LECTURE FOURTEEN Flashcards
BEHAVIORISM
- Focus on observable behavior
- Mostly focused on learning
- What is learning?
- Associating stimulus with response
- Associating concepts with other concepts (semantic
network) - Associating inputs with outputs (machine learning)
FORMING ASSOCIATIONS
- Key with learning is to associate things together
- From classical conditioning to elaborative encoding
- Hebbian learning
- Neurons that fire together wire together
- Correlative, associative learning
- Note this type of basic associative learning is very sensitive to correlation – but not cause
CLASSICAL CONDITIONING
- Unconditioned stimulus (UCS)
Some stimulus that elicits a natural behavior response - Unconditioned response (UCR)
The response elicited from the UCS - Conditioned stimulus (CS)
Some stimulus that doesn’t elicit the UCR originally (neutral stimulus), but will after training - Conditioned response (CR)
The same initial UCR now elicited by the CS after training
CLASSICAL CONDITIONING (CC)
- Pair CS (bell) with UCS (food) to elicit UCR (salivation)
- CS (bell) becomes associated with UCS (food)
- After repeated pairings, CS (bell) now elicits CR
(salivation) - CS now associated with UCS (triggers CR because UCS
triggers UCR)
ASPECTS OF CC
- Aspects of conditioning:
- Acquisition – learn CS/CR association
- Extinction – remove UCS – now CS/CR association
weakens - Recovery – restoration of CS/CR association (can be
elicited by repairing UCS with CS, or spontaneous) - Generalization – stimuli similar to CS can also elicit CR
- Discrimination – stimuli different to CS do not elicit CR
- Note that generalization and discrimination are in
contest for any CS
OPERANT CONDITIONING
- No longer passive association – now the animal must
“do” something (“operate”) - Basic procedure
- Animal performs some action
- Animal is rewarded for this action – behavior reinforced
- Animal is punished for this action – behavior undermined
- Reinforced behaviors become more common and sought out, undermined become less frequent and avoided
PLEASURE
- Electrical stimulation of certain brain regions leads to acute pleasurable feeling; rewarding sensation
- Reward circuit in the limbic system
- Mesolimbic dopamine pathway
- VTA to nucleus accumbens
- Rats (and people…) will compulsively “press the lever”
when it stimulates this area - Appears like addiction
- Rats will forego food (and starve) in favor of stimulation
REWARD CIRCUIT
- Purpose of the reward circuit
- Provide behavioral sense of reward upon fulfilling unmet biological need
- “Reinforcers”
- Food, sex, comfort/safety, belongingness
- Can be more abstracted; money, high fives, Instagra followers
- Tightly linked with learning and memory
- Hippocampus, amygdala
- Must learn/remember what you did to achieve reward
- People, places, objects, activities, actions all become associated with reward
LEARNED HELPLESSNESS
- What if, no matter what the animal does, they are
punished? - Learned helplessness
- Punishment regardless of action
- No way to avoid punishment
- After some time, animal will no longer “try”
- Animal has learned that it is helpless
LEARNED HELPLESSNESS THEORY OF DEPRESSION
- Is this what is happening in depression?
- Bad things happen to a person and are judged as completely outside their control
- Chronic, un-relievable stress is like unavoidable punishment
- Nothing the person does avoids punishment
- Natural response? Stop trying and accept it – learned helplessness
- Note this is much more than sadness - “Why bother”
- How to treat? Cognitive therapy and/or drugs that promote learning/unlearning (plasticity and neurogenesis)
- Antidepressants and psychedelics both have this in common
DRUGS AND BEHAVIORISM
- Drugs are “natural reinforcers” – DoA operate directly on brain’s reward system
- Thus, the drug reward is both a reinforcer for
operant conditioning and an UCS in classical conditioning - Drugs themselves then become CS and sought after
DRUG ADDICTION
- Implications of behavioral drug addiction viewpoint:
- Addiction is a learned behavior
- Context of drug use becomes CS – triggers CR, cravings, desire to take drugs
- Treatment from a behavioral perspective:
- Eliminate CS contexts – go to rehab
- Eliminate CR – block or interfere with drugs from having their rewarding effects in the brain
- Chantix, naltrexone, etc.
