markov chain monte carlo with people tom griffiths department of psychology cognitive science...
Post on 21-Dec-2015
223 views
TRANSCRIPT
![Page 1: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/1.jpg)
Markov chain Monte Carlowith people
Tom GriffithsDepartment of Psychology
Cognitive Science Program
UC Berkeley
with Mike Kalish, Stephan Lewandowsky, and Adam Sanborn
![Page 2: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/2.jpg)
Inductive problems
blicket toma
dax wug
blicket wug
S X Y
X {blicket,dax}
Y {toma, wug}
Learning languages from utterances
Learning functions from (x,y) pairs
Learning categories from instances of their members
![Page 3: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/3.jpg)
Computational cognitive science
Identify the underlying computational problem
Find the optimal solution to that problem
Compare human cognition to that solution
For inductive problems, solutions come from statistics
![Page 4: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/4.jpg)
Statistics and inductive problems
Cognitive science
Categorization
Causal learning
Function learning
Language
…
Statistics
Density estimation
Graphical models
Regression
Probabilistic grammars
…
![Page 5: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/5.jpg)
Statistics and human cognition
• How can we use statistics to understand cognition?
• How can cognition inspire new statistical models?– applications of Dirichlet process and Pitman-Yor process
models to natural language– exchangeable distributions on infinite binary matrices via
the Indian buffet process (priors on causal structure)– nonparametric Bayesian models for relational data
![Page 6: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/6.jpg)
Statistics and human cognition
• How can we use statistics to understand cognition?
• How can cognition inspire new statistical models?– applications of Dirichlet process and Pitman-Yor process
models to natural language– exchangeable distributions on infinite binary matrices via
the Indian buffet process (priors on causal structure)– nonparametric Bayesian models for relational data
![Page 7: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/7.jpg)
Statistics and human cognition
• How can we use statistics to understand cognition?
• How can cognition inspire new statistical models?– applications of Dirichlet process and Pitman-Yor process
models to natural language– exchangeable distributions on infinite binary matrices via
the Indian buffet process– nonparametric Bayesian models for relational data
![Page 8: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/8.jpg)
Are people Bayesian?
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
Reverend Thomas Bayes
![Page 9: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/9.jpg)
Bayes’ theorem
€
P(h | d) =P(d | h)P(h)
P(d | ′ h )P( ′ h )′ h ∈H
∑
Posteriorprobability
Likelihood Priorprobability
Sum over space of hypothesesh: hypothesis
d: data
![Page 10: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/10.jpg)
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
People are stupid
![Page 11: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/11.jpg)
Predicting the future
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
How often is Google News updated?
t = time since last update
ttotal = time between updates
What should we guess for ttotal given t?
![Page 12: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/12.jpg)
The effects of priors
![Page 13: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/13.jpg)
Evaluating human predictions
• Different domains with different priors:– a movie has made $60 million [power-law]
– your friend quotes from line 17 of a poem [power-law]
– you meet a 78 year old man [Gaussian]
– a movie has been running for 55 minutes [Gaussian]
– a U.S. congressman has served for 11 years [Erlang]
• Prior distributions derived from actual data
• Use 5 values of t for each
• People predict ttotal
![Page 14: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/14.jpg)
peopleparametric priorempirical prior
Gott’s rule
![Page 15: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/15.jpg)
A different approach…
Instead of asking whether people are rational, use assumption of rationality to investigate cognition
If we can predict people’s responses, we can design experiments that measure psychological variables
![Page 16: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/16.jpg)
Two deep questions
• What are the biases that guide human learning?– prior probability distribution P(h)
• What do mental representations look like?– category distribution P(x|c)
€
limt →∞
P(x(t ) = i | x(0)) = π i
![Page 17: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/17.jpg)
Two deep questions
