On the importance of being flexible: dynamic brain networks and their potential functional significances

Abstract

In this theoretical review, we begin by discussing brains and minds from a dynamical systems perspective, and then go on to describe methods for characterizing the flexibility of dynamic networks. We discuss how varying degrees and kinds of flexibility may be adaptive (or maladaptive) in different contexts, specifically focusing on measures related to either more disjoint or cohesive dynamics. While disjointed flexibility may be useful for assessing neural entropy, cohesive flexibility may potentially serve as a proxy for self-organized criticality as a fundamental property enabling adaptive behavior in complex systems. Particular attention is given to recent studies in which flexibility methods have been used to investigate neurological and cognitive maturation, as well as the breakdown of conscious processing under varying levels of anesthesia. We further discuss how these findings and methods might be contextualized within the Free Energy Principle as a fundamental principle of brain organization, and describe potential methodological advances from this paradigm. Finally, with relevance to computational psychiatry, we propose a research program for obtaining a better understanding of ways that dynamic networks may relate to different forms of psychological flexibility, which may be the single most important factor for ensuring human flourishing.

Integrated world modeling theory expanded: Implications for the future of consciousness

“…In considering workspace dynamics as implementing Bayesian model selection, it may be the case that brains obtain the best of both discrete and probabilistic modeling by “dividing and conquering” across different phases of cognitive cycles, or possibly across brain areas ( Gazzaniga, 2018 ; McGilchrist, 2019 ). Alternating workspace modes—potentially reflected by the formation/dissolution of mesoscale connectomic modularity ( Betzel et al, 2016 ; Safron et al, 2021b )—could allow periods where multiple competing and cooperating hypotheses can remain in play, followed by winner-take-all dynamics when this information is integrated into larger scale networks and models ( Cheung et al, 2019 ), and then “broadcasted” back to modules as they re-form.…”

“…In terms of generalized synchrony, direction of entraining influence may potentially switch between peripheral and core networks before and after critical ignition events ( Safron et al, 2021b ). Theoretically, peripheral sensory hierarchies may asymmetrically entrain deeper levels with core connectivity, seeding them with ascending prediction errors, communicated via driving inputs at gamma frequencies.…”

“…If integrated information corresponds to self-model-evidence as described by FEP-AI, then this would be a very special measure of dynamical complexity, potentially indicating the ability of whole systems to be both stable, adaptive, and even autonomous ( Albantakis, 2017 ). Indeed, connections between integrated information and self-organized criticality further suggests that we may be dealing with a measure that applies to all systems capable of not just persisting, but evolving ( Arsiwalla and Verschure, 2016 ; Arsiwalla et al, 2017 ; Hoffmann and Payton, 2018 ; Takagi, 2018 ; Safron et al, 2021b ).…”

“…Perhaps more fundamentally, one of the primary functions of workspaces and their realization by dynamic cores of effective connectivity may be the ability to flexibly bind information in different combinations in order to realize functional synergies ( Singer, 2001 ; Baars et al, 2013 ; Greff et al, 2020 ; Safron et al, 2021b ). While an FNN could theoretically achieve adaptive binding with respect to a single state estimate, this would divorce the integrating processes from its environmental couplings and historicity as an iterative process of generating inferences regarding the contents of experience, comparing these predictions against sense data, and then updating these prior expectations into posterior beliefs as priors for subsequent rounds of predictive modeling.…”

AIXI, FEP-AI, and integrated world models: Towards a unified understanding of intelligence and consciousness

“…The flexible small-worldness of these "highway nets"-and their entailed recurrent processing-has potentially strong functional correspondences with aspects of brains thought to enable workspace architectures via connectomic "rich clubs," which may be capable of supporting "dynamic cores" of re-entrant signaling, so allowing for synergistic processing via balanced integrated and segregated processing [82][83][84][85]. Notably, such balanced integration and segregation via small-worldness is also a property of systems capable of both maximizing integrated information and supporting self-organized critical dynamics [1,86], the one regime in which (generalized) evolution is possible [20,21,[87][88][89][90]. As described elsewhere with respect to the critique of Aaronson's critique of Integrated Information Theory (IIT) [91], small-world networks can also be used for error-correction via expander graphs (or low-density parity checking codes), which enable systems to approach the Shannon limit with respect to handling noisy, lossilyand irreversibly [92]-compressed data.…”