Cog & Bio Uncertainty And Prediction Flashcards
(19 cards)
What is perceptual uncertainty?
Uncertainty about what we are seeing - illusions or in everyday
What is uncertainty about events?
Being unsure what it going to happen
What is uncertainty about causal structures?
What is causing this thing to happen
What is uncertainty about intentions / social uncertainty?
Why did they do that
What is one solution to resolving uncertainty?
Normative models of cognition
What is normative models of cognition?
Specify the best possible solution to a problem - the ideal decision-making process
What is one example of a normative model?
Bayesian inference
What is bayesian inference?
Combines prior knowledge/ beliefs / predictions with new data / experiences to form updated beliefs
ONE METHOD FOR UPDATING BELIEFS BASED ON PROBABILITIES
Likelihood =
New data
Posterior =
Updated belief based on the new data (likelihood)
Very slight shift away from prior towards likelihood
What is certainty of information?
Putting more trust in the certainty because you see it with your own eyes
What are the benefits of normative approaches to cognition?
- probabilistic (Bayesian) approaches are the best way to handle uncertainty
- they describe the ideal decision - making process
- we would expect humans to behave probabilistically due to the huge survival advantages in an uncertain world
What are the limitations of normative approaches to cognition?
- normative arguments are controversial - debate if we do function in this normative way
- evolution doesn’t require perfection, but rather ‘minimum viable product’
- arguable human cognition is far from optimal (base rate fallacy)
- may be a better description of some areas than others
What is bayesian inference and multisensory integration?
According to normative models, we integrate the information we have to generate the best possible solution e.g integrate prior beliefs and new info
Multisensory integration - visual and haptic info
How is visual and haptic info combined to generate the best possible guess
What is the most simple approach?
Take the average - equal weighting, and put less weight on info that is less certain - however this is a poor fit to the data
What is the best approach, according to normative models?
Take a weighted average, precision weighting - better fit to the data
Social perception = ?
Expectations + sensory input
What is precision weighting?
How much the brain relies on expectations vs sensory input
