According to a growing trend in theoretical neuroscience, the human perceptual system is akin to a Bayesian machine. The aim of this article is to clearly articulate the claims that perception can be considered Bayesian inference and that the brain can be considered a Bayesian machine, some of the epistemological challenges to these claims; and some of the implications of these claims. We address two questions: (i) How are Bayesian models used in theoretical neuroscience? (ii) From the use of Bayesian (...) models in theoretical neuroscience, have we learned or can we hope to learn that perception is Bayesian inference or that the brain is a Bayesian machine? From actual practice in theoretical neuroscience, we argue for three claims. First, currently Bayesian models do not provide mechanistic explanations; instead they are useful devices for predicting and systematizing observational statements about people's performances in a variety of perceptual tasks. That is, currently we should have an instrumentalist attitude towards Bayesian models in neuroscience. Second, the inference typically drawn from Bayesian behavioural performance in a variety of perceptual tasks to underlying Bayesian mechanisms should be understood within the three-level framework laid out by David Marr ( [1982] ). Third, we can hope to learn that perception is Bayesian inference or that the brain is a Bayesian machine to the extent that Bayesian models will prove successful in yielding secure and informative predictions of both subjects' perceptual performance and features of the underlying neural mechanisms. (shrink)

The success of a piece of behaviour is often explained by its being caused by a true representation (similarly, failure falsity). In some simple organisms, success is just survival and reproduction. Scientists explain why a piece of behaviour helped the organism to survive and reproduce by adverting to the behaviour’s having been caused by a true representation. That usage should, if possible, be vindicated by an adequate naturalistic theory of content. Teleosemantics cannot do so, when it is applied to simple (...) representing systems (Godfrey-Smith 1996). Here it is argued that the teleosemantic approach to content should therefore be modified, not abandoned, at least for simple representing systems. The new ‘infotel-semantics’ adds an input condition to the output condition offered by teleosemantics, recognising that it is constitutive of content in a simple representing system that the tokening of a representation should correlate probabilistically with the obtaining of its specific evolutionary success condition. (shrink)

Brains, it has recently been argued, are essentially prediction machines. They are bundles of cells that support perception and action by constantly attempting to match incoming sensory inputs with top-down expectations or predictions. This is achieved using a hierarchical generative model that aims to minimize prediction error within a bidirectional cascade of cortical processing. Such accounts offer a unifying model of perception and action, illuminate the functional role of attention, and may neatly capture the special contribution of cortical processing to (...) adaptive success. This target article critically examines this approach, concluding that it offers the best clue yet to the shape of a unified science of mind and action. Sections 1 and 2 lay out the key elements and implications of the approach. Section 3 explores a variety of pitfalls and challenges, spanning the evidential, the methodological, and the more properly conceptual. The paper ends (sections 4 and 5) by asking how such approaches might impact our more general vision of mind, experience, and agency. (shrink)

The brain is often taken to be a paradigmatic example of a signaling system with semantic and representational properties, in which neurons are senders and receivers of information carried in action potentials. A closer look at this picture shows that it is not as appealing as it might initially seem in explaining the function of the brain. Working from several sender-receiver models within the teleosemantic framework, I will first argue that two requirements must be met for a system to support (...) genuine semantic information: 1. The receiver must be competent —that is, it must be able to extract rewards from its environment on the basis of the signals that it receives. 2. The receiver must have some flexibility of response relative to the signal received. In the second part of the paper, this initial framework will be applied to neural processes, pointing to the surprising conclusion that signaling at the single-neuron level is only weakly semantic at best. Contrary to received views, neurons will have little or no access to semantic information (though their patterns of activity may carry plenty of quantitative, correlational information) about the world outside the organism. Genuine representation of the world requires an organism - level receiver of semantic information, to which any particular set of neurons makes only a small contribution. (shrink)

In this paper, we set out to develop a theoretical and conceptual framework for the new field of Radical Embodied Cognitive Neuroscience. This framework should be able to integrate insights from several relevant disciplines: theory on embodied cognition, ecological psychology, phenomenology, dynamical systems theory, and neurodynamics. We suggest that the main task of Radical Embodied Cognitive Neuroscience is to investigate the phenomenon of skilled intentionality from the perspective of the self-organization of the brain-body-environment system, while doing justice to the phenomenology (...) of skilled action. In previous work, we have characterized skilled intentionality as the organism's tendency toward an optimal grip on multiple relevant affordances simultaneously. Affordances are possibilities for action provided by the environment. In the first part of this paper, we introduce the notion of skilled intentionality and the phenomenon of responsiveness to a field of relevant affordances. Second, we use Friston's work on neurodynamics, but embed a very minimal version of his Free Energy Principle in the ecological niche of the animal. Thus amended, this principle is helpful for understanding the embeddedness of neurodynamics within the dynamics of the system “brain-body-landscape of affordances.” Next, we show how we can use this adjusted principle to understand the neurodynamics of selective openness to the environment: interacting action-readiness patterns at multiple timescales contribute to the organism's selective openness to relevant affordances. In the final part of the paper, we emphasize the important role of metastable dynamics in both the brain and the brain-body-environment system for adequate affordance-responsiveness. We exemplify our integrative approach by presenting research on the impact of Deep Brain Stimulation on affordance responsiveness of OCD patients. (shrink)

