Predictive Processing Part 2

1. What is Precision Weighting?

The process by which the predictive system assigns greater or lesser weight to prediction errors depending on how reliable or informative the sensory signal is estimated to be

Why does we need Precision Weighting in Predictive Processing?

Any mismatch that reconstructs leads to instability, sensory systems encounter noise and uncertainty constantly so error needs to be weighted

1. What’s High-precision Sensory Evidence?

You trust the evidence and prediction gets more weight, current prediction may need substantial updating

1. What’s Low-precision Sensory Evidence?

The evidence isn’t reliable, prediction error gets less weight and existing prediction/prior exert relatively greater influence

1. Connection, how does Precision Weighting complete Bayesian Inference?

Precision shapes perception so it’s sensitive to the estimated reliability of the available information

2. Connection, how does Precision related to attention?

If a particular source of information is judged highly relevant/reliable, the system can effectively increase the attention towards it

3. Connection, how does Precision related to attention in terms of Predictive Processing?

Possible way of understanding aspects of selective attention, prioritizing some error signals over others than treat all sensory information equally88788878

2. What does Precision in Precision Weighting?

Concerns the estimated reliability of a signal and therefore determines how strongly a prediction error should be weighted

1. Relevance, how does Precision Weighting related to Predictive Processing?

Allows PP to deal with uncertainty, so the brain shouldn’t update its model equally in response to reliably sensory evidence and noisy

2. Connection, how does Precision Weighting connect to Bayesian Inference?

Precision helps determine the relative influence of existing expectations and incoming evidence

What does Perceptual Inference mean?

Change the model/prediction to better fit sensory input

What does Active Inference mean?

Act to change sensory motor so it better fits predicted states, you act on the world to bring sensory input alignment with the predicted state

How does Active Inference relate to Predictive Processing?

Gives PP a route from perception to action rather than treating them as completely separate systems

1. What happens in Sine-wave Speech?

It strips speech down into unusual acoustic patterns so it can sound like meaningless sounds without knowing what the sentence says

2. What happens in Sine-wave Speech?

Once you’re told what the sentence says, you can suddenly hear the words and it’s difficult go to back to meaningless sounds

What does Sine-wave Speech help demonstrate?

How prior knowledge can dramatically alter perception of an ambiguous sensory signal

2. Relevance, how does Predictive Processing explain Sine-wave Speech?

The sensory stimulus hasn’t changed, what changed was your prior knowledge

3. Relevance, what does Sine-wave Speech support in Predictive Processing?

Perception results from an interaction between bottom-up sensory evidence and top-down expectations

1. What happens in the Hollow-Face Illusion?

Look at the concave inside of a face mask and the sensory information indicates face curves inward yet people perceive it projecting outward

2. What happens in the Hollow-Face Illusion, why is the perception different?

Because your generative model has a very strong prior that faces normally bulge outward

So what does the Hollow-face Illusion show about why we perceive faces the way we do?

When that sensory evidence is compatible with multiple interpretations that strong prior can pull toward normal outward face

What does the Hollow-face Illusion help demonstrate?

The influence of prior expectations on perceptual inference

What does the angry-Jack example demonstrate?

A potentially self-reinforcing perception-action cycle, in which prior beliefs influence precision weighting, perception guides action, and action alters subsequent sensory evidence

Why is the angry-Jack example so theoretically important?

Because prediction error minimization doesn’t automatically guarantee truth

2. Why is the angry-Jack example so theoretically important, what’s one reason?

A system can potentially prediction error by changing its model to fit the evidence

3. Why is the angry-Jack example so theoretically important, what’s one reason?

Through action, changing the evidence it encounters so that if fits its model

How does the angry-Jack example relate to Active Inference?

If prediction guide action, our generative models don’t merely affect how we perceive but how we change and act

So in the angry-Jack example, what can be said about inaccurate models?

They can potentially become self-reinforcing