Lecture 4 - Biological Backdrop of Learning Flashcards
What theoretical framework emerged around World War II
for conceptualizing human problem solving?
Information processing!!!!
any kind of signal (process) could be abstractly represented and then manipulated (whether that’s something that a computer or human is doing
Example of this: Universal Turing machines (hypothetical computer: if you can represent symbols and do basic manipulations in discrete steps then with a enough time you can do any mathematical computation and if you can do this you can solve any logic thing)
mounting evidence against behaviorism (something going on inside the head during learning -animals)
Behaviorism’s increasing shortcomings coincided with
the additional attention given to computation and human factors research during World War II.
Digital computers soon became a dominant model for how the mind may operate
British psychologist Donald Broadbent (1958) developed the first flow diagram of a cognitive process based on the design of a digital computer.
some kind of input (keystroke) and within the computer it has to interpret that info then it gives output
supposedly you could take that output and create a different kind of input
Broadbent’s ‘filter model’ of attention
signal and noise coming in, filter in the head gets rid of just enough of the noise and takes what’s left to the detector (yes that was a plane)
a very serial process
Cognitive Psychology (trying to account for internal proceses) became prominent in the late 1960s.
- Digital computers were the basic model for how information might be processed in the brain.
- Reaction time studies (mental chronometry) were a natural fit –determining how long a mental process might take (serial process: if you could time how long it took the input to get to diff stages you can break down those stages/steps in the process).
- Studying the biases and errors of the cognitive system provides clues as to how the internal ‘pieces’ fit together (give the system “bad” data, how long does it take for that to get through, where does it break? intuitive way of seeing how the system operates).
• The experimental methods of the behaviorists were still
important, but were more powerful when you were allowed to infer internal mental processes once again.
The multiplicity of approaches used for
understanding the mind eventually led to……
…..the interdisciplinary field of Cognitive Science.
The goal of Cogsci is still to…
….understand the nature of the human mind, drawing on any methods and theoretical foundations available from lots of different disciplines (philosophy, psychology, AI, neuroscience, anthropology, linguistics)
In recent years, the advent of methodologies in
cognitive neuroscience
(e.g. fMRI, EEG/ERP, TMS, etc.) have led to more detailed models.
note that we are still using
behavioral measures to infer what is happening in the mind.
INFERENCES
we do not have direct access to mental events!!
And we still argue over how much is innate and
how much is learned…
Behaviorism rejected
any appeal to innate or ‘mental’ processes. It
thought that psychology should focus only on the study of
behaviors and outcomes.
Classical and operant conditioning
were the main methodological tools in the study of behaviorist learning.
decline of behaviorism
Failures to explain delayed responses (Tinklepaugh), spatial
learning in rats (Tolman’s cognitive maps), and unreinforced
language learning in children (Chomsky) contributed to its
decline.
Information processing and modeling
based on digital computers provided a new theoretical framework for understanding internal mental processes.
Cognitive Psychology and Cognitive Science re-remerged
to study mental events with the aid of tools from many disciplines.
Early Cognitive Psychology was born out of
a response to Behaviorism and still used many of its paradigms and methods.
• The focus was on explaining observable behaviors in terms of processes that (now) involved cognitive operations.
• However, not all behaviors seem to need cognitive
involvement (e.g. reflex actions and instinctual behaviors - innate).
- Also, some behaviors seem to emerge at specific periods in development (start at some time after birth -learned? or something else?).
- How can we explain these behaviors?
Fixed-action patterns
Some behavioral sequences seem to be innate (instinctive) and don’t need to be learned.
• Species specific - adaptive behavior: These are usually adaptive behaviors that help the organism survive in its
environment.
• They are initiated by releasers – specific stimuli that trigger the behavior.
• Once the behavior pattern is started, it may run its course automatically until completed.
− EX: Greylag goose and egg rolling (still does the movement even if the egg is removed).
releaser
specific stimuli that trigger the behavior within fixed-action patterns
the egg in Greylag goose rolling
Some fixed-action patterns can become ‘hijacked’ (exploited) by
supernormal stimuli.
something can exploit this behavioral trait
has to be similar but exaggerated
supernormal stimuli
examples?
