Transcript Ch09xxxx

Remember/Know
100
Synonym
Rhyme
90
Lure
80
Percent Correct
70
60
50
40
30
20
10
0
Remember
Know
Not on list
False Memory
• Concepts and Knowledge
Percent of Reports
120
100
80
60
40
20
0
In original list
Normal Distractor
Special Distractor
Knowledge
• Knowledge is - presumably - what
memories create
• It can be gained and lost
• Where is it?
– We don’t know
• How is it organized?
– We have some guesses
Categories and Terms
• Category: Movie Stars
– Exemplar – Individual members of a
category (Harrison Ford, Halle Berry)
– Rule – A precise definition of the criteria
for a category (Must appear in a Movie)
• Prototype – Specifies the properties
that are most likely to be true of a
category
Stereotype or Prototype
• Prototype - A template
against which new
examples are compared
– Flexibility is acceptable
• Stereotype - A fixed set of
traits that a member of a
category is assumed to
possess
– No flexibility
PROTOTYPICALITY:
For each category, list five traits associated with
each category
VEHICLES
CLOTHING
FRUIT
PROTOTYPICALITY:
For each category, assign the number “1” to the
best exemplar, “2” to the second best, “3” to the
third best, “4” to the fourth best, and “5” to the
worst member of the category.
VEHICLES
car
elevator
sled
tractor
train
CLOTHING
jacket
mittens
necklace
pajamas
pants
FRUIT
olive
grapefruit
orange
pear
honeydew
Prototypicality
• Low values are assigned to those items with most
of the features assigned to the prototype
– Greatest family resemblance
– These are also most likely to be listed if asked to list
members of each group
• Consider your results from Part 1
– Items with low scores probably had most of
the traits you listed
• Culture and personal experience play a role (olive)
Reaction Times and Prototypes
Task
1. Primed
“Green”
2. Asked
“Same?”
3. Respond
“Yes/No”
Rosch’s (1975b) Priming Experiment
Reaction Times and Spreading
Activation
• We know that neurons are connected and that
electrical activity passes through the network
of neurons
• Reaction times are assumed to be a reflection
of how information travels from point-to-point
through the network/brain
• Fast reaction times are associated with short
distances between the prime and target
Typicality Gradients
• Sentence verification experiments
Each approach leads to the same results. Do
they all support the existence of cognitive
prototypes?
Prototypes vs. Exemplars
• Prototype theory – We have one “ideal”
member of a category and we make
judgments by comparing a stimulus to the
ideal (which may not be an exemplar!).
• Exemplar theory – We have lots of
exemplars stored and we make judgments
by comparing a stimulus to all exemplars
and adding up the result.
Support for Exemplars
Medin et al.’s (1982) “burlosis” experiment.
Results of the burlosis experiment. Is it
settled?
Exemplars vs. Prototypes
• Classification is probably based on both
strategies
• One approach is probably favored when
there is insufficient data for the other
Semantic Networks
• Might reflect the organization of the
brain, but much can be learned without
the direct neural link
How does Classification Emerge?
Where do you start when…
Teaching a language to children/new learners?
Which terms do you use when discussing basic
elements of stories?
How does Classification Emerge?
Rosch provided evidence for the idea that the basic level is
“psychologically privileged.” We start at the middle.
- Categories at the middle level are most consistent
across cultures, easiest to process, and members are
more clearly grouped (What else do you notice?)
Colins and Quillian (1969) Semantic Networks
The objective was to develop a model of memory/knowledge that
could be tested with computers!
Cognitive Economy
Common properties can be
associated with categories,
rather than individuals.
The time required to retrieve information can be
compared to the time it takes to “travel”
through the semantic network.
RT and Network
Spreading Activation could explain “Priming Effects.”
CogLab: Lexical Decision Task
Meyer and Schvaneveldt (1971)
Problem: Rips et al. (1973)
A pig is a mammal. RT = 1476 ms
A pig is an animal. RT = 1268 ms
Challenges to Spreading Activation
• The trouble with models
– Don’t assume the network is fixed from
moment to moment
– Don’t assume that the network is the
same from person to person
The Solution!
Collins and Loftus (1975).
The hierarchy is
abandoned in
favor of a
network that is
based on
experience!
The Connectionist Approach: Remember the Neuron!
Knowledge doesn’t “live at a node,” but is, instead
represented by a combined firing of neurons.
Weights
Weights
A parallel-distributed-processing (PDP) network.
All seven units carry the message for both
animals. The pattern of activity determines
what you are thinking about.
Learning in a PDP network.
(a) Initially presenting canary
causes a pattern of activation
in the output units that is
different than the pattern that
stands for canary. (b) An error
signal is transmitted back
through the network to
indicate how weights need to
be changed to achieve the
correct output response.
continued on next slide
Figure 8.30c (p. 300)
(c) After the processes in (a) and (b) are repeated many times, the
network has learned to respond correctly to canary.
Pros and Cons of PDP
• Uses the rules of the nervous system
• Uses rules of learning
– Try, get feedback, adjust, try again
• Exhibits graceful degradation
• Can’t explain fast learning
• New learning will compromise old knowledge
• May explain some kind of learning, but not all.
Neuro-physiological Evidence
• We have “category” neurons (i.e. neurons that
respond to the image of any dog)
• Specific forms of agnosia
CogLab
• Lexical Decision (done)
• Prototypes
Parallel Distributed Processing
(PDP) as a Model of Cognition
•
•
•
•
Immune to minor damage
Work when input is noisy or incomplete
Allow retrieval by content
Retrieve typical instances of categories
(prototypes)
PDP Models
• McClelland and Rummelhart(1986)
• Knowledge is distributed
• Computations take place in parallel
Conclusions are based on consensus!
PDP vs. Traditional Computer
• Processing is parallel, not serial
– In a serial system, one broken link stops
everything
• Information is distributed, not local
Distributed Information Storage
• The “benign” qualities of a tumor live in the same set of
weights and connections as the malignant qualities
Fuzzy Nature of Neurons
• The response of neuron to any particular
input is probabalistic, not fixed
Light ON
Fuzzy Nature of Neurons
• Q: How can perceptions be consistent if neurons
aren’t?
• A: Perceptions arise from many active neurons
and the responses (opinions) are averaged.
– No neuron has the final say
– No neuron is indespensable
Access by Content
• Connectionist networks allow access by
content, rather than address.
– Phone book: Access by address
• You can find a number if you know name
• You can’t find a name if you know a number
• You cant find a number if you know an address
– Access by content allows you to work either
way. Any bit of info can lead to all other bits
Access by Content
• Each piece of information is linked to the other
pieces and, therefore, can activate them (bring
them into awareness).
• Access is fault-tolerant, so small errors in input
can still lead to a correct solution. A
distributed system will look for a “best-fit.”
– An erroneous spelling will not lead to a correct
phone number
Jets and Sharks
Access by Address
Conventional Database
Works well if you start with
a name.
• Is Fred a pusher?
• Do you know a pusher?
• Are the Sharks educated?
• Are Jets likely to be
divorced?
• What are Jets like?
Access by Content
Content Addressable Database
You can start from anywhere!
• Is Fred a pusher?
• Do you know a pusher?
• Are the Sharks educated?
• Are Jets likely to be
divorced?
• What are Jets like?
A Closer Look
Excitatory
Connections
Nodes
Knowledge
Areas
Inhibitory
Connections
Instance
Units
A Closer Look
• All nodes start at the same level
of activity.
• Input (question) activates a
node and “information” starts
flowing through the network.
• Eventually, the network
stabilizes, and the solution is
represented by the most active
nodes