The secret life of predictive brains: what's spontaneous activity for?

Abstract

Brains at rest spontaneously generate dynamical activity that is not mere noise but highly structured in space and time. As our most vivid dreams exemplify, spontaneous brain activity can give rise to the most sophisticated cognition. We suggest that spontaneous brain activity as in rest, mind wandering, or dreaming, underlies top-down dynamics of the generative models used to engage in everyday tasks. However, these models have different functions during active tasks (online) and in the absence of overt behavior (offline). During active tasks, generative models provide top-down predictive signals to carry out perceptual, cognitive, and motor tasks. When stimuli are weak or absent as in dreaming or eyes closed rest, top-down dynamics support the optimization of the generative model for future interactions by maximizing the entropy of explanations and minimizing model complexity. Specifically, spontaneous fluctuations of correlated activity within and across brain regions may reflect transitions between "generic priors" of the generative model, which represent low dimensional latent variables and connectivity patterns of the most common perceptual, motor, cognitive, and interoceptive states. These states are not tied to any specific explanation but useful to code future interactions, both familiar and novel. Therefore, even at rest, brains are proactive and predictive.

Disentangling Predictive Processing in the Brain: A Meta-analytic Study in Favour of a Predictive Network

“…Moreover, since the DMN is located at the extreme in a continuum of integration and hetero-modal functioning within the human connectome 91,92 , it is even more surprising that it did not result from our functionally heterogeneous PC-related dataset. A possible reason for its absence could be that, under the hypothesis that the DMN is responsible for the integration of predictions in former internal modelsacting in a sort of "autopilot mode" 93,94 , almost no experimental paradigm, included in the current analysis, tested this kind of automatic activity. Indeed, given the functional nature of the DMN, the speci c impact of prediction on this network is di cult to test experimentally.…”

Disentangling predictive processing in the brain: a meta-analytic study in favour of a predictive network

“…Moreover, since the DMN is located at the extreme end of a continuum of integration and hetero-modal functioning within the human connectome 91 , 92 , it is even more surprising that it did not result from our functionally heterogeneous dataset. A possible reason for its absence could be that, under the hypothesis that the DMN is responsible for the integration of predictions in prior internal models—acting in a sort of “autopilot mode” 93 , 94 —almost no included experimental paradigm tested this kind of automatic activity. High temporal resolution techniques, computational, and meta-analytic approaches to functional neuroimaging data can be valuable tools to investigate the role of the DMN in predictive processing in the future.…”

Modelling ourselves: what the free energy principle reveals about our implicit notions of representation

“…A second line of argument in favour of the representationalist interpretation of sensorimotor models is that they afford forms of internal manipulation and inference-for example, when they are used to steer sensorimotor predictions of action consequences-and for this they are sometimes called "action-oriented" or "embodied representations" (Clark & Grush, 1999;Grush, 2004;Pezzulo, 2008). This line of argument is in agreement with many foundational works in cognitive science, according to which internal representations are important because they can be used prior to or instead of acting in the real world-and even "vicariously", in the absence of their referent (Craik, 1943;Piaget, 1954).…”

“…On the one hand, one can argue that the notion of sensorimotor model can stretch to any kind of complexity (Bruineberg et al, 2016), as sensorimotor loops become able to incorporate increasingly distal consequences. Furthermore, the sensorimotor dynamics created by sensorimotor models may be progressively internalized to support mental operations detached from the sensory-motor cycle (Hesslow, 2002;Ito, 2008;Buzsáki et al, 2014;Pezzulo, Donnarumma, et al, 2017;Pezzulo, Kemere, et al, 2017;Pezzulo et al, 2020;Stoianov et al, 2020).…”

“…Another way to address the issue of what a representation is and what it is not is by asking what functional role representations play within a hierarchical architecture. An idea that dates back at least to Piaget (1954) is that representations should vicariously "stand for" something external in its absence and afford vicarious operations, i.e., mental operations using an internal vehicle that are executed before acting on the external referent of the vehicle (e.g., consider mentally which route to take to go home), or even when the external referent of the vehicle is absent (e.g., consider mentally whether one would enjoy eating a pizza-with no pizza is in sight). 16 Relatedly, this functional role aspect of representations has been expressed in terms of whether the agent's internal operations are detached from its action-perception cycle and hence autonomously generated, versus determined or sustained by external stimuli (Gardenfors, 1996;Pezzulo & Castelfranchi, 2007).…”