A stimulus that shares important traits with the releaser may dominate the response.
• An oystercatcher will normally sit on the largest egg available – even if it is not one of it’s own.
• The Australian jewel beetle was attracted to discarded beer bottles because it had the right color and
dimpled texture of a female.
Humans exhibit a number of fixed-action patterns.
- Babies will grasp (strongly) when something touches their hands. Head-turning and sucking reflexes also aid in nursing.
- Babies have a respiratory occlusion reflex triggered by low air flow.
- Yawning is thought to be a fixed action pattern, but the explanation is less clear.
respiratory occlusion reflex
triggered by low air flow
if it’s not getting enough oxygen it will exhibit a system of behaviors: − Pull back head, wipe face, and cry.
Beyond fixed-action patterns that are innate, animals have
other constraints on their learning.
species-specific preferences
critical periods
— Imprinting
species-specific preferences for certain types of stimuli
pigeons -> vision,
rats -> taste
critical periods (also called sensitive periods)
Some behaviors seem to be learned at specific times of
development.
are marked by rapid learning of specific skills or capacities.
Perhaps they have more plasticity (neural architectural is malleable to learn that new skill) at that time.
Imprinting
is a fast attachment made shortly after birth.
e.g. baby ducks, will attach to the first they they imprint to (even if that’s not their mother)
critical periods - birds
Some birds seem to learn their species-specific
birdsong in specific time windows after hatching.
• The white-crowned sparrow must hear its species’
song only when 10 – 50 days old (Marler, 1970).
- Won’t develop its own song until 150 – 200 days.
- Can learn from a recording of the song.
- Cannot learn the song of the closely related song sparrow.
- It is suggested that it has an internal template that needs to be activated by strongly similar songs and can later provide feedback for its own song - if it doesn’t get it within a particular window of time it is gone.
has it’s own internal feedback created from that template!
Phonemic discrimination
Are there critical periods in human language development?
(e. g. hearing the difference between /l/ and /r/) exists in all children up to 8 months (on average).
- -> Afterwards, the ability diminishes without exposure to the language sounds (you have to build up that category internally at a certain time).
• So, you may never get the same level of perceptual
fluency if you learn a second language later in life.
• An abused child “Genie” was deprived of language input until age 13. Her language skills never developed beyond simple utterances (e.g. “want food”).
Human language seems to be similar to learned birdsong in that
learning a language earlier leads to much better performance.
Newport (1990)
suggested the “less is more hypothesis” in which
learning is limited to simple mappings at first so it isn’t overwhelmed by computational complexity (want a young child to make the simplest possible associations: one sound to one object).
- Can only learn a few symbols (words) to a few words at a time.
- Suppose you have 3 words: a, b, c.
- You would like to map them to the right, three different meanings: m, n, o.
- There are many (7 * 7 = 49) possible mappings!
- If language mapping emerges at a specific time window and with simple mapping (say, due to less myelinated neural connections available - neurally limited), you limit the possibilities (the computational complexity): you can really only do a few. if you build up those simple associations, then later on that organism has the ability to gain more and more once it’s gained that
if you don’t do those simple mappings early on, that’s when the animal runs into trouble
• This is still very controversial…
Coming from the tradition of behaviorism, there were said
to be two basic forms of learning (we now recognize
many more).
- Non-associative learning
2. Associative learning: “What goes with what”
Non-associative learning
− Habituation
− Sensitization
− Perceptual learning
Associative learning: “What goes with what”
“What goes with what”
− Classical conditioning
− Operant conditioning
Habituation
is a reduction (or halting) of an organism’s behavioral response to repetitive, predictable stimuli.
- Innocuous (mild) stimuli work better for habituation.
- It is usually stimulus-specific (has to be the same sound).
• Usually short-term, but can be made long term
with many repetitions over many sessions.
Dishabituation
Spontaneous recovery
Dishabituation:
A novel stimulus interrupts the habituated stimulus, causing
the original behavioral response.
e.g. bom bom bom bom bom BING bom bom bom bom
Spontaneous recovery
After a long delay in stimulus presentation, the response may return to full strength.
Habituation will happen very quickly to mild stimuli.
mouse in cage: habituation to the acoustic startle reflex.
