Predictive Processing Extension

What does Interoception mean?

The perception of internal bodily states, such as heart rate, blood sugar, visceral activity, and musculoskeletal signals

What does Interoception play in Predictive Processing?

The brain is not only predicting external sensory states, it also integrates inward-looking bodily signals with outward-looking sensory information and prior to form a best guess

Why is Interoception relevant in Predictive Processing?

Makes PP more embodied because bodily states are part of the predictive system rather than something added after cognition

1. What’s Valence?

Whether an experience feels positive or negative, pleasant or unpleasant

2. How is Valence described?

As the positive/negative quality of experience and as a kind of bodily signal about how the organism is doing

1. How is Valence (wrongly) explained in terms of PP, why?

A correct prediction feels good while a prediction error feels bad, that is too simple because we can enjoy surprising events where prediction error increases

2. Relevance. Instead, what's a more subtle explanation for how valence is understood in terms of PP, why?

PP is now trying to explain not only what we perceive but also why certain situation feel good or bad and motivate behaviour

3. Relevance, how is valence then summarized in PP?

Interprets the felt quality of experience as embodied feedback about how well prediction is progressing

How is error evaluated in Error Dynamics?

Where errors are evaluate relative to expectations about the rate at which similar errors have

What does Error Dynamics mean?

The change in the rate of prediction-error reduction, allowing the organism to track not only whether it is wrong but also how efficiently it is improving

1. Relevance, how does Error Dynamics connect to Valence?

Valence gives the organism a felt signal about error dynamics

2. Relevance, if predictive progress is unexpectedly improving then there’s positive valence, what’s the opposite?

If predictive progress is going worse than expected, there’s negative valence, not merely less error equals to happiness

In Error Dynamics, what’s First-order Prediction Error?

Information in the system, used to update predictions and guide behaviour

In Error Dynamics, what’s Second-order Prediction Error?

Information for the system about how effectively and efficiently prediction-error reduction itself is going

What’s the Goldilocks Zone?

Intermediate level of uncertainty or complexity where prediction errors are challenging but still learnable and reducible

When studying for example, what would you want to seek in question?

A challenging question just slightly above your current skill level so you can make errors but figure them out

What does the Goldilocks zone create?

Prediction error, successful reduction, predictive progress, a positive affect that leads to a desire to continue

What does the Goldilocks Zone result or explain?

Curiosity, exploration, play, learning

What’s the objection in the Dark Room Puzzle?

PP often sounds like organisms try to minimize prediction error, so what if we were put in the most predictable environment possible

How does the Dark Room Puzzle perceive Predictive Processing, what’s wrong about it?

PP isn’t about having as little error as possible at every instant, error dynamics matter

1. What is the response to the Dark Room Puzzle?

There’s almost no opportunity to improve your predictions and predictive progress stagnates

1. What is the responses to the Dark Room Puzzle, what actually happens instead?

Organisms are drawn toward situations where there’s prediction error and the error can be successfully reduced, the Goldilocks zone

What is generally concluded about organisms from the Dark Room Puzzle?

Organisms actively seek manageable errors that can be reduced and that affect helps tune them toward good opportunities for error reduction

How does Binocular Rivalry occur?

When each eye is simultaneously shown a different visual image that causes a conscious perception to alternate between the two rather than blending them together

What happens in Binocular Rivalry?

One eye may see a house and the other a face, instead of experiencing a mixture of the two people tend to perceive in a bi-stable alternation

Why is Binocular Rivalry relevant to Predictive Processing?

PP explains this as a competition between hypotheses and the system selects the hypothesis that best explains the sensory input

2. Why is Binocular Rivalry relevant to Predictive Processing?

That winning hypothesis sends top-down predictions that “explain away” that sensory information compatible with it

3. Why is Binocular Rivalry relevant to Predictive Processing?

One eye is still supplying incompatible information, some prediction error remains and travels upward, that unresolved error makes the competing interpretation more successful so perception flips

Why is it we don’t perceived a merged perception in Binocular Rivalry?

The hybird is often an improbably hypothesis even thought it could technically explain pieces of both sensory signals

What’s a Hyperprior?

A very general prior about how the world normally works

How do houses and faces illustrate a Hyperprior?

Houses and faces do not normally occupy the exact same place at the same scale at the same time so it’s an inprobable hypothesis

In relevance, how does Binocular Rivalry support Predictive Processing view?

That perception is an active competition among hypotheses constrained by both sensory evidence and prior evidence, rather than just passive sensory input