomena like risk-taking in adolescents and overeating in obese individuals?
Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions
Adolescents take risks
• 200% increase in morbidity and mortality rates in adolescence compared to childhood (Dahl, 2004)
• Adolescents are physically healthier and stronger than both children and adults (Dey et al., 2004)
• Increase mostly attributed to risky behaviors: car accidents, alcohol and substance abuse, violence, eating disorders, unsafe sex (Reyna and Farley, 2006)
• Not due to flawed reasoning capabilities, poor decision- making skills or failure to understand the consequences of their actions (Reyna and Farley, 2006)
Subjects
Age \ Female Male Total Gender 12-17 17 16 33 21-25 18 16 34 30-50 17 15 32 65-90 18 17 35 Total: 70 64 134
In collaboration with Paul Glimcher
Adolescents vs. adults Uncertainty Summary 2
• Adolescents are more risk averse, but less ambiguity averse than adults • Young organisms need to learn about their world
Do people treat risk and ambiguity similarly in different domains?
In collaboration with Terri Fried Overview
• Introduction to neuroeconomics • Decision under uncertainty – Brain and behavior – Adolescent behavior – Medical decisions
“You were involved in a car accident and as a result suffered traumatic brain injury. You were immediately rushed to the nearest hospital and were informed by the doctor that without immediate treatment you will not survive.”
Gains and losses in medical decisions
Gains: cognitive improvement
No effect Slight improvement Moderate improvement Major improvement Recovery
Worst outcome Best outcome
Major improvement = Mild cognitive disability: mild memory impairment resulting in forgetting some appointments, forgetting people’s names, needing a list to do food shopping No effect
Slight improvement or
recovery Major improvement
Slight improvement or
No effect Gains and losses in medical decisions
Losses: headache as an adverse effect
Critical headache Severe headache Moderate headache Mild headache Recovery
Worst outcome Best outcome
Moderate headache: improves but does not resolve with acetaminophen (Tylenol); requires you to lie down occasionally to relieve pain; occurs a couple of times a week. moderate headache
Mild headache or
recovery Severe headache
Mild headache or
recovery Decision under risk 1 Positive outcomes 1 Negative outcomes 0.9 0.9
0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 Proportionof choices lottery 0.1 0.1 0 0 0.13 0.25 0.38 0.5 0.75 0.13 0.25 0.38 0.5 0.75 Outcome probability Outcome probability Money Medical N = 29 Decision under ambiguity 1 Money 1 Medical
0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3
Proportionof choices lottery 0.2 0.2 0.1 0.1 0 0 gain loss improvement adverse effect
Ambiguity level: 0 0.25 0.50 0.75 N = 29 Uncertainty Summary 3
• Ambiguity aversion was observed both under gains and under losses when making medical decisions
Summary • Economic models can be used to make sense of neural data • High variability across subjects in attitudes towards risk and ambiguity • Adolescents are more risk averse and less ambiguity averse than adults • Ambiguity aversion in medical decision- making for both positive and negative outcomes
Acknowledgements
Lab Collaborators Sarah Abdallah Paul Glimcher, NYU Jennifer Fanning Aldo Rustichini, Minnesota U Ellen Furlong Agnieszka Tymula, NYU Patrick Kenney Amy Roy, Fordham Genny Ladiges Terri Fried, Yale Kirk Manson Scott Huettel, Duke Helen Pushkarskaya Linda Mayes, Yale Lior Rosenberg Belmaker Michael Crowley, Yale Lital Ruderman Ashley Gearhardt, Yale Sana Samnani Eric Jackson, Yale Jeannie Tran Daniela Schiller, Mount Sinai Zhihao Zhang
Funding NIA, Pepper Center
Sense of incompetence (Heath and Tversky, 1991)
People prefer to bet on events in their field of expertise, even when they judge the probabilities as equal Comparative ignorance (Fox and Tversky, 1995)
How much will you pay for playing the lottery?
Within subject Between subject > =
More recent study: Ambiguity aversion is reduced but not abolished (Chow and Sarin 2001) Informed opponent (Kuhberger and Perner, 1991)
Subjects chose the ambiguous option more when the person who filled the bag was a partner than when it was an opponent Experimental design Risk
Ambiguity
Real bags!
Experimental design
• Subjects were endowed with $125 • Gain and loss trials • Choice between a lottery and a certain amount (±$5) • 3 ambiguity levels: 0.25, 0.5, 0.75 • 5 risk levels: 0.75, 0.5, 0.38, 0.25, 0.13 • 5 outcome levels: ±$5, ±$8, ±$20, ±$50, ±$125 • 320 trials • 1 trial randomly selected and played for real money Behavioral model MaxMin, Gilboa and Schmeidler 1989
subjective value
ambiguity level − β A α ( p ) · V risk 2 preference amount probability ambiguity aversion
stochastic choice model S2: gains S2: losses
α = 0.58, β = -0.03 α = 0.86, β = 0.72
p = 0.75
p = 0.50 p = 0.38 p = 0.25 p = 0.13 A = 0.25 A = 0.50 p (chose p (chose lottery) A = 0.75
Amount [$] Amount [$] Subjective value in ambiguity defined regions
MPFC striatum
%signal change %signal change
risk ambiguity risk ambiguity Subjective value in risk defined regions
MPFC striatum
%signal change %signal change
risk ambiguity risk ambiguity Adolescents take risks
• 200% increase in morbidity and mortality rates in adolescence compared to childhood (Dahl, 2004)
• Adolescents are physically healthier and stronger than both children and adults (Dey et al., 2004)
• Increase mostly attributed to risky behaviors: car accidents, alcohol and substance abuse, violence, eating disorders, unsafe sex (Reyna and Farley, 2006)
• Not due to flawed reasoning capabilities, poor decision- making skills or failure to understand the consequences of their actions (Reyna and Farley, 2006)
And in the brain…
• Gray matter maturation processes in PFC and striatum continue into adolescence (Giedd et al., 1996, 1999, 2004) • Frontal increase in white matter occurs late and extends into adulthood (Fuster, 2002) • Structural atrophy and decline in dopamine receptors in striatum and PFC in aging (Backman et al., 2000; Volkow et al., 1998) • Altered striatal activation during gain anticipation in adolescents compared to adults (Ernst et al., 2005; Galvan et al., 2006; Bjork et al., 2004) • Reduction in activation in striatal areas during loss anticipation in older adults (Samanez-Larkin et al., 2007).
Adolescents vs. adults Controls Controls Controls “Cognitive” blocks No effect = The treatment failed. You end up in a vegetative state. Slight improvement = Severe cognitive disability: severe memory impairment resulting in inability to recognize your loved ones. Moderate improvement = Moderate cognitive disability: moderate memory impairment resulting in inability to work and participate in leisure activities such as playing cards or doing crossword puzzles. Major improvement = Mild cognitive disability: mild memory impairment resulting in forgetting some appointments, forgetting people’s names, needing a list to do food shopping. Recovery = return to your initial cognitive ability prior to the accident. “Headache” blocks Recovery = successful treatment with no side effects. Mild headache: responds to acetaminophen (Tylenol); does not interfere with daily activities; occurs a couple of times a week. Moderate headache: improves but does not resolve with acetaminophen (Tylenol); requires you to lie down occasionally to relieve pain; occurs a couple of times a week. Severe headache: not responsive to acetaminophen (Tylenol); requires stronger pain medication, which does not fully relieve pain; requires you to lie down frequently to relieve pain; occurs daily. Critical Headache: Severe headache (as above) accompanied by other symptoms, such as nausea and vomiting.