Why Two Worlds Collide: the Savage Mind From the Perspective of a Free Energy Principle Hypothesis

In this paper, we argue for a theoretical separation of the free-energy principle from Helmholtzian accounts of the predictive brain. The free-energy principle is a theoretical framework capturing the imperative for biological self-organization in information-theoretic terms. The free-energy principle has typically been connected with a Bayesian theory of predictive coding, and the latter is often taken to support a Helmholtzian theory of perception as unconscious inference. If our interpretation is right, however, a Helmholtzian view of perception is incompatible with Bayesian predictive coding under the free-energy principle. We argue that the free energy principle and the ecological and enactive approach to mind and life make for a much happier marriage of ideas. We make our argument based on three points. First we argue that the free energy principle applies to the whole animal–environment system, and not only to the brain. Second, we show that active inference, as understood by the free-energy principle, is incompatible with unconscious inference understood as analagous to scientific hypothesis-testing, the main tenet of a Helmholtzian view of perception. Third, we argue that the notion of inference at work in Bayesian predictive coding under the free-energy principle is too weak to support a Helmholtzian theory of perception. Taken together these points imply that the free energy principle is best understood in ecological and enactive terms set out in this paper.

Active inference introduces a theory describing action-perception loops via the minimisation of variational (and expected) free energy or, under simplifying assumptions, (weighted) prediction error. Recently, active inference has been proposed as part of a new and unifying framework in the cognitive sciences: predictive processing. Predictive processing is often associated with traditional computational theories of the mind, strongly relying on internal representations presented in the form of generative models thought to explain different functions of living and cognitive systems. In this work, we introduce an active inference formulation of the Watt centrifugal governor, a system often portrayed as the canonical "anti-representational" metaphor for cognition. We identify a generative model of a steam engine for the governor, and derive a set of equations describing "perception" and "action" processes as a form of prediction error minimisation. In doing so, we firstly challenge the idea of generative models as explicit internal representations for cognitive systems, suggesting that such models serve only as implicit descriptions for an observer. Secondly, we consider current proposals of predictive processing as a theory of cognition, focusing on some of its potential shortcomings and in particular on the idea that virtually any system admits a description in terms of prediction error minimisation, suggesting that this theory may offer limited explanatory power for cognitive systems. Finally, as a silver lining we emphasise the instrumental role this framework can nonetheless play as a mathematical tool for modelling cognitive architectures interpreted in terms of Bayesian (active) inference.

The free energy principle (FEP) portends to provide a unifying principle for the biological and cognitive sciences. It states that for a system to maintain non-equilibrium steady-state with its environment it must minimise its (information-theoretic) free energy. Under the FEP, to minimise free energy is equivalent to engaging in approximate Bayesian inference. According to the FEP, therefore, inference is at the explanatory base of biology and cognition. In this paper, we discuss a specific challenge to this inferential formulation of adaptive self-organisation. We call it the universal ethology challenge: it states that the FEP cannot unify biology and cognition, for life itself (or adaptive self-organisation) does not require inferential routines to select adaptive solutions to environmental pressures (as mandated by the FEP). We show that it is possible to overcome the universal ethology challenge by providing a cautious and exploratory treatment of inference under the FEP. We conclude that there are good reasons for thinking that the FEP can unify biology and cognition under the notion of approximate Bayesian inference, even if further challenges must be addressed to properly draw such a conclusion.

This paper examines, for the first time, the relationship between realization relations and the free energy principle in cognitive neuroscience. I argue, firstly, that the free energy principle has ramifications for the wide versus narrow realization distinction: if the free energy principle is correct, then organismic realizers are insufficient for realizing free energy minimization. I argue, secondly, that the free energy principle has implications for synchronic realization relations, because free energy minimization is realized in dynamical agent-environment couplings embedded at multiple time scales.

In the last decade, the free energy principle (FEP) and active inference (AIF) have achieved many successes connecting conceptual models of learning and cognition to mathematical models of perception and action. This effort is driven by a multidisciplinary interest in understanding aspects of self-organizing complex adaptive systems, including elements of agency. Various reinforcement learning (RL) models performing active inference have been proposed and trained on standard RL tasks using deep neural networks. Recent work has focused on improving such agents’ performance in complex environments by incorporating the latest machine learning techniques. In this paper, we build upon these techniques. Within the constraints imposed by the FEP and AIF, we attempt to model agents in an interpretable way without deep neural networks by introducing Free Energy Projective Simulation (FEPS). Using internal rewards only, FEPS agents build a representation of their partially observable environments with which they interact. Following AIF, the policy to achieve a given task is derived from this world model by minimizing the expected free energy. Leveraging the interpretability of the model, techniques are introduced to deal with long-term goals and reduce prediction errors caused by erroneous hidden state estimation. We test the FEPS model on two RL environments inspired from behavioral biology: a timed response task and a navigation task in a partially observable grid. Our results show that FEPS agents fully resolve the ambiguity of both environments by appropriately contextualizing their observations based on prediction accuracy only. In addition, they infer optimal policies flexibly for any target observation in the environment.

