VL18 Predictive Processing in development

• feedback: top down, expectation

• feedforward: bottom up

• feedback - feedforward = prediction error

• feedforward and feedback denote direction in which information flows

• prediction = information that flows in feedback direction

• prediction error = information that flows in feedforward direction

• computed as difference between feedback and feedforward

• feedback - feedforward

1. prior meets evidence

2. model fine tuning with precision weighting

3. hierarchical structure and functionally distinct units for prediction & prediction error

• statistical learning

• associative learning —> conditioned tone-eyeblink response with airpuff

• habituation to repeated stimuli

• dishabituation to novel stimuli

• violation of expectation effects

they are sensitive to regularities and highyl motivated to explore -> seek to form predictive models to reduce uncertainties

• sensorimotor and cognitive capacities

• sophisticated internal models of the environment

• predominance local connections and slow development of long-range interactions

option a) with feedforward -> precursor to feedback mechanisms

option b) fundamental feedback architecture as scaffolg for development of efficient internal models

• 36 6-month olds

• fnirs

• unexpected visual omission

• control condition a+ v-

• a+v+ —> v+

• activation in temporal roi and occipital roi

• already in 6 month olds —> more oxygenated hemoglobin in temporal and occipital lobe than in control condition

• potentially not long-range connections yet but through associative learning (basal ganglia)