• What are the biases that guide human learning?– prior probability distribution on hypotheses, P(h)
• What do mental representations look like?– distribution over objects x in category c, P(x|c)
Develop ways to sample from these distributions
![Page 18: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/18.jpg)
Outline
Markov chain Monte Carlo
Sampling from the prior
Sampling from category distributions
![Page 19: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/19.jpg)
Outline
Markov chain Monte Carlo
Sampling from the prior
Sampling from category distributions
![Page 20: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/20.jpg)
• Variables x(t+1) independent of history given x(t)
• Converges to a stationary distribution under easily checked conditions (i.e., if it is ergodic)
x x x x x x x x
Transition matrixT = P(x(t+1)|x(t))
Markov chains
![Page 21: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/21.jpg)
Markov chain Monte Carlo
• Sample from a target distribution P(x) by constructing Markov chain for which P(x) is the stationary distribution
• Two main schemes:– Gibbs sampling– Metropolis-Hastings algorithm
![Page 22: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/22.jpg)
Gibbs sampling
For variables x = x1, x2, …, xn and target P(x)
Draw xi(t+1) from P(xi|x-i)
x-i = x1(t+1), x2
(t+1),…, xi-1(t+1)
, xi+1(t)
, …, xn(t)
![Page 23: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/23.jpg)
Gibbs sampling
(MacKay, 2002)
![Page 24: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/24.jpg)
Metropolis-Hastings algorithm(Metropolis et al., 1953; Hastings, 1970)
Step 1: propose a state (we assume symmetrically)
Q(x(t+1)|x(t)) = Q(x(t))|x(t+1))
Step 2: decide whether to accept, with probability
Metropolis acceptance function
Barker acceptance function
![Page 25: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/25.jpg)
Metropolis-Hastings algorithm
p(x)
![Page 26: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/26.jpg)
Metropolis-Hastings algorithm
p(x)
![Page 27: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/27.jpg)
Metropolis-Hastings algorithm
p(x)
![Page 28: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/28.jpg)
Metropolis-Hastings algorithm
A(x(t), x(t+1)) = 0.5
p(x)
![Page 29: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/29.jpg)
Metropolis-Hastings algorithm
p(x)
![Page 30: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/30.jpg)
Metropolis-Hastings algorithm
A(x(t), x(t+1)) = 1
p(x)
![Page 31: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/31.jpg)
Outline
Markov chain Monte Carlo
Sampling from the prior
Sampling from category distributions
![Page 32: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/32.jpg)
Iterated learning(Kirby, 2001)
What are the consequences of learners learning from other learners?
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
![Page 33: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/33.jpg)
Analyzing iterated learning
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
PL(h|d): probability of inferring hypothesis h from data d
PP(d|h): probability of generating data d from hypothesis h
PL(h|d)
PP(d|h)
PL(h|d)
PP(d|h)
![Page 34: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/34.jpg)
Iterated Bayesian learning
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
PL(h|d)
PP(d|h)
PL(h|d)
PP(d|h)
€
PL (h | d) =PP (d | h)P(h)
PP (d | ′ h )P( ′ h )′ h ∈H
∑
Assume learners sample from their posterior distribution:
![Page 35: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/35.jpg)
Analyzing iterated learning
d0 h1 d1 h2PL(h|d) PP(d|h) PL(h|d)
d2 h3PP(d|h) PL(h|d)
d PP(d|h)PL(h|d)h1 h2d PP(d|h)PL(h|d)
h3
A Markov chain on hypotheses
d0 d1h PL(h|d) PP(d|h)d2h PL(h|d) PP(d|h) h PL(h|d) PP(d|h)
A Markov chain on data
![Page 36: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/36.jpg)
Stationary distributions
• Markov chain on h converges to the prior, P(h)
• Markov chain on d converges to the “prior predictive distribution”
€
P(d) = P(d | h)h
∑ P(h)
(Griffiths & Kalish, 2005)
![Page 37: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/37.jpg)
Explaining convergence to the prior
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (Uncompressed) decompressor
are needed to see this picture.
PL(h|d)
PP(d|h)
PL(h|d)
PP(d|h)
• Intuitively: data acts once, prior many times
• Formally: iterated learning with Bayesian agents is a Gibbs sampler on P(d,h)
(Griffiths & Kalish, in press)
![Page 38: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/38.jpg)
Revealing inductive biases
• Many problems in cognitive science can be formulated as problems of induction– learning languages, concepts, and causal relations
• Such problems are not solvable without bias(e.g., Goodman, 1955; Kearns & Vazirani, 1994; Vapnik, 1995)
• What biases guide human inductive inferences?