Biological agents can act in ways that express a sensitivity to context-dependent relevance. So far it has proven difficult to engineer this capacity for context-dependent sensitivity to relevance in artificial agents. We give this problem the label the “problem of meaning”. The problem of meaning could be circumvented if artificial intelligence researchers were to design agents based on the assumption of the continuity of life and mind. In this paper, we focus on the proposal made by enactive cognitive scientists to design artificial agents that possess sensorimotor autonomy—stable, self-sustaining patterns of sensorimotor interaction that can ground values, norms and goals necessary for encountering a meaningful environment. More specifically, we consider whether the Free Energy Principle (FEP) can provide formal tools for modeling sensorimotor autonomy. There is currently no consensus on how to understand the relationship between enactive cognitive science and the FEP. However, a number of recent papers have argued that the two frameworks are fundamentally incompatible. Some argue that biological systems exhibit historical path-dependent learning that is absent from systems that minimize free energy. Others have argued that a free energy minimizing system would fail to satisfy a key condition for sensorimotor agency referred to as “interactional asymmetry”. These critics question the claim we defend in this paper that the FEP can be used to formally model autonomy and adaptivity. We will argue it is too soon to conclude that the two frameworks are incompatible. There are undeniable conceptual differences between the two frameworks but in our view each has something important and necessary to offer. The FEP needs enactive cognitive science for the solution it provides to the problem of meaning. Enactive cognitive science needs the FEP to formally model the properties it argues to be constitutive of agency. Our conclusion will be that active inference models based on the FEP provides a way by which scientists can think about how to address the problems of engineering autonomy and adaptivity in artificial agents in formal terms. In the end engaging more closely with this formalism and its further developments will benefit those working within the enactive framework.

The free energy principle says that organisms act to maintain themselves in their expected states and that they achieve this by minimizing their free energy. This corresponds to the brain’s job of minimizing prediction error, selective sampling of sensory data, optimizing expected precisions, and minimizing complexity of internal models. These in turn map on to perception, action, attention, and model selection, respectively. This means that the free energy principle is extremely ambitious: it aims to explain everything about the mind. The principle is bound to be controversial, and hostage to empirical fortune. It may also be thought preposterous: the theory may seem either too ambitious or too trivial to be taken seriously. This chapter introduces the ideas behind the free energy principle and then proceeds to discuss the charge of preposterousness from the perspective of philosophy of science. It is shown that whereas it is ambitious, controversial and needs further evidence in its favour, it is not preposterous. The argument proceeds by appeal to: (i) the notion of inference to the best explanation, (ii) a comparison with the theory of evolution, (iii) the notion of explaining-away, and (iv) a “biofunctionalist” account of Bayesian processing. The heuristic starting point is the simple idea that the brain is just one among our bodily organs, each of which has an overall function. The outcome is not just a defence of the free energy principle against various challenges but also a deeper anchoring of this theory in philosophy of science, yielding an appreciation of the kind of explanation of the mind it offers.

Two striking claims are advanced on behalf of the free energy principle (FEP) in cognitive science and philosophy: (i) that it identifies a condition of the possibility of existence for self-organising systems; and (ii) that it has important implications for our understanding of how the brain works, defining a set of process theories—roughly, theories of the structure and functions of neural mechanisms—consistent with the free energy minimising imperative that it derives as a necessary feature of all self-organising systems. I argue that the conjunction of claims (i) and (ii) rests on a fallacy of equivocation. The FEP can be interpreted in two ways: as a claim about how it is possible to redescribe the existence of self-organising systems (the Descriptive FEP), and as a claim about how such systems maintain their existence (the Explanatory FEP). Although the Descriptive FEP plausibly does identify a condition of the possibility of existence for self-organising systems, it has no important implications for our understanding of how the brain works. Although the Explanatory FEP would have such implications if it were true, it does not identify a condition of the possibility of existence for self-organising systems. I consider various ways of responding to this conclusion, and I explore its implications for the role and importance of the FEP in cognitive science and philosophy.

The issue of the relationship between predictive processing (PP) and the free energy principle (FEP) remains a subject of debate and controversy within the research community. Many researchers have expressed doubts regarding the actual integration of PP with the FEP, questioning whether the FEP can truly contribute significantly to the mechanistic understanding of PP or even undermine such integration altogether. In this paper, I present an alternative perspective. I argue that, from the viewpoint of the constraint-based mechanisms approach, the FEP imposes an important constraint, namely variational free energy, on the mechanistic architecture proposed by PP. According to the constraint-based mechanisms approach, high-level cognitive mechanisms are integral parts of extensive heterarchical networks that govern the physiology and behavior of agents. Consequently, mechanistic explanations of cognitive phenomena should incorporate constraints and flows of free energy as relevant components, given that the implemented constraints operate as long as free energy is available. Within this framework, I contend that the FEP provides a relevant constraint for explaining at least some biological cognitive mechanisms described in terms of Bayesian generative models that minimize prediction errors.

