The Many Roles of Precision in Action

Entropy, Journal Year: 2024, Volume and Issue: 26(9), P. 790 - 790

Published: Sept. 14, 2024

Active inference describes (Bayes-optimal) behaviour as being motivated by the minimisation of surprise one’s sensory observations, through optimisation a generative model (of hidden causes data) in brain. One active inference’s key appeals is its conceptualisation precision biasing neuronal communication and, thus, within models. The importance perceptual evident—many studies have demonstrated ensuring estimates are correct for normal (healthy) sensation and perception. Here, we highlight many roles plays action, i.e., processes that rely on adequate precision, from decision making action planning to initiation control muscle movement itself. Thereby, focus recent development hierarchical, “mixed” models—generative models spanning multiple levels discrete continuous inference. These kinds open up new perspectives unified description hierarchical computation, implementation, action. how these reflect action—from execution—and associated pathologies if estimation goes wrong. We also discuss potential biological implementation message passing, focusing role neuromodulatory systems mediating different precision.

Language: Английский

Inferring when to move

Neuroscience & Biobehavioral Reviews, Journal Year: 2024, Volume and Issue: 169, P. 105984 - 105984

Published: Dec. 17, 2024

Most of our movement consists sequences discrete actions at regular intervals-including speech, walking, playing music, or even chewing. Despite this, few models the motor system address how brain determines interval which to trigger actions. This paper offers a theoretical analysis problem timing movements. We consider scenario in we must align an alternating with external (auditory) stimulus. assume that brains employ generative world include internal clocks various speeds. These allow us associate temporally sensory input clock, and parts clock cycle. treat this as process inferring best explains input. way choices might emerge from continuous process. is not straightforward, particularly if each those unfolds during time has (possibly unknown) duration. develop route for translation neurology, context Parkinson's disease-a disorder characteristically slows down The effects are often elicited clinic by find it possible reproduce behavioural electrophysiological features associated parkinsonism disrupting specific parameters-that determine priors inferences made brain. observe three core disease: amplitude decrement, festination, breakdown repetitive Our simulations provide mechanistic interpretation pathology therapeutics influence behaviour neural activity.

Language: Английский