Skyrms, building on the work of Dretske, has recently developed a novel information-theoretic account of propositional content in simple signalling systems. Information-theoretic accounts of content traditionally struggle to accommodate the possibility of misrepresentation, and I show that Skyrms’s account is no exception. I proceed to argue, however, that a modified version of Skyrms’s account can overcome this problem. On my proposed account, the propositional content of a signal is determined not by the information that it actually carries, but by the (...) information that it would carry at the nearest separating equilibrium of the underlying evolutionary dynamics. I show that this amended account yields reasonable ascriptions of false propositional content in a well-known formal model of the evolution of communication , and close with a discussion of the serious but perhaps not insuperable difficulties we face in applying the account to examples of signalling in the real world. (shrink)

Aboutness has been studied from any number of angles. Brentano made it the defining feature of the mental. Phenomenologists try to pin down the aboutness-features of particular mental states. Materialists sometimes claim to have grounded aboutness in natural regularities. Attempts have even been made, in library science and information theory, to operationalize the notion. But it has played no real role in philosophical semantics. This is surprising; sentences have aboutness-properties if anything does. Aboutness is the first book to examine through (...) a philosophical lens the role of subject matter in meaning. A long-standing tradition sees meaning as truth-conditions, to be specified by listing the scenarios in which a sentence is true. Nothing is said about the principle of selection--about what in a scenario gets it onto the list. Subject matter is the missing link here. A sentence is true because of how matters stand where its subject matter is concerned. Stephen Yablo maintains that this is not just a feature of subject matter, but its essence. One indicates what a sentence is about by mapping out logical space according to its changing ways of being true or false. The notion of content that results--directed content--is brought to bear on a range of philosophical topics, including ontology, verisimilitude, knowledge, loose talk, assertive content, and philosophical methodology. Written by one of today's leading philosophers, Aboutness represents a major advance in semantics and the philosophy of language. (shrink)

Acknowledgements Andrea Scarantino, Nicholas Shea, Mark Sprevak, and three anonymous referees provided incisive and constructive comments, for which I am very grateful. In 2012, earlier versions of this paper were delivered in Edinburgh, at the Joint Session in Stirling, and at a workshop on natural information in Aberdeen. I thank participants for their feedback.

This chapter focuses on what’s novel in the perspective that the prediction error minimization (PEM) framework affords on the cognitive-scientific project of explaining intelligence by appeal to internal representations. It shows how truth-conditional and resemblance-based approaches to representation in generative models may be integrated. The PEM framework in cognitive science is an approach to cognition and perception centered on a simple idea: organisms represent the world by constantly predicting their own internal states. PEM theories often stress the hierarchical structure of (...) the generative models they posit. The novel explanatory power of the PEM account derives largely from the way in which pairs of generative and recognition models interact. “Predictive coding” refers to an encoding strategy in which predicted portions of an input signal are subtracted from the actual signal received, so that only the difference between the two is passed as output to the next stage of information processing. (shrink)

For this second edition, the editors have returned to Austin's original lecture notes, amending the printed text where it seemed necessary.

This is a welcome reprint of a book that continues to grow in importance.

Evaluativism is best thought of as a way of enriching a perceptual view of pain to account for pain’s unpleasantness or painfulness. Once it was common for philosophers to contrast pains with perceptual experiences (McGinn 1982; Rorty 1980). It was thought that perceptual experiences were intentional (or content-bearing, or about something), whereas pains were representationally blank. But today many of us reject this contrast. For us, your having a pain in your toe is a matter not of your sensing “pain-ly” (...) or encountering a sense-datum, but of your having an interoceptive experience representing (accurately or inaccurately) that your toe is in a particular experience-independent condition, such as undergoing a certain “disturbance” or being damaged or in danger (Armstrong 1962; Tye 1995). But even if such representational content makes an experience a pain, a further ingredient seems required to make the pain unpleasant. According to evaluativism, the further ingredient is the experience’s possession of evaluative content: its representing the bodily condition as bad for the subject. In this chapter, I elaborate evaluativism, locate it among alternatives, and explain its attractions and challenges. (shrink)

The mathematical theory of communication.