If iterated learning converges to the prior, then it may provide a method for investigating biases
![Page 39: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/39.jpg)
Serial reproduction(Bartlett, 1932)
• Participants see stimuli, then reproduce them from memory
• Reproductions of one participant are stimuli for the next
• Stimuli were interesting, rather than controlled– e.g., “War of the Ghosts”
![Page 40: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/40.jpg)
General strategy
• Use well-studied and simple stimuli for which people’s inductive biases are known– function learning– concept learning– color words
• Examine dynamics of iterated learning– convergence to state reflecting biases– predictable path to convergence
![Page 41: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/41.jpg)
Iterated function learning
• Each learner sees a set of (x,y) pairs
• Makes predictions of y for new x values
• Predictions are data for the next learner
data hypotheses
(Kalish, Griffiths, & Lewandowsky, in press)
![Page 42: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/42.jpg)
Function learning experiments
Stimulus
Response
Slider
Feedback
Examine iterated learning with different initial data
![Page 43: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/43.jpg)
1 2 3 4 5 6 7 8 9
IterationInitialdata
![Page 44: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/44.jpg)
Identifying inductive biases
• Formal analysis suggests that iterated learning provides a way to determine inductive biases
• Experiments with human learners support this idea– when stimuli for which biases are well understood are used,
those biases are revealed by iterated learning
• What do inductive biases look like in other cases?– continuous categories– causal structure– word learning– language learning
![Page 45: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/45.jpg)
• Iterated learning for MAP learners reduces to a form of the stochastic EM algorithm– Monte Carlo EM with a single sample
• Provides connections between cultural evolution and classic models used in population genetics– MAP learning of multinomials = Wright-Fisher
• More generally, an account of how products of cultural evolution relate to the biases of learners
Statistics and cultural evolution
![Page 46: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/46.jpg)
Outline
Markov chain Monte Carlo
Sampling from the prior
Sampling from category distributions
![Page 47: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/47.jpg)
Categories are central to cognition
![Page 48: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/48.jpg)
Sampling from categories
Frog distribution
P(x|c)
![Page 49: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/49.jpg)
A task
Ask subjects which of two alternatives comes from a target category
Which animal is a frog?
![Page 50: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/50.jpg)
A Bayesian analysis of the task
Assume:
![Page 51: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/51.jpg)
Response probabilities
If people probability match to the posterior, response probability is equivalent to the Barker acceptance function for target distribution p(x|c)
![Page 52: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/52.jpg)
Collecting the samplesWhich is the frog? Which is the frog? Which is the frog?
Trial 1 Trial 2 Trial 3
![Page 53: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/53.jpg)
Verifying the method
![Page 54: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/54.jpg)
Training
Subjects were shown schematic fish of different sizes and trained on whether they came from the ocean
(uniform) or a fish farm (Gaussian)
![Page 55: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/55.jpg)
Between-subject conditions
![Page 56: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/56.jpg)
Choice task
Subjects judged which of the two fish came from the fish farm (Gaussian) distribution
![Page 57: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/57.jpg)
Examples of subject MCMC chains
![Page 58: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/58.jpg)
Estimates from all subjects
• Estimated means and standard deviations are significantly different across groups
• Estimated means are accurate, but standard deviation estimates are high– result could be due to perceptual noise or response gain
![Page 59: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/59.jpg)
Sampling from natural categories
Examined distributions for four natural categories: giraffes, horses, cats, and dogs
Presented stimuli with nine-parameter stick figures (Olman & Kersten, 2004)
![Page 60: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/60.jpg)
Choice task
![Page 61: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/61.jpg)
Samples from Subject 3(projected onto plane from LDA)
![Page 62: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/62.jpg)
Mean animals by subject
giraffe
horse
cat
dog
S1 S2 S3 S4 S5 S6 S7 S8
![Page 63: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/63.jpg)
Marginal densities (aggregated across subjects)
Giraffes are distinguished by neck length, body height and body tilt
Horses are like giraffes, but with shorter bodies and nearly uniform necks
Cats have longer tails than dogs
![Page 64: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/64.jpg)
Relative volume of categoriesMinimum Enclosing Hypercube
Giraffe Horse Cat Dog
0.00004 0.00006 0.00003 0.00002
Convex hull content divided by enclosing hypercube content
Convex Hull
![Page 65: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/65.jpg)
Discrimination method(Olman & Kersten, 2004)
![Page 66: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/66.jpg)
Parameter space for discrimination
Restricted so that most random draws were animal-like
![Page 67: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/67.jpg)
MCMC and discrimination means
![Page 68: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/68.jpg)
Conclusion
• Markov chain Monte Carlo provides a way to sample from subjective probability distributions
