Conditioning and Cognition P2 Flashcards
Squire’s types of memory
Declarative and nondeclarative
Declarative memory
conscious retrieval of facts and events, manifests in episodic and semantic knowledge
Nondeclarative memory
procedural memory
Why talk about memory in learning?
theorist aren’t sure about the best way to define memory and learning and where the distinctions are
Propositional vs non-propositional leanring focuses on…
Lovibond and Shanks
Focuses on the symbolic nature of learning
Propositional learning
Assumes conceptual processing and symbolic representation.
Involves higher cognitive functions.
Non-propositional learning
Don’t need any symbolic connection; learning that does not involve “symbolic” representation.
Example of propositional learning
“The tone is usually followed by a food pellet”
Have to be able to classify events
Example of non-propositional learning
CS-US and CS-CR links generate behaviour via more direct mechanisms
Shouldn’t require conscious expectancy or conscious control of action
What could non-propositional learning be interrupted by? And example
top down control, cognitive system is also involved.
But goes against assumption.
Instance based learning vs rule based learning
Learning instances of stimuli and events, physical features and associative nature - very lawful, extent depends on similarity
vs
Abstraction of a rule, active hypothesis testing, generalises according to the applicability of the rule
Shanks argued learning does not occur without
awareness, but still believes there are important distinctions in types of learning like rule distinction
How to induce rule vs instance based learning
Positive and negative patterning
A - US
B - US
AB - no US
Relatively easy to work out via rule
Difficult by association
Shanks and Darby: Causal learning task
Used negative patterning.
If you learn rule you can apply it to different training conditions of negative patterning
Results of shanks and darby causal learning task
Feature based learning leads to slower association than rule based learning
People who learn quickly and efficiently are rule based learning
Double dissociation
Results of Wills et al follow up study from shanks and darby
Whether rule abstraction is more cognitive dependent
They trained people well, either had a high working memory load or not.
When you have a high load you lose all rule transfer and becomes feature learning (unsophisticated)
Harder to pick up a concept when you are distracted
Model-based vs model free choice
Model-free behaviour - habitual learning, reinforced by the presence of a reward, in line with Thorndike’s law of effect
Model-based behaviour is goal-directed, mental simulation of relations between events,
How to distinguish model based vs model free behaviour
Daw et al (2011) Two-step sequential choice task
What is the two-step sequential choice task?
Daw et al (2011)
on every trial you may receive money, you have to make 2 choices at every trial to see if you make money.
The first choice is abstract stimuli, most of the time you go to green or blue screen depending on choice. (70-30% colour leads from picture)
At colour 2 choices of pictures with different probabilities of reward.
How does the two step sequential choice task distinguish between choice types
Participants should have understood what picture leads to colour and then reward.
If the opposite to what we think leads to the actual reward what does this do to how we think?
Model free prediction we choose depending on rewards
Model based decision there is an interaction. For a common result it follows normal process, for unusual transition we apply model to get back to where we think reward is.
(this assumes they are aware of the 70-30 ratio)
Connectionism
NOT IN THE EXAM
You have processing units devoted to sensory inputs that form links with internal brain regions that aren’t linked to stimuli and outputs to respond, links strengthened and weakened by similar algorithms in simpler learning models.
Simple operations on a sufficiently large scale can lead to intelligence