An introduction to learning Flashcards
What is learning and what is it not
Inferred change in organism’s mental state resulting from experience + which influences organism’s potential for subsequent adaptive behaviour
Sources of know how:
- Learned by experience
- Know how my design (innate-> natural selection on response mechanism)
It is not:
- Performance
- Developement (e.g. sexual responses at adolescence) -> independent of social experience
- motivation (e.g. hunger, sleep, arousal) -> short term
Types of learning
Non-associative learning (responding to repetition of single stimulus)
o eg. habituation, sensitisation
Associative (associated 2 or more events)
o CLASSICAL (Pavlovian) conditioning - associate external events
o OPERANT (Instrumental) conditioning – associate own behaviour and outcomes
Concept learning – acquiring responses to classes of stimuli (concepts)
Relational learning – picking up abstract rules reg. discriminate equal vs different (where sequence of images all look the same or look different)
Specialised learning - Imprinting and song (learning w/ pre-defined structure + function)
Social learning – learning from others
State-dependent learning – learning + performance depend on state during acquisition eg. if hungry mor likely learn to travel to food
Insight learning – sudden reorganisation of experience leading to discontinuous behaviour
Machine learning – reinforcement learning used in AI and robotics
Non-associative learning
Modification of a response to a single stimulus
HABITUATION – decrease innate response to a stimulus w/ repeated presentations
->Adaptive: ignoring of irrelevant events (w/ no consequences of ignorance)
SENSITISATION – strength of response increases w/ repeat exposure to the stimulus
-> adaptive as can protect from harm (e.g. predation)
-> example: Dog taking multiple trips in the car to the vet
Habituation: examples and general principles
Example:
- Chaffinch alarm call to oil
- Rock squirrel to humans taking a photo
- Gill withdrawal in sea hare (deregulation of neurones)
General principles of habituation (presented in Gill withdrawal experiment)
* Shorter intervals between stimuli-> learn faster to habituate
* Stimulus = weaker -> learn faster to habituate
* Dehabituation to new stimuli
* If stimulus is witheld, response recovers.
Habituation is considered non-associative because it involves only one stimulus and one pre-existing response, but it is true learning, not just fatigue, or forgetting, or drop in perception
Associative learning: Classical (pavlovian)
Pre-existing behaviour is directed towards a novel stimulus
- Unconditioned stimulus – biologically relevant stimulus eliciting characteristic response
- Unconditioned response – characteristic response to a US
- Conditioned stimulus - stimulus that does NOT initially elicit UR
- Conditioned response – response elicited by CS after paring w/ US
Example: Pavlov’s apparatus and dog
Conditions:
a) Contiguity – shorter interval between CS and US (altho long time-lag for food aversion post food poisoning)
b) Appropriateness – relevance of CS for US
-> Natural selection constrains associations to those likely to be causally linked
->Eg. rats associated taste (not noise) w/ illness eg. sweet vs noisy water
-> Eg. rats associated noisy water (not sweet water) w/ shock
c) Contingency – correlation between CS and US
-> The CS must predict the US (excitatory / inhibitory)
Associative learning: Operant
Association of own behaviour with the consequences of the behavior.
Example: Thorndike’s puzzle box
- cats press lever by chance + escapes box
- learns to do it w/ more practice + will get faster
Example: Your parents reward you for getting an ‘A’ on a test that requires you to study hard
Blocking
There must be surprise to allow learning
If training begins with a phase in which only CSA is paired with the US (Noise + shock), and is then followed by a phase in which both CSA and CSB are paired with the US (Noise/ light + shock), then CSB fails to produce the CR
-> previous conditioning is preventing future conditioning as no longer a surprise as US predicts CS.
Associative learning model: Rescorla-Wagner Model
associative strength of a CS with a given US is proportional to difference between the max associative strength the US can support and the current associative strength of all the CSs currently present
-> CS must contain new info
Function of classical conditioning
a CS can signal an impending event
Example:
Enhanced aggression or courtship associated w/ flash of light signal that signals removal of barrier between males or male + female
Concept learning
Acquiring responses to classes of stimuli / concepts
Generalisation within a class of stimuli
Discrimination between classes of stimuli
Example
-> pigeons trained to peck at pictures w/ people in – this behaviour reinforced w/ reward
-> learnt to classify based on complex criterion “people” as could do w/ new images (shows not just remembering reinforced images)
How do they do this:
* learning lists
* Feature theory – look for unique feature present in each category
* Exemplar theory – classify image based on their similarity to images previously classified
* Prototype theory – if correspond to average of members of the category (have generalised image that gets better over time)
Relational learning
Animals can discriminate between “Equal” Vs “different” images
Speciliased learning
Imprinting: learning occurring at particular life stage that is rapid + apparently independent of the consequences of behaviour
Example:
a) FILIAL imprinting – newborns learn to recognise parents (eg. nidifugous birds-> ducks follow first adult they see after hatching)
b) SEXUAL imprinting – juveniles learn characteristics of desirable mates when reach sexual maturity
c) SONG learning (Aural inprinting) – nestlings learn song from hearing their fathers
