Lecture 3 - Modern Approaches in Cognitive Science Flashcards
What was a problem with analytic introspection according
to behaviorists?
The results were highly variable between individuals.
Watson
focus on behavior
Behaviors no longer allowed inference to an internal state.
Behaviors were, themselves, the thing of interest.
‘Mental’ entities
something you only know from the first person perspective: feelings, thoughts, beliefs, etc…
something only the individual has direct access too
Watson and Rayner (1920) used
classical conditioning in the “Little
Albert” experiment.
trying to show a complex mental behavior (phobia of animals) could be explained purely in terms of experience
learning process: conjunction of sound and animal can cause phobia and fear (behavior)
• A nine-month-old became frightened of a rat (and rabbit) after a loud noise was made every time they were presented (and then anything furry caused him to cry).
− The goal was to show that simple learned responses could be trained without appeal to ‘mental’ processes.
− Fear of animal [crying] = presence of animal + bad
experience
Watson and nurture over nature
if you could control everyone’s experiences, you can create any kind of person you want
nurture is the deciding factor - almost nothing is innate
B. F. Skinner took behaviorism even further
- Skinner believed that free will was a myth. All
behavior was the result of rewards and reinforcements of behavior.
− Psychology should be just the experimental analysis of behavior.
− His work mapped different schedules of reinforcement to observable behavioral outcomes.
− operant conditioning.
− All behavior can be explained in terms of experience (not innate mental processes).
- Under Skinner (and others), behaviorism become the
dominant movement in Psychology for 50 years.
operant conditioning
more radical version of behaviorism
Behaviors are shaped (trained) by rewards or punishments.
start off with some sort of stimulus in the environment which leads to some sort of behavioral response in an individual and that response will have some sort of outcome: that outcome is either positive or negative reinforcement
focuses more on the behavioral response and outcome: how likely are you to perform this response again based on the outcome
classical conditioning
focuses more on the stimulus and the response
Skinner was particularly famous because
he published novels that helped popularize behaviorist ideas (e.g. Walden Two, 1948).
book about a Utopia that using operant principles you could make a perfect society
Many of its tools (e.g. classical conditioning and operant
conditioning) are still used
and have found practical application in modern settings.
However, behaviorism was most useful in very simple
(impoverished) stimulus conditions. It was less successful in
dealing with
complex learning.
− That is, it failed to overcome problems of computational
complexity…
very few stimuli that are acting in these lab settings but in the real world people exist in complex environments
behaviorism can’t explain how to cross the street
No cognitive system can deal with
the computational
complexity of all events happening simultaneously in the
world.
We use ____ , innate and learned, to narrow our focus
on what is relevant.
biases
[Cognitive Psychology
The study of these biases and tendencies can help us build
models of the mind.
mental chronometry
Early experiments used measurements of behavior to make
inferences about mental events
computational complexity
the idea of trying to deal with a large number of stimuli in real time
too many things to see, memorize, problems to solve
you simplify the situation to make it manageable
Behaviorists rejected the study of
mental events as too subjective and untestable.
Behaviorists focused on how to
learn how behaviors could reliably be produced over time.
Behaviorism, despite its (generally) broad
acceptance for a time, had several
problems.
innateness
It had a difficult time explaining differences in
learning by species (recall that rats learn taste
associations and pigeons learn visual
associations).
− If learning is determined by
reinforcement, what accounts for innate
differences?
Behaviorism also couldn’t explain delayed
responses (when the stimulus was no
longer present).
Otto Tinklepaugh (1928, 1932)
had monkeys and chimpanzees observe as
food (e.g. banana) was placed under one of two cups.
− They chimp was then taken out of the room. Meanwhile, the banana was replaced by lettuce or a carrot.
− The chimp was returned several minutes later and went to the correct cup location to retrieve the banana… and seemed to confused (and upset) to find the lettuce. Won’t eat the lettuce.
− It seemed very difficult to explain the confused behavior without the existence of memory (expectation and preference).
Edward Tolman (1938)
performed an experiment with rats that was a direct challenge to behaviorism: exhibited a different behavior!
Memory - something being processes or operated on
- Rats explored a maze (initial period)
- They were then trained over time to go to a specific location for food (several repetitions).
- Later, they were placed in a new initial position…
* They performed different behaviors (left turns) to get the reward!
* Tolman suggested the rat had a cognitive map that was stored in its memory from the initial stage. Food location had been remembered in reference to an internal spatial map.
A pivotal challenge came when Skinner tried to
explain the complex learned behavior of
language in his book Verbal Behavior (1957).
• He argued that children learn language through reinforced learning (operant conditioning).
− Children imitate what they hear.
− Correct speech is rewarded.
− Children then generalize to new situations and new reinforcement begins once again.
Linguist Noam Chomsky (1959) published a
scathing review of Skinner’s work, pointing out
several flaws in the account:
- How do we talk about past experiences? What present stimulus triggers that response? You need memory.
- How do children learn sentences that aren’t reinforced? (e.g. “I hate you!”)
- How do children talk about imaginary things if there
isn’t a stimulus there to trigger it and they haven’t heard it before?
Chomsky argued that there must be some
innate biological factors (the syntax - structure for the language was already in the brain - universal) that determine how language is learned, not merely external reinforcement.
wasn’t merely conditioning
Behaviorism’s increasing shortcomings coincided with the
additional attention given to computation and human factors
research during World War II.
• Mathematicians worked with cryptographers in England to
break German codes.
• Psychologists tried to help the Air Ministry improve human
performance in radar detection and aircraft control.
• In applied human factors research, issues were framed as
problems of information processing.
What do you consider to be the strongest empirical
evidence against behaviorism (and why)?
Tinklepaugh’s finding that chimpanzees seemed to
remember where food had been placed after a delay.
Tolman’s finding that rats used cognitive maps to
execute different behaviors to receive a reward.
- a little stronger: simple behavior, the idea that an animal under simple stimulus condition was using a cognitive map was a big deal
Alan Turing’s work on code-breaking led
him to reconceive and generalize
intelligent behavior.
• It was demonstrated mathematically that a general problem solving computer, called a universal Turing machine (1936), could carry out any mathematical
computation.
• ‘Knowledge’ could be conceptualized as symbolic information processed in discrete steps.
• This suggested that intelligent capabilities (e.g. logic, learning, and memory) are not inherently biological.
They can be implemented in a machine.
The Turing Test (1950)
Is a universal Turing machine really intelligent?
was devised as a way to determine if the computer was truly
performing in the same manner as a human (and maybe ‘thinking’ like a human).
- had an individual talking to a computer or to a human, if you couldn’t distinguish between the two than the machine is “intelligent.”
- If a computer can produce communication that is indistinguishable from that of a human operator, then the computer can be said to be intelligent.
- This idea proved to be essential in the formulation of Artificial Intelligence.
- If a Turing machine can produce ‘intelligent’ behavior, then maybe the human mind can be represented as a computer!