• Many interesting questions can be framed in terms of subjective probability distributions– inductive biases (priors)
– mental representations (category distributions)
• Other MCMC methods may provide further empirical methods…– Gibbs for categories, adaptive MCMC, …
![Page 69: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/69.jpg)
A different approach…
Instead of asking whether people are rational, use assumption of rationality to investigate cognition
If we can predict people’s responses, we can design experiments that measure psychological variables
Randomized algorithms Psychological experiments
![Page 70: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/70.jpg)
![Page 71: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/71.jpg)
€
PL (h | d)∝PP (d | h)P(h)
PP (d | ′ h )P( ′ h )′ h ∈H
∑
⎡
⎣
⎢ ⎢ ⎢
⎤
⎦
⎥ ⎥ ⎥
r
r = 1 r = 2 r =
From sampling to maximizing
![Page 72: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/72.jpg)
• General analytic results are hard to obtain– (r = is Monte Carlo EM with a single sample)
• For certain classes of languages, it is possible to show that the stationary distribution gives each hypothesis h probability proportional to P(h)r
– the ordering identified by the prior is preserved, but not the corresponding probabilities
(Kirby, Dowman, & Griffiths, in press)
From sampling to maximizing
![Page 73: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/73.jpg)
Implications for linguistic universals
• When learners sample from P(h|d), the distribution over languages converges to the prior– identifies a one-to-one correspondence between inductive
biases and linguistic universals
• As learners move towards maximizing, the influence of the prior is exaggerated– weak biases can produce strong universals– cultural evolution is a viable alternative to traditional
explanations for linguistic universals
![Page 74: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/74.jpg)
![Page 75: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/75.jpg)
Iterated concept learning
• Each learner sees examples from a species
• Identifies species of four amoebae
• Iterated learning is run within-subjects
data hypotheses
(Griffiths, Christian, & Kalish, in press)
![Page 76: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/76.jpg)
Two positive examples
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
data (d)
hypotheses (h)
![Page 77: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/77.jpg)
Bayesian model(Tenenbaum, 1999; Tenenbaum & Griffiths, 2001)
€
P(h | d) =P(d | h)P(h)
P(d | ′ h )P( ′ h )′ h ∈H
∑d: 2 amoebaeh: set of 4 amoebae
€
P(d | h) =1/ h
m
0
⎧ ⎨ ⎩
d ∈ h
otherwise
m: # of amoebae in the set d (= 2)|h|: # of amoebae in the set h (= 4)
€
P(h | d) =P(h)
P( ′ h )h '|d ∈h'
∑Posterior is renormalized prior
What is the prior?
![Page 78: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/78.jpg)
Classes of concepts(Shepard, Hovland, & Jenkins, 1958)
Class 1
Class 2
Class 3
Class 4
Class 5
Class 6
shape
size
color
![Page 79: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/79.jpg)
Experiment design (for each subject)Class 1Class 2Class 3Class 4Class 5Class 6Class 1Class 2Class 3Class 4Class 5Class 6
6 iterated learning chains
6 independent
learning “chains”
![Page 80: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/80.jpg)
Estimating the prior
data (d)hy
poth
eses
(h)
![Page 81: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/81.jpg)
Estimating the prior
Class 1Class 2
Class 3
Class 4
Class 5
Class 6
0.8610.087
0.009
0.002
0.013
0.028
Prior
r = 0.952
Bayesian modelHuman subjects
![Page 82: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/82.jpg)
Two positive examples(n = 20)
Prob
abil
ity
Iteration
Prob
abil
ity
Iteration
Human learners Bayesian model
![Page 83: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/83.jpg)
Two positive examples(n = 20)
Prob
abil
ity
Bayesian model
Human learners
![Page 84: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/84.jpg)
Three positive examples
data (d)
hypotheses (h)
![Page 85: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/85.jpg)
Three positive examples(n = 20)
Prob
abil
ity
Iteration
Prob
abil
ity
Iteration
Human learners Bayesian model
![Page 86: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/86.jpg)
Three positive examples(n = 20)
Bayesian model
Human learners
![Page 87: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/87.jpg)
![Page 88: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/88.jpg)
Classification objects
![Page 89: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/89.jpg)
Parameter space for discrimination
Restricted so that most random draws were animal-like
![Page 90: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/90.jpg)
MCMC and discrimination means
![Page 91: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/91.jpg)
Problems with classification objects
Category 1
Category 2
Category 1
Category 2
![Page 92: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/92.jpg)
Problems with classification objectsMinimum Enclosing Hypercube
Giraffe Horse Cat Dog
0.00004 0.00006 0.00003 0.00002
Convex hull content divided by enclosing hypercube content
Convex Hull
![Page 93: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/93.jpg)
![Page 94: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/94.jpg)
Allowing a Wider Range of Behavior
An exponentiated choice rule results in a Markov chain with stationary distribution corresponding to an exponentiated version of the category distribution, proportional to p(x|c)
![Page 95: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/95.jpg)
Category drift
• For fragile categories, the MCMC procedure could influence the category representation
• Interleaved training and test blocks in the training experiments
![Page 96: Markov chain Monte Carlo with people Tom Griffiths Department of Psychology Cognitive Science Program UC Berkeley with Mike Kalish, Stephan Lewandowsky,](https://reader036.vdocuments.us/reader036/viewer/2022062516/56649d6c5503460f94a4c846/html5/thumbnails/96.jpg)