Transcript abstract constructs

Turn to Cognitive Perspective
Brief history
Inspiration from other fields
Inspiration from within psychology
A framework: we’re talking about
scientific method
Observe
Test
Theorize
Example: what’s
this?
1. Fool with it.
2. Develop a hypothesis
3. Try it to see if you’re right
Pot trammel, used to position cooking
pots higher or lower over a fire.
1. Fool with it.
Observe
3. Try it to see if
you’re right
Test
Theorize
2. Develop a
hypothesis
Philosophy (prior to mid
Observe
Test
Theorize
th
18 )
Introspectionism
• First attempt to apply scientific method to
thought (1880s).
• Wilhelm Wundt. Edward Titchener
• Goal: description of the contents of
consciousness; find irreducible “elements of
consciousness.
• Method: introspection.
Introspectionism
Observe
Only use
introspection to
test
Test
Theorize
Introspection
•
•
•
•
Only covers conscious processing
Poor reliability between subjects
Watching a mental process changes it
Not making much progress
Turn to behaviorism
• Focus on observables only
• Theory must be parsimonious
• Break behavior down into irreducible
concepts
Focus on observables
Note that psychology became a
science not of the mind, not of
thought, but of behavior.
Parsimony
Shared with all sciences; this was
lacking in introspectionism, and was
really emphasized in behaviorism.
E.g., all of behavior boiled down to a
few simple laws of learning: operant,
and classically conditioned response
Irreducible concepts
Again, all sciences have this as their
goal; introspectionism had it as a goal
as well, but couldn’t come up with good
irreducible concepts. Behaviorism had
good candidates: free operant and the
conditioned response.
What was lacking in this perspective,
and where are we now?
“Observables
only” in theory
development
was too
restrictive
Observe
Test
Theorize
Overview of changes
• Inability to account for all animal behavior
indicated something might be wrong.
• Inability to go from animal models to
human behavior indicated it was incomplete
• Posing abstract constructs suggested as
what was needed to solve these problems.
• Inspiration from other fields to use abstract
constructs.
Inability to account for all animal
behavior
Work from ethology--1
Critical period: A time when the animal
is able to learn particular information
rapidly and with little exposure; if the
time window is missed, the animal learns
with greater effort or not at all.
Critical period--example
Some birds (e.g., ducks, geese) follow
the first large thing that they see when
they are hatched--usually first large
thing is mom.
The tendency to follow first large thing
has a critical period.
What happens if first large thing is not
mom?
Lorenz as mom
Lorenz as mom
Critical period in humans?
Johnson and Newport (1989)
A case for the critical period
Subjects
 46 native Chinese or Korean learners of English
 In the US for at least 5 years
 Age of arrival: 3-39 years old
Method
 Grammaticality Judgment Task
Results
 Accuracy on grammaticality test correlated
with age of arrival for subjects who arrived in
US before puberty
 Accuracy on grammaticality NOT correlated
with age of arrival for subjects who arrived in
US after puberty
Other work from ethology
Fixed action pattern--this is a complex
behavior that emerges, full-blown,
with little opportunity for practice or
reward.
Fixed action pattern--stickleback
Fixed action pattern in humans?
Eyebrow flick
Fixed action pattern--coy
• Catch eye
• Look down, smiling
• Catch eye, smile,
cover mouth
Fixed action pattern
Overview of changes
• Inability to account for all animal behavior
• Inability to go from animal models to
human behavior
• Apparent need to pose abstract constructs.
• Inspiration from other fields to use abstract
constructs.
Going from animals to humans
Language
Skinner, 1957
Chomsky, 1959
Language is generative
Skinner, 1957
The behaviorist account of language:
Child utters sounds at random; is
reinforced for utterances that are
close to appropriate. Language is
shaped.
Chomsky, 1959
Behaviorist account is wrong
Behaviorist accounts ignore that
language is generative. This means
that virtually everything you say and
hear is novel. It can’t be the case that
you understand it because of
reinforcement in the past, because
you’ve never heard it before.
Need for abstract constructs
Generativity seems to call for rules,
similar to y = mx +b
Thoughts
about
baby
trying to
hug self
Translation
rules
“That is so
cute!”
Thoughts
about
baby
Translation
rules
“How many hands
did Aristotle have?”
“That is so
cute!”
Speech
Interpretation
Find answer
in memory
Phrase the
answer
Interpretation of
question
Make decision:
answer or not?
Create motor
commands to
lips, tongue, etc
“Two”
Need for abstract constructs
In memory research, data seemed
to call for some discussion of
subjects’ strategies.
E.g., subjects hear a mixed list, but report it
back in categories. When asked, they say that
they are doing so because it helps them to
remember.
Abstract constructs
A theoretical set of processes and
representations (e.g., a rule for
language, or a strategy in memory)
E.g., why do people forget phone numbers after 30 seconds?
You could propose that there is a short term memory, which
holds information for 30 seconds or so.
Short term memory has representations (e.g. of the number “8”
and processes (e.g., refresh). It’s like a mini-theory.
Abstract constructs
This drove behaviorist nuts, because
it violated one of their tenets.
Which one?
Abstract constructs
This drove behaviorist nuts, because it
violated one of their tenets.
Which one?
Deal only with observables.
“Where is this short term memory?
I can’t see it.” (uttered in snide voice)
Abstract constructs--other fields
Computer science (artificial
intelligence research) was crucial.
Their whole field was based on
abstract constructs, with no loss of
rigor.
In the 1950’s, AI had progressed beyond “glorified
adding machines.”
Newell & Simon’s Logic
Theorist (1956)
• Discovered proofs in symbolic logic, of the kind originally
presented in Whitehead & Russell’s Principia Mathematica
• Program contains a list of axioms, a list of previously proven
theorems, and a manner in which axioms may be combined
• Program given new statement which it must prove.
• Successfully proved 38 of the first 52 theorems in Principia
Mathematica (usually in less than 5 minutes each).
• Not just a cool program--they claimed that people really solve
problems this way.
• Other programs were simply cool.
Greater usefulness of brain data
“What happened to you?”
Speech
Interpretation
Find answer
in memory
Phrase the
answer
What’s wrong with this patient?
Interpretation of
question
Make decision:
answer or not?
Create motor
commands to
lips, tongue, etc
“bus”
Abstract constructs--recap
1. It appeared that certain phenomena could
only be accounted for by using abstract
constructs, not behaviorist principles.
2. Abstract constructs were forbidden in
behaviorist thinking--they were rejected as
“mystical,” “unscientific”
3. Other fields (artificial intelligence,
neurology) used abstract constructs with no
apparent loss of rigor.
What replaced behaviorism?
We’ve just been listing problems with
behaviorism--question:
What could replace behaviorism?
Observable
stimuli
Observable
stimuli
Observable
responses
Abstract
constructs
(mind)
Observable
responses
Huge question
How can you discover abstract
constructs?
Example
Sternberg task: given a “set,” e.g., A, K, Y, W:
size of the set varies.
Stimulus appears, K,and you must
say “yes” or “no” if it is in the set.
Encode
K
Search
AKYW
Decide
AKYW
K
Respond
Hit
“yes”
button
How does the search work?
• Serial, exhaustive
• Serial, self-terminating
• Parallel
Serial, exhaustive
Serial, self-terminating
Parallel
Sternberg’s data
Yes=filled
No=open
Key features of cognitive strategy
• Develop alternate models of processing
• Derive signature predictions of each model
• Obtain data that allow comparison