Learning to associate the probability + value of behavioural outcomes with stimuli is essential for rational decision making.However in demanding cog conditions access to learned values might be constrained by limited attentional capacity.
1 of 13
Raymond and Obrien (2009)
Shows that model-based and model-free can occur at the same time.
2 of 13
Learning Task
Shown pairs of faces, told to choose a face + win as much as you can. Limited instructional info.
3 of 13
Three Pairs of Faces
There was actually 3 different pairs of faces. One pair where if you picked one there was an 80% chance you'd win + other you'll win 20% of time. 2nd pair of faces will lose 20% of time with one face + 80% of time with other. 3rd pair where nothing
4 of 13
Three Pairs of Faces
happened as a mere exposure control. P's learn the monetary associations.
5 of 13
Attention Task
2 stimuli, have target 1 then a mask and an interval of 200ms or 800ms. This is then followed by a face then a mask. p's needed to say whether it was a new face or one from previous task. Target 1 was simple circle and square pattern.
6 of 13
Attentional Blink
Measures capacity for central resources. Should be more capacity in 800ms interval.
7 of 13
200ms Results
Predicted that in 200ms interval, should be unable to excel in this task. Should have trouble recognising face as old or new. Hardly any capacity to make judgement about 2nd face after engaging in 1st target. Do worse as haven't got as much central
8 of 13
200ms Results
capacity. Attentional blink is found when loss associated stimuli are displayed: shows when something has value it gains access to central resources, has priority. Example of valence sensitive. Example of model based learning. Doesnt work for
9 of 13
200ms Results
negative things, stimulus with high subjective value have no attentional blink.
10 of 13
800ms Results
Should do well, no problem with central resources. Actually see a u-shaped function of learning. Do well when high expected value+ when low, but dramatic dip in middle. Processing is easy for stimuli with high pos or neg expected value: value
11 of 13
800ms Results
insensitive. Model free learning as its valence insensitive. Doesnt matter if its bad or good, it's merely the high quantity associated with it (winning or losing lots)
12 of 13
Conclusions
Supports notion of value sensitive and insensitive mechanism in brain and that both can happen at the same time.
13 of 13
Other cards in this set
Card 2
Front
Raymond and Obrien (2009)
Back
Shows that model-based and model-free can occur at the same time.
Comments
No comments have yet been made