An extended argument that cognitive phenomena—perceiving, imagining, remembering—can be best explained in terms of an interface between contentless and content-involving forms of cognition. -/- Evolving Enactivism argues that cognitive phenomena—perceiving, imagining, remembering—can be best explained in terms of an interface between contentless and content-involving forms of cognition. Building on their earlier book Radicalizing Enactivism, which proposes that there can be forms of cognition without content, Daniel Hutto and Erik Myin demonstrate the unique explanatory advantages of recognizing that only some forms (...) of cognition have content while others—the most elementary ones—do not. They offer an account of the mind in duplex terms, proposing a complex vision of mentality in which these basic contentless forms of cognition interact with content-involving ones. -/- Hutto and Myin argue that the most basic forms of cognition do not, contrary to a currently popular account of cognition, involve picking up and processing information that is then used, reused, stored, and represented in the brain. Rather, basic cognition is contentless—fundamentally interactive, dynamic, and relational. In advancing the case for a radically enactive account of cognition, Hutto and Myin propose crucial adjustments to our concept of cognition and offer theoretical support for their revolutionary rethinking, emphasizing its capacity to explain basic minds in naturalistic terms. They demonstrate the explanatory power of the duplex vision of cognition, showing how it offers powerful means for understanding quintessential cognitive phenomena without introducing scientifically intractable mysteries into the mix. (shrink)

Clark has recently suggested that predictive processing advances a theory of neural function with the resources to put an ecumenical end to the “representation wars” of recent cognitive science. In this paper I defend and develop this suggestion. First, I broaden the representation wars to include three foundational challenges to representational cognitive science. Second, I articulate three features of predictive processing’s account of internal representation that distinguish it from more orthodox representationalist frameworks. Specifically, I argue that it posits a resemblance-based (...) representational architecture with organism-relative contents that functions in the service of pragmatic success, not veridical representation. Finally, I argue that internal representation so understood is either impervious to the three anti-representationalist challenges I outline or can actively embrace them. (shrink)

A naturalistic scheme of primitive conceptual representations is proposed using the statistical measure of mutual information. It is argued that a concept represents, not the class of objects that caused its tokening, but the class of objects that is most likely to have caused it (had it been tokened), as specified by the statistical measure of mutual information. This solves the problem of misrepresentation which plagues causal accounts, by taking the representation relation to be determined via ordinal relationships between conditional (...) probabilities. The scheme can deal with statistical biases and does not rely on arbitrary criteria. Implications for the theory of meaning and semantic content are addressed. (shrink)

A naturalistic scheme of primitive conceptual representations is proposed using the statistical measure of mutual information. It is argued that a concept represents, not the class of objects that caused its tokening, but the class of objects that is most likely to have caused it (had it been tokened), as specified by the statistical measure of mutual information. This solves the problem of misrepresentation which plagues causal accounts, by taking the representation relation to be determined via ordinal relationships between conditional (...) probabilities. The scheme can deal with statistical biases and does not rely on arbitrary criteria. Implications for the theory of meaning and semantic content are addressed. (shrink)

What is the relationship between information and representation? Dating back at least to Dretske (1981), an influential answer has been that information is a rung on a ladder that gets one to representation. Representation is information, or representation is information plus some other ingredient. In this paper, I argue that this approach oversimplifies the relationship between information and representation. If one takes current probabilistic models of cognition seriously, information is connected to representation in a new way. It enters as a (...) property of the represented content as well as a property of the vehicles that carry that content. This offers a new, conceptually and logically distinct way in which information and representation are intertwined in cognition. (shrink)

Both the quantity of information and the informational content of a signal are defined in the context of signaling games. Informational content is a generalization of standard philosophical notions of propositional content. It is shown how signals that initially carry no information may spontaneously acquire informational content by evolutionary or learning dynamics. It is shown how information can flow through signaling chains or signaling networks.

1. Introduction 2. Reward-Guided Decision Making 3. Content in the Model 4. How to Deflate a Metarepresentational Reading Proust and Carruthers on metacognitive feelings 5. A Deflationary Treatment of RPEs? 5.1 Dispensing with prediction errors 5.2 What is use of the RPE focused on? 5.3 Alternative explanations—worldly correlates 5.4 Contrast cases 6. Conclusion Appendix: Temporal Difference Learning Algorithms.

