From allostatic agents to counterfactual cognisers: active inference, biological regulation, and the origins of cognition

Abstract

What is the function of cognition? On one influential account, cognition evolved to co-ordinate behaviour with environmental change or complexity (Godfrey-Smith in Complexity and the function of mind in nature, Cambridge Studies in Philosophy and Biology, Cambridge University Press, Cambridge, 1996). Liberal interpretations of this view ascribe cognition to an extraordinarily broad set of biological systems-even bacteria, which modulate their activity in response to salient external cues, would seem to qualify as cognitive agents. However, equating cognition with adaptive flexibility per se glosses over important distinctions in the way biological organisms deal with environmental complexity. Drawing on contemporary advances in theoretical biology and computational neuroscience, we cash these distinctions out in terms of different kinds of generative models, and the representational and uncertainty-resolving capacities they afford. This analysis leads us to propose a formal criterion for delineating cognition from other, more pervasive forms of adaptive plasticity. On this view, biological cognition is rooted in a particular kind of functional organisation; namely, that which enables the agent to detach from the present and engage in counterfactual (active) inference.

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; Toward Solving the Hard Problem and Characterizing Agentic Causation

“…This deep organization allows these systems to model not only transient events at lower levels, but also their organization into more temporally extended sequences at higher levels (Hawkins and Blakeslee, 2004 ; Baldassano et al, 2017 ; Friston et al, 2017c ). Further, deep internal dynamics create a potential for functional decoupling between modeling and the unfolding of particular sensorimotor engagements (Tani, 2016 ; Sormaz et al, 2018 ; Corcoran et al, 2020 ), thus enabling counterfactual simulations (Kanai et al, 2019 ) with temporal “thickness”/“depth” (Humphrey, 2017 ; Friston, 2018 ), which when conscious enable imagination and explicit planning. These capacities afford the possibility of constructing rich causal world models (Hassabis and Maguire, 2009 ; Buchsbaum et al, 2012 ; Pearl and Mackenzie, 2018 ; MacKay, 2019 ), and as discussed below, preconditions for coherent conscious experience.…”

“…Conscious access—and possibly phenomenal consciousness—likely requires generative processes capable of counterfactual modeling (Friston, 2018 ; Pearl and Mackenzie, 2018 ; Kanai et al, 2019 ; Corcoran et al, 2020 ) with respect to selfhood and self-generated actions.…”

“…The DMN and its imaginative capacities (Beaty et al, 2014 , 2015 , 2018 ; Hassabis et al, 2014 ) may instantiate this kind of (informational) creative dynamo, constituting sources of strongly internally coherent predictions, thus being capable of temporarily absorbing and then releasing free energy via the shaping of perception and driving of action. The network properties of the DMN are ideally suited to serve these functions, having both high centrality—and so high potential for integrating information and exerting control (Kenett et al, 2018 )—while also being located distally from primary modalities, and so being capable of supporting dynamics that are more decoupled from immediate sensorimotor engagements (Sormaz et al, 2018 ; Corcoran et al, 2020 ). Further, the DMN is likely to support some of the most stable inferences available to embodied–embedded persons, with major nodes allowing for egocentric perspective—i.e., providing a subjective point of view in generating world models with spatial, temporal, and causal coherence—integrated memory, and even the foundations of selfhood (Dennett, 1992 ; Hassabis and Maguire, 2009 ; Northoff, 2012 ; Brewer et al, 2013 ; Davey and Harrison, 2018 ).…”

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

“…In suggesting that the ability to deploy detached operations on internal variables is an evolutionary function, we are suggesting that the fact that such representational capacities are found today in some organisms can be explained (partially) by the selection pressures that the capacity of detached representation allowed such organisms to successfully respond to. Any detailed speculative account however as to what the presence of detached representational capacities in certain organisms suggests about the kinds of environmental problem spaces that they have responded to in the past falls beyond the scope of this paper (but see for example Millikan, 1989Millikan, , 2004Sterelny, 2003;Schulz, 2018;Corcoran et al, 2020).…”

“…If this is the case, then not only can a process that does not start out as a representation become a representation at short timescales (think use/re-use accounts), but we should expect as much given some kind of evolutionary continuity between putatively non-representational processes and those detached processes that have evolved from them. In other words, full-fledged decoupled representational capacities might "shade-off" (Godfrey-Smith, 1996) into other cognitive or minimal cognitive processes, the dynamics of which are increasingly 21 Allostatic control refers to anticipatory behaviour that by enslaving wide-range homeostatic variables allows systems to avoid deviation from essential homeostatic setpoints prior to the onset of deviation (for various and yet somewhat related views of allostatic control see Stephan et al (2016), Corcoran et al (2020), Kiverstein and Sims (2021). more coupled to the environment through action and sensory feedback.…”