Active inference offers a unified theory of perception, learning, and decision-making at computational and neural levels of description. In this article, we address the worry that active inference may be in tension with the belief–desire–intention (BDI) model within folk psychology because it does not include terms for desires (or other conative constructs) at the mathematical level of description. To resolve this concern, we first provide a brief review of the historical progression from predictive coding to active inference, enabling us to distinguish between active inference formulations of motor control (which need not have desires under folk psychology) and active inference formulations of decision processes (which do have desires within folk psychology). We then show that, despite a superficial tension when viewed at the mathematical level of description, the active inference formalism contains terms that are readily identifiable as encoding both the objects of desire and the strength of desire at the psychological level of description. We demonstrate this with simple simulations of an active inference agent motivated to leave a dark room for different reasons. Despite their consistency, we further show how active inference may increase the granularity of folk-psychological descriptions by highlighting distinctions between drives to seek information versus reward—and how it may also offer more precise, quantitative folk-psychological predictions. Finally, we consider how the implicitly conative components of active inference may have partial analogues (i.e., “as if” desires) in other systems describable by the broader free energy principle to which it conforms.

Free-energy principle inspired visual quality assessment: An overview

Adaptive agents must act in intrinsically uncertain environments with complex latent structure. Here, we elaborate a model of visual foraging—in a hierarchical context—wherein agents infer a higher-order visual pattern (a “scene”) by sequentially sampling ambiguous cues. Inspired by previous models of scene construction—that cast perception and action as consequences of approximate Bayesian inference—we use active inference to simulate decisions of agents categorizing a scene in a hierarchically-structured setting. Under active inference, agents develop probabilistic beliefs about their environment, while actively sampling it to maximize the evidence for their internal generative model. This approximate evidence maximization (i.e., self-evidencing) comprises drives to both maximize rewards and resolve uncertainty about hidden states. This is realized via minimization of a free energy functional of posterior beliefs about both the world as well as the actions used to sample or perturb it, corresponding to perception and action, respectively. We show that active inference, in the context of hierarchical scene construction, gives rise to many empirical evidence accumulation phenomena, such as noise-sensitive reaction times and epistemic saccades. We explain these behaviors in terms of the principled drives that constitute the expected free energy, the key quantity for evaluating policies under active inference. In addition, we report novel behaviors exhibited by these active inference agents that furnish new predictions for research on evidence accumulation and perceptual decision-making. We discuss the implications of this hierarchical active inference scheme for tasks that require planned sequences of information-gathering actions to infer compositional latent structure (such as visual scene construction and sentence comprehension). This work sets the stage for future experiments to investigate active inference in relation to other formulations of evidence accumulation (e.g., drift-diffusion models) in tasks that require planning in uncertain environments with higher-order structure.

Uncertainty and stress: Why it causes diseases and how it is mastered by the brain

Predictive processing theories are increasingly popular in philosophy of mind; such process theories often gain support from the Free Energy Principle (FEP)—a normative principle for adaptive self-organized systems. Yet there is a current and much discussed debate about conflicting philosophical interpretations of FEP, e.g., representational versus non-representational. Here we argue that these different interpretations depend on implicit assumptions about what qualifies (or fails to qualify) as representational. We deploy the Free Energy Principle (FEP) instrumentally to distinguish four main notions of representation, which focus on organizational, structural, content-related and functional aspects, respectively. The various ways that these different aspects matter in arriving at representational or non-representational interpretations of the Free Energy Principle are discussed. We also discuss how the Free Energy Principle may be seen as a unified view where terms that traditionally belong to different ontologies—e.g., notions of model and expectation versus notions of autopoiesis and synchronization—can be harmonized. However, rather than attempting to settle the representationalist versus non-representationalist debate and reveal something about what representations are simpliciter, this paper demonstrates how the Free Energy Principle may be used to reveal something about those partaking in the debate; namely, what our hidden assumptions about what representations are—assumptions that act as sometimes antithetical starting points in this persistent philosophical debate.

The free energy principle (FEP) in the neurosciences stipulates that all viable agents induce and minimize informational free energy in the brain to fit their environmental niche. In this study, we continue our effort to make the FEP a more physically principled formalism by implementing free energy minimization based on the principle of least action. We build a Bayesian mechanics (BM) by casting the formulation reported in the earlier publication (Kim in Neural Comput 30:2616–2659, 2018, https://doi.org/10.1162/neco_a_01115) to considering active inference beyond passive perception. The BM is a neural implementation of variational Bayes under the FEP in continuous time. The resulting BM is provided as an effective Hamilton’s equation of motion and subject to the control signal arising from the brain’s prediction errors at the proprioceptive level. To demonstrate the utility of our approach, we adopt a simple agent-based model and present a concrete numerical illustration of the brain performing recognition dynamics by integrating BM in neural phase space. Furthermore, we recapitulate the major theoretical architectures in the FEP by comparing our approach with the common state-space formulations.