Recent work combining cognitive neuroscience with computational modelling suggests that distributed patterns of neural firing may represent probability distributions. This paper asks: what makes it the case that distributed patterns of firing, as well as carrying information about (correlating with) probability distributions over worldly parameters, represent such distributions? In examples of probabilistic population coding, it is the way information is used in downstream processing so as to lead to successful behaviour. In these cases content depends on factors beyond bare information, (...) contra Brian Skyrms’ (2010) view that representational content can be fully characterised in information-theoretic terms. (shrink)

Our understanding of communication and its evolution has advanced significantly through the study of simple models involving interacting senders and receivers of signals. Many theorists have thought that the resources of mathematical information theory are all that are needed to capture the meaning or content that is being communicated in these systems. However, the way theorists routinely talk about the models implicitly draws on a conception of content that is richer than bare informational content, especially in contexts where false content (...) is important. This article shows that this concept can be made precise by defining a notion of functional content that captures the degree to which different states of the world are involved in stabilizing senders’ and receivers’ use of a signal at equilibrium. A series of case studies is used to contrast functional content with informational content, and to illustrate the explanatory role and limitations of this definition of functional content. _1_ Introduction _2_ Modelling Framework _3_ Two Kinds of Content _3.1_ Informational content _3.2_ Functional content _4_ Cases _4.1_ Case 1: Simplest case _4.2_ Case 2: Partial pooling _4.3_ Case 3: Bottleneck _4.4_ Case 4: Partial common interest _4.5_ Case 5: Deception _4.6_ Case 6: A further problem arising from divergent interests _5_ Discussion Appendix. (shrink)

Abstract: According to the Veridicality Thesis, information requires truth. On this view, smoke carries information about there being a fire only if there is a fire, the proposition that the earth has two moons carries information about the earth having two moons only if the earth has two moons, and so on. We reject this Veridicality Thesis. We argue that the main notions of information used in cognitive science and computer science allow A to have information about the obtaining of (...) p even when p is false. (shrink)

By virtue of what do alarm calls and facial expressions carry natural information? The answer I defend in this paper is that they carry natural information by virtue of changing the probabilities of various states of affairs, relative to background data. The Probabilistic Difference Maker Theory of natural information that I introduce here is inspired by Dretske's [1981] seminal analysis of natural information, but parts ways with it by eschewing the requirements that information transmission must be nomically underwritten, mind-independent, and (...) knowledge-yielding. PDMT includes both a qualitative account of information transmission and a measure of natural information in keeping with the basic principles of Shannon's communication theory and Bayesian confirmation theory. It also includes a new account of the informational content of a signal, understood as the combination of the incremental and overall support that the signal provides for all states of affairs at the source. Finally, I compare and.. (shrink)

A critique of the purely semantic, minimalist notion of 'what is said'.

This article attempts to clarify the commitments of a predictive coding approach to perception. After summarizing predictive coding theory, the article addresses two questions. Is a predictive coding perceptual system also a Bayesian system? Is it a Kantian system? The article shows that the answer to these questions is negative.

A list of groceries, Professor Anscombe once suggested, might be used as a shopping list, telling what to buy, or it might be used as an inventory list, telling what has been bought (Anscombe 1957). If used as a shopping list, the world is supposed to conform to the representation: if the list does not match what is in the grocery bag, it is what is in the bag that is at fault. But if used as an inventory list, the (...) representation is supposed to conform to the world: if the list does not match what is in the bag, it is the list that is at fault. The first kind of representation, where the world is supposed to conform to the list, can be called "directive"; it represents or directs what is to be done. The second, where the list is supposed to conform to the world, can be called "descriptive"; it represents or describes what is the case. I wish to propose that there exist representations that face both these ways at once. With apologies to Dr. Doolittle, I call them pushmi-pullyu representations or PPRs. (shrink)

Information is widely perceived as essential to the study of communication and representation; still, theorists working on these topics often take themselves not to be centrally concerned with "Shannon information", as it is often put, but with some other, sometimes called "semantic" or "nonnatural",kind of information. This perception is wrong. Shannon's theory of information is the only one we need. -/- I intend to make good on this last assertion by canvassing a fully (Shannon) informational answer to the metasemantic question (...) of what makes something a representation, for a certain important family of cases. This answer and the accompanying theory, which represents a significant departure from the broadly Dretskean philosophical mainstream, will show how a number of threads in the literature on naturalistic metasemantics, aimed at describing the purportedly non-informational ingredients in representation, actually belong in the same coherent, purely information-theoretic picture. (shrink)

Recent work on signaling has mostly focused on communication between organisms. The Lewis–Skyrms framework should be equally applicable to intra-organismic signaling. We present a Lewis–Skyrms signaling-game model of painful signaling, and use it to argue that the content of pain is predominantly imperative. We address several objections to the account, concluding that our model gives a productive framework within which to consider internal signaling.

