Decision making and free will Flashcards
Perceptual Decision making
Perceptual decision making involves the judgement a property or identity of your environment. It could be a discrete property (is this a house) or continuous property (is this a beautiful face?). The process is about taking a stimulus and transforming it into an output that somehow reflects the
properties of the stimulus. These simple decisions have a lot in common with more technical scenarios.
Signal detection
Detecting a feature in the presence of noise.
Example:
A military radar operator has to detect planes on the radar. The radar has a baseline brightness according to the map of the area. When a plane enters the radar, it creates small alterations in the brightness of the radar map. The difference in signal is what is used to detect the plane, but it is complicated due to the presence of noise in the signal caused by pidgeons and clouds and amazon’s fucking drones. When the brightness increases beyond a certain level, then its fair to say that it’s a plane, this is setting a criterion, a threshold.
In the brain, this is conceptually the same, except brightness of the radar map is substituted with neurons and action potentials.
Response distribution
The distribution of probabilities that the incoming stimuli is e.g. either a bird or a plane.
What happens if the response distributions overlap?
If the response distributions overlap, it is impossible to avoid errors, as there will be false positives beyond the threshold (e.g. plane-sized birds), and false negatives below the threshold (e.g. bird-sized planes).
The overlap between response distributions determines the amount of discriminability.
ROC curve
The ROC curve is made by moving the criterion along all possible outcome values, and plotting the true positives vs the false positives.
This can be used to determine the criterion based on how we weight the relationship between false negatives and false positives.
In Covid-testing, some people experience false positives. This could be avoided by moving the criterion so that higher values of nasty virus stuff in your nose is needed for a positive outcome. However, that would cause us to miss some true positives as well.
Discriminability
The degree to which response distributions overlap. It can be calculated in different ways and is notated: d’ (d prime)
What are potential sources of variability in perceptual decision making?
Stimulus fluctuations (e.g. Photon noise) Sensory processing (The brain has imperfections and is state dependent, which causes variability that has nothing to do with the stimulus) Central processing (e.g. attention)
Random dot kinematogram
A video showing black and white dots in motion. The motion can be more or less coherent, based on the proportion of dots moving in the same direction.
Participants are shown this video and has to determine whether the coherent dots are moving to the left or to the right. By carrying out this experiment, it is possible to measure at which motion coherence level people are able to access the information.
How was perceptual decision making studied using random dot kinematogram and macaque monkeys?
In the study, monkeys were shown two fixation points, one to the left and one to the right. They were then shown the random dot kinematogram and had to make a saccade either to the left or the right fixation point in order to decide the direction of motion. While performing the experiment, single cell recordings were made in the motion area for neurons responding to the direction of the stimulus. Also, recordings in the LIP of the monkeys which corresponds to the intraparietal sulcus in humans were made, which is an area involved starting a saccade.
The monkeys completed trials with different levels of motion coherence, and the aim was examine the discriminability in the motion neuron recordings. When plotting the curves for proportion of correct answers, they did so using the cell recordings (neurometric) and using the monkeys fixations (psychometric). Interestingly, the neurometric measure was performing better than the psychometric, which shows that in the brain there are cells that contain information which we are not able to use behaviorally.
Common explanation of reaction time distributions
The distribution of reaction times seems to be heavy-tailed. The common explanation of this distribution is that information about some stimuli is accumulated over time.
E.g. when seeing the random dot kinematogram, the sensory system is gathering evidence for either right or left, and when it reaches a threshold of certainty, you answer. The speed of the accumulation is thus dependent on the quality of the stimulus.
How was evidence accumulation expressed in the macaque study using random dot kinematograms?
With the activity measured in the LIP of the macaque (intraparietal sulcus in humans, involved with starting a saccade), the recordings showed that the cells firing rate increased over time in a manner that is coherent with accumulation models. It was clearly shown that accumulation rate is proportional to the strength of the sensory evidence (motion coherence). When the firing rate reached a threshold, the saccade was made.
Is evidence accumulation always necessary?
No, the large focus on evidence accumulation might in part be an effect of the way we design experimental stimuli. There are several examples of perceptual decisions that happen “at a glance” and lead to early choice signals in the higher cortical regions.
Is evidence accumulation always necessary?
No, the large focus on evidence accumulation might in part be an effect of the way we design experimental stimuli. There are several examples of perceptual decisions that happen “at a glance” and lead to early choice signals in the higher cortical regions.
