The Discrete and Continuous Brain: From Decisions to Movement—And Back Again

Abstract

To act upon the world, creatures must change continuous variables such as muscle length or chemical concentration. In contrast, decision making is an inherently discrete process, involving the selection among alternative courses of action. In this article, we consider the interface between the discrete and continuous processes that translate our decisions into movement in a Newtonian world—and how movement informs our decisions. We do so by appealing to active inference, with a special focus on the oculomotor system. Within this exemplar system, we argue that the superior colliculus is well placed to act as a discrete-continuous interface. Interestingly, when the neuronal computations within the superior colliculus are formulated in terms of active inference, we find that many aspects of its neuroanatomy emerge from the computations it must perform in this role.

An Investigation of the Free Energy Principle for Emotion Recognition

“…The brain's neuromodulatory mechanisms appear to perform gain control operations in a manner analogous to the precision optimization (see Box 4). This underwrites attentional processing in computational models of predictive coding and active inference (Parr and Friston, 2013).…”

Deep Active Inference and Scene Construction

“…This can be analogised to the idea of perception as inference, wherein perception constitutes optimizing the parameters an approximate posterior distribution over hidden states , under a particular policy π . 3 In the context of neural systems, it is theorised that the parameters of these posterior beliefs about states are encoded by distributed neural activity in the agent’s brain (Friston, 2008; Friston & Kiebel, 2009; Huang & Rao, 2011; Bastos et al, 2012; Parr & Friston, 2018c). Parameters of the generative model itself (such as the likelihood mapping P ( o | s )) are hypothesised to be encoded by the network architectures, synaptic strengths, and neuromodulatory elements of the nervous system (Bogacz, 2017; Parr, Benrimoh, Vincent, & Friston, 2018; Parr, Markovic, Kiebel, & Friston, 2019).…”

An Integrated World Modeling Theory (IWMT) of consciousness: Combining Integrated Information and Global Neuronal Workspace Theories with the Free Energy Principle and Active Inference Framework; towards solving the Hard problem and characterizing agentic causation

“…This neurobiological account in which neuronal systems dynamically move between more integrated and segregated processing is referred to by Dehaene et al (2001; as Global Neuronal Workspace Theory (GNWT). From an FEP-AI (and IWMT) perspective, these phase transitions may correspond to discrete updating and Bayesian model selection with respect to perception and action (Friston et al, 2012a;Hohwy, 2012;Parr and Friston, 2018b). GNWT has been increasingly described in terms of Bayesian inference (Dehaene, 2020;Mashour et al, 2020), including in a recently proposed Predictive Global Neuronal Workspace model (Whyte and Smith, 2020).…”

“…This hypothesis is consistent with Clark's (2018) suggestion that coherent and precise inference stems from requirements for engaging with environments via sensorimotor couplings (Clark, 2016). Along these lines, by enabling the generation of inferences with rapidity and reliability, SOHMs could afford approximate models capable of guiding action-perception cycles and decision-making (Von Uexküll, 1957;Fuster, 2009;Madl et al, 2011;Vul et al, 2014;Linson et al, 2018;Parr and Friston, 2018b). Further, these sensorimotor engagements may promote SOHM-formation by providing coherent sources of correlated information, so affording the possibility of learning even more sophisticated models (Pfeifer and Bongard, 2006;Safron, 2019cSafron, , 2019a.…”