- Once drug use is extinguished, must watch for
spontaneous recovery of learned association (relapse)
ADDICTION IN GENERAL
- These same principles hold true for any
addiction, not just drugs. - Gambling addiction
- Food addiction (particularly sugar)
- Social media addiction
- All involve learning, all involve some sort of reward
PROGRAMMING
- Two main methods for programming computers
- Algorithms
- Line by line coding (standard programming)
- Explicit instructions on how to analyze things
LEARNING IN HUMANS
- For adult humans:
- Learning typically involves a “teacher” or some sort of
desired outcome that can easily be checked against - Supervised learning
- Biologically relevant learning involves reinforcement
- Do something clearly biologically relevant (e.g., eat food, have sex), get a reinforcing effect in the brain (reward circuit)
LEARNING IN MACHINES
- How do (some) neural networks learn?
- Provide input and desired output (training data set)
- Use a learning algorithm (Hebbian, back-propagation, etc.) to change connection weights to get the right answer
- Provide input without desired output and evaluate accuracy of the model (testing data set)
- With enough good training data, network can learn to generalize to new stimuli!
- Reinforcement learning – “right answer” is reinforced (not explicit)
DEEP LEARNING NETWORKS
- Neural networks with more than one hidden layers = deep networks
- Take a longer time to learn and process info
- Massive amounts of computational power needed to train
- Can perform complex analyses (like object recognition!)
DEEP NETWORKS
- This sounds like vision in the brain!
- Network provides emergent solution to
complex problem - If we can “peek into” hidden layers it’s like
opening the black box of cognition
DEEP DREAMING
- Take a network that has learned to classify objects
- Peek into hidden layers to see how it is doing it
- Network has learned to identify animals (but
nothing else) - Show it a picture of something without animals
- It “sees” everything as animals
- Recurrent connections and feedback
- Things become more and more animal like
INCEPTIONISM
What if you only feed in static?
And then feed that back in…
And then feed it back in again…
AI ART
- Currently, multiple programs using image recognition
software to generate images - Train network with many labeled pieces of art
- Generate new piece through a written prompt
- Good at some things…bad at others (text!) but improving
SO WHAT DOES THIS ALL HAVE TO DO WITH THE MEMORY AGAIN
Conditioning (classical and operant) and associative
learning is not explicit memory
* Can be demonstrated in all animals
* Does not require sense of self….probably
* Can be remembered without being consciously
recognized (esp. classical conditioning)
* Does not require effortful encoding and elaboration –
just repetition
Part of implicit memory
* Memory for skills, procedures, and basic learning
associations
ANTEROGRADE AMNESIA
- Inability to form new explicit memories
- Implicit memory still intact – can form new
procedural memories and use old procedures: However, no explicit memory of learning a new
procedure - Damage to the hippocampus (subcortical):
- Surgery
- Korsakoff’s Syndrome (alcoholism complications)
- Viral infection
WHAT IS IMPLICIT MEMORY
- Any change in mental processes or behavior as a result of experience that is not explicit memory
- Probably not a single system
- many brain processes for storing experience have built up over evolutionary history
- explicit memory may be just the most recent “layer”
- Types of implicit memory
classical conditioning
operant conditioning?
perceptual fluency
PRIMING
- Brief exposure to a stimulus aids subsequent
processing of that stimulus (or related stimuli) - Pairwise priming
- Implicit priming
- Masked priming
- Semantic priming
- Block priming
EXPLICIT MEMORY
- Variable tasks
- Effortful recollection
- Able to communicate
- Complex consequences and prediction
- Adapt to changing situations
IMPLICIT MEMORY
- Stereotyped tasks
- Automatic access
- Unable to communicate
- Simple consequences and conditioning
- Automatize frequent tasks