84
nstitute for Theoretical Physics and Mathema Tehran January, 2006 Value based decision making: behavior and theory

Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Embed Size (px)

Citation preview

Page 1: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Institute for Theoretical Physics and MathematicsTehran

January, 2006

Value based decision making: behavior and theory

Page 2: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Greg Corrado

Leo Sugrue

Page 3: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

SENSORY INPUT

DECISION MECHANISMS

ADAPTIVE BEHAVIOR

low level sensory analyzers

motor output structures

Page 4: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 5: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

SENSORY INPUT

DECISION MECHANISMS

ADAPTIVE BEHAVIOR

low level sensory analyzers

motor output structures

REWARD HISTORY

representationof stimulus/action value

Page 6: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

How do we measure value?

Herrnstein RJ, 1961

Page 7: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

The Matching Law

Ch

oic

e F

ract

ion

Page 8: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Behavior: What computation does the monkey use to ‘match’?

Theory: Can we build a model that replicatesthe monkeys’ behavior on the matching task?How can we validate the performance of the?

model? Why is a model useful?

Physiology: What are the neural circuits and signal transformations within the brain

that implement the computation?

Page 9: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

An eye movement matching task

Bai

ting

Fra

ctio

n1:1

6:1

1:6

6:1

1:2

2:1

1:2

Page 10: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Dynamic Matching Behavior

Page 11: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Rewards

Dynamic Matching Behavior

Page 12: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

ResponsesRewards

Dynamic Matching Behavior

Page 13: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Relation Between Reward and Choice is Local

ResponsesRewards

Page 14: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

How do they do this?

What local mechanism underlies the monkey’s choices in this game?

To estimate this mechanism we need a modeling framework.

Page 15: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Linear-Nonlinear-Poisson (LNP) Models

of choice behavior

Strategy estimation is straightforward

Page 16: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

How do animals weigh past rewards in determining current choice?

Estimating the form of the linear stage

Page 17: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 18: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 19: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 20: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

How is differential value mapped onto the animal’s instantaneous probability of

choice?

Estimating the form of the nonlinear stage

DifferentialValue

Page 21: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Differential Value (rewards)

Monkey F Monkey G

Pro

babi

lity

of

Cho

ice

(red

)

Page 22: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Our LNP Model of Choice Behavior

Model Validation• Can the model predict the monkey’s next choice?• Can the model generate behavior on its own?

Page 23: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Can the model predict the monkey’s next choice?

Page 24: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Predicting the next choice: single experiment

Page 25: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Predicting the next choice: all experiments

Page 26: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Can the model generate behavior on its own?

Page 27: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Model generated behavior: single experiment

Page 28: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Distribution of stay durations summarizes behavior across all experiments

Stay Duration (trials)

Page 29: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Model generated behavior: all experiments

Stay Duration (trials)

Page 30: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Model generated behavior: all experiments

Stay Duration (trials)

Page 31: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

1. Explore second order behavioral questions

2. Explore neural correlates of valuation

Ok, now that you have a reasonable model what can you do with it?

Page 32: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

1. Explore second order behavioral questions

2. Explore neural correlates of valuation

Ok, now that you have a reasonable model what can you do with it?

Page 33: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

0000111110011choice history:

Surely ‘not getting a reward’ also has some influence on the monkey’s behavior?

0000010100001reward history:

Choice of Model Input

Page 34: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

0000111110011choice history:

0000010100001reward history:

the value of an unrewarded choice

hybrid history: 0000010100001

Choice of Model Input

Page 35: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

• Systematically vary the value of • Estimate new L and N stages for the model• Test each new model’s ability to

a) predict choice and b) generate behavior

hybrid history: 0000010100001

Can we build a better model by taking unrewarded choices into account?

Page 36: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Value of Unrewarded Choices () Value of Unrewarded Choices ()

Predictive Performance Generative Performance

Unrewarded choices: The value of nothin’

Page 37: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Value of Unrewarded Choices () Value of Unrewarded Choices ()

Predictive Performance Generative Performance

S tay

Du r

atio

n H

i sto

gram

Ov e

r lap

(%

)

Unrewarded choices: The value of nothin’

Page 38: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Contrary to our intuition inclusion of information about unrewarded choices does not

improve model performance

Choice of Model Input

Page 39: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Optimality of Parameters

Page 40: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Weighting of past rewards

Is there an ‘optimal’ weighting function to maximize the rewards a player can harvest in this game?

Page 41: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 42: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 43: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 44: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 45: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 46: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 47: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 48: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 49: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 50: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 51: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 52: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 53: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 54: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

• The tuning of the 2 (long) component of the L-stage affects foraging efficiency. Monkeys have found this optimum.

Weighting of past rewards

• The 1 (short) component of the L-stage does not affect foraging efficiency. Why do monkeysoverweight recent rewards?

• The tuning of the , the nonlinear function relating value to p(choice) affects foraging efficiency. The monkeys have found this optimum also.

Page 55: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 56: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 57: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 58: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

The differential model is a better predictor of monkey choice

Page 59: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

• Monkeys match; best LNP model

• Model predicts and generates choices

• Monkeys find optimal 2 and ; 1 not critical

• Unrewarded choices have no effect

• Differential value predicts choices better than fractional value

Page 60: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

?

Page 61: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Best LNP model:

Candidate decision variable, differential value:

g(v1 - v2) = pc

Page 62: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 63: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 64: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Aside: what would Bayes do?1) maintain beliefs over baiting probabilities

2) be greedy or use dynamic programming

Page 65: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Firing rates in LIP are related to target value on a trial-by-trial basis

LIP

http://brainmap.wustl.edu/vanessen.html

gm020b

intoRF

outof RF

Target Value

Page 66: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

The differential model also accounts for more variance in LIP firing rates

Page 67: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

• How we control/measure value• An experimental task based on that principle • A simple model of value based choice• How we validate that model• How we use the model to explore behavior• How we use the model to explore value related signals in the brain

What I’ve told you:

• Our Linear-Nonlinear-Poisson model

• Hybrid models, optimality of reward weights• Neural firing in area LIP correlates with ‘differential value’ on a trial-by-trial basis

• A dynamic foraging task• The matching law

• Predictive and generative validation

Page 68: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 69: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Foraging Efficiency Varies as a Function of 2

Page 70: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Foraging Efficiency Does Not Vary as a Function of 1

Page 71: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

What do animals do?

Matching is a probabilistic policy:

pchoose = f pbait , pbait( )

Matching is almost optimal within the set of probabilistic policies.

Animals match.

Page 72: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 73: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

+ the changeover delay

Page 74: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

Greg Corrado

Page 75: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 76: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 77: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 78: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 79: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 80: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 81: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 82: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 83: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory
Page 84: Institute for Theoretical Physics and Mathematics Tehran January, 2006 Value based decision making: behavior and theory

How do we implement the change over delay?

only one ‘live’ target at a time