Prefrontal Computation as Active Inference

Abstract

The prefrontal cortex is vital for a range of cognitive processes, including working memory, attention, and decision-making. Notably, its absence impairs the performance of tasks requiring the maintenance of information through a delay period. In this paper, we formulate a rodent task—which requires maintenance of delay-period activity—as a Markov decision process and treat optimal task performance as an (active) inference problem. We simulate the behavior of a Bayes optimal mouse presented with 1 of 2 cues that instructs the selection of concurrent visual and auditory targets on a trial-by-trial basis. Formulating inference as message passing, we reproduce features of neuronal coupling within and between prefrontal regions engaged by this task. We focus on the micro-circuitry that underwrites delay-period activity and relate it to functional specialization within the prefrontal cortex in primates. Finally, we simulate the electrophysiological correlates of inference and demonstrate the consequences of lesions to each part of our in silico prefrontal cortex. In brief, this formulation suggests that recurrent excitatory connections—which support persistent neuronal activity—encode beliefs about transition probabilities over time. We argue that attentional modulation can be understood as the contextualization of sensory input by these persistent beliefs.

Neuromodulatory Control and Language Recovery in Bilingual Aphasia: An Active Inference Approach

“…Variational message passing can be formulated as a gradient descent on variational free energy, using a mean-field approximation [ 33 , 34 ]. This has been shown to provide a plausible account of neuronal dynamics [ 14 , 35 , 36 ].…”

“…This perspective allows us to associate in-silico lesions with anatomical lesions. When formulating model inversion in terms of neuronal message passing, the likelihood mapping takes the form of extrinsic connections (i.e., between different neuronal populations encoding posterior expectations) while the transition mapping takes the form of intrinsic connections (i.e., connectivity within neuronal populations encoding posterior expectations over time) [ 35 ].…”

Inferring What to Do (And What Not to)

“…Framed as a generative model [ 79 ], this says that we make use of hidden states comprising the task context and choice to predict outcomes that include exteroceptive and interoceptive stimuli. Exteroceptive stimuli depend only upon the task context, while the task context and choice conspire to predict the interoceptive input—as the juice is only received following the correct saccade.…”

“…The consequence of this is that lesions to different parts of the prefrontal cortex are accompanied by dissociable behavioural phenotypes. This offers a simple way of thinking about the distinction between medial and lateral prefrontal syndromes [ 79 ]. The formulation here suggests that lateral lesions should impair the ability to infer (and retain) beliefs about the experimental context—thereby precluding value-seeking behaviour, as the association between choices and their anticipated consequences are inappropriately contextualised [ 63 ].…”

Integrated World Modeling Theory (IWMT) Revisited

“…A recurrent anterior pole portion that mediates between affect and action selection via its evolving/bifurcating/non-linear attractor dynamics (Izquierdo-Torres and Bührmann, 2008;Izquierdo et al, 2013;Tani, 2016). (1) and (2) could be frontal analogues to the "what" and "where" pathways for vision (Pezzulo et al, 2018;Parr et al, 2019c)-with macroscale connectivity reflecting these functional relationships-except here we are dealing with (1) what-where (via coupling with the hippocampal complex) and (2) how-where (via coupling with the parietal lobes). Taken together (which is how these systems are likely to work under most circumstances), these different parts of frontal cortices could all be understood in a unified sense as implementing policy selection via predictions and precision-weighting.…”