The evolution of brain architectures for predictive coding and active inference

Abstract

This article considers the evolution of brain architectures for predictive processing. We argue that brain mechanisms for predictive perception and action are not late evolutionary additions of advanced creatures like us. Rather, they emerged gradually from simpler predictive loops (e.g. autonomic and motor reflexes) that were a legacy from our earlier evolutionary ancestors—and were key to solving their fundamental problems of adaptive regulation. We characterize simpler-to-more-complex brains formally, in terms of generative models that include predictive loops of increasing hierarchical breadth and depth. These may start from a simple homeostatic motif and be elaborated during evolution in four main ways: these include the multimodal expansion of predictive control into an allostatic loop; its duplication to form multiple sensorimotor loops that expand an animal's behavioural repertoire; and the gradual endowment of generative models with hierarchical depth (to deal with aspects of the world that unfold at different spatial scales) and temporal depth (to select plans in a future-oriented manner). In turn, these elaborations underwrite the solution to biological regulation problems faced by increasingly sophisticated animals. Our proposal aligns neuroscientific theorising—about predictive processing—with evolutionary and comparative data on brain architectures in different animal species. This article is part of the theme issue ‘Systems neuroscience through the lens of evolutionary theory’.

The Human Affectome

“…Immediate to distal concerns provide a gradient or scale pertaining to things in the environment one can have affective states about, which can be interpreted as the complexity of actions needed to address the allostatic concern, the timescale necessary to achieve homeostatic impact, or its abstractness and semantic meaning (McEwen and Seeman, 1999;Pezzulo, 2012;Pezzulo et al, 2015). In formal and computational terms, this can be construed as hierarchical depth .…”

Enactive-Dynamic Social Cognition and Active Inference

“…An adaptive system's action to maximise their adjustment to their local environment can be translated into the constructs of minimisation of uncertainty, entropy, or surprisal. As a 'first principles' approach to understanding behaviour and the brain, it is framed in terms of a single imperative to minimise free energy given a generative model (Pezzulo et al, 2022). The Free Energy Principle (FEP) states that natural systems remain in non-equilibrium steady states by restricting themselves to a limited number of states.…”

On religious practices as multiscale active inference: Certainties emerging from recurrent interactions within and across individuals and groups

“…Contrary to what was supposed by the Modularity of the Mind, evidence indicates top-down modulation over lower levels of cognition. This evidence has led to predictive coding becoming the basis for a variety of frameworks to understand neural activity (Rao and Ballard, 1999;Spratling, 2016;Smith, Badcock and Friston, 2021;Kramer et al 2022;Pezzulo, Parr and Friston, 2022) and it has inspired predictive accounts of cognition, which extend predictive coding to cognitive activity to the agent scale, such as thinking, dreaming or perceiving (Clark, 2015;Lupyan and Clark, 2015;Van de Cruys, 2017;Fabry, 2018;Wilkinson et al 2019;Walsh et al 2020;Hohwy, 2020). Andersen (2019) specifically applies predictive coding to the detection of supernatural agency.…”