Lecture 8 Learning And Knowledge Flashcards
What is salience
Tells how much the cue grabs attention
In the absence of validity, high salience cues will attract attention
4 effects that shows us attention is important for learning?
Trade offs between salience and validity
Blocking
Highlighting
Learning rules of different complexity
Posner cuing task? What’d he Do
Shown lots of arrows, need to detect which arrow will more likely point to the cue over time.
Arrow that points to the cue = high validity
Can use two different arrows - high salience arrow (flashing arrow) and low salience arrow (non salience) —> initially people rely on the salience of the cue (flashing), but then learn about validity which is built up over successive trials
What happened when Posner manipulated the validity of his cues?
When the high salience cue was more valid - people learnt to use this really fast
When high salience and low salience had the same validity - there was a trade off, high salience still used more, less than before, and low salience more.
When low salience was more valid - people learnt to use the low salience cue more, but it wasn’t that much more than high salience cue (still a trade off )
How does blocking work?
A -> X
A + B -> X
C + D -> Y
B + D -> Y
Happens because since A was shown alone first, all attention was shifted to A for predicting X, none left for B. So when B and D is shown, D will drive response
Therefore attention important to leRning
How does highlighting work?
A + B -> X
A + B -> X
A + D -> Y
B + D -> Y
Happens because A and B are already paired with X, all attention shifted to D because it alone predicts the unusual event Y.
Interesting bc B has been shown twice as often
What does simple learning theory suggest?
Co occurrences lead to a strengthening between cues and outcomes.
(Associations over time)
Increases in strength is determined by a learning rate, and the weight of old cues, and the prediction of would happen, and what actually happened.
Rescorla Wagner model / delta rule
Attentional learning theory differs bc it also incorporates relevance into the weights (aka incorporates attention)
What is the rescorla Wagner model?
W(new)= w(old) + alpha(prediction-actual)
We update our old weights by the difference of what we predict and what actually happens
Alpha is learning rate
Attentional learning theory
Selective attention allows for some aspects of an event of be associated more strongly than others.
What did the krusche study, with the stimuli that varied on height and position of a line (Represented in a hexagon) show?
Used a filtration condition - where only height of the shape needed to be attended to
And condensation condition - where both height and position of line need to be seen
People learnt slower in the condensation task and never really reached the same accuracy as the filtration, because had to attend to 2 dimensions.
Predictions of this matches only when SELECTIVE ATTENTION is accounted for in the theory.
Therefore attention is very important to learning - we can exclude things!!!!
What is fast mapping?
Learning by exclusion, based on what is already known
Expectations provide hypothesis which we use to explain what we’re trying to learn, and evaluate our experiences
What did blicket detector test show us?
Blicket is anything that activates the box
Demonstrates fast mapping by building up prior knowledge - expectations are hypothesises that you use to evaluate the world.
Can bias us with the blickets.
What are the two ways in which prior knowledge influences our hypothesises?
Holmesian deduction
And
Judicial exoneration
What’s holmesian deduction?
Once you eliminate the impossible, whatever remains, however improbable must be the truth
This type of reasoning is used in cognition
What’s judicial exoneration?
If you have a hypothesis that explains what you want really really well, then you can let everything else go.
So if one suspect confesses, then let all the other suspects go.
Also called inference to the best explanation.