Despite of its formal precision and its great many applications, Shannon’s theory still offers an active terrain of debate when the interpretation of its main concepts is the task at issue. In this article we try to analyze certain points that still remain obscure or matter of discussion, and whose elucidation contribute to the assessment of the different interpretative proposals about the concept of information. In particular, we argue for a pluralist position, according to which the different views about information (...) are no longer rival, but different interpretations of a single formal concept. (shrink)

_ Convention_ was immediately recognized as a major contribution to the subject and its significance has remained undiminished since its first publication in 1969. Lewis analyzes social conventions as regularities in the resolution of recurring coordination problems-situations characterized by interdependent decision processes in which common interests are at stake. Conventions are contrasted with other kinds of regularity, and conventions governing systems of communication are given special attention.

Shannon information is commonly assumed to be the wrong way in which to conceive of information in most biological contexts. Since the theory deals only in correlations between systems, the argument goes, it can apply to any and all causal interactions that affect a biological outcome. Since informational language is generally confined to only certain kinds of biological process, such as gene expression and hormone signalling, Shannon information is thought to be unable to account for this restriction. It is often (...) concluded that a richer, teleosemantic sense of information is needed. I argue against this view, and show that a coherent and sufficiently restrictive theory of biological information can be constructed with Shannon information at its core. This can be done by paying due attention some crucial distinctions: between information quantity and its fitness value, and between carrying information and having the function of doing so. From this I construct an account of how informational functions arise, and show that the “subject matter” of these functions can easily be seen as the natural information dealt with by Shannon’s theory. (shrink)

In this paper, we consider how certain longstanding philosophical questions about mental representation may be answered on the assumption that cognitive and perceptual systems implement hierarchical generative models, such as those discussed within the prediction error minimization framework. We build on existing treatments of representation via structural resemblance, such as those in Gładziejewski :559–582, 2016) and Gładziejewski and Miłkowski, to argue for a representationalist interpretation of the PEM framework. We further motivate the proposed approach to content by arguing that it (...) is consistent with approaches implicit in theories of unsupervised learning in neural networks. In the course of this discussion, we argue that the structural representation proposal, properly understood, has more in common with functional-role than with causal/informational or teleosemantic theories. In the remainder of the paper, we describe the PEM framework for approximate Bayesian inference in some detail, and discuss how structural representations might arise within the proposed Bayesian hierarchies. After explicating the notion of variational inference, we define a subjectively accessible measure of misrepresentation for hierarchical Bayesian networks by appeal to the Kullbach–Leibler divergence between posterior generative and approximate recognition densities, and discuss a related measure of objective misrepresentation in terms of correspondence with the facts. (shrink)

The lore is that standard information theory provides an analysis of information quantity, but not of information content. I argue this lore is incorrect, and there is an adequate informational semantics latent in standard theory. The roots of this notion of content can be traced to the secret parallel development of an information theory equivalent to Shannon’s by Turing at Bletchley Park, and it has been suggested independently in recent work by Skyrms and Bullinaria and Levy. This paper explicitly articulates (...) the semantics latent in information theory and defends it as an adequate theory of information content, or natural meaning. I argue that this theory suggests a new perspective on the classic misrepresentation worry for correlation-based semantics. (shrink)

Predictive Processing accounts of Cognition, PPC, promise to forge productive alliances that will unite approaches that are otherwise at odds. Can it? This paper argues that it can’t—or at least not so long as it sticks with the cognitivist rendering that Clark and others favor. In making this case the argument of this paper unfolds as follows: Sect. 1 describes the basics of PPC—its attachment to the idea that we perceive the world by guessing the world. It then details the (...) reasons why so many find cognitivist interpretations to be inevitable. Section 2 examines how prominent proponents of cognitivist PPC have proposed dealing with a fundamental problem that troubles their accounts—the question of how the brain is able to get into the great guessing game in the first place. It is argued that on close inspection Clark’s solution, which he calls bootstrap heaven is—once we take a realistic look at the situation of the brain—in fact bootstrap hell. Section 3 argues that it is possible to avoid dwelling in bootstrap hell if one adopts a radically enactive take on PPC. A brief sketch of what this might look like is provided. (shrink)

Drawing on models of communication due to Lewis and Skyrms, I contrast sender-receiver systems as they appear within and between organisms, and as they function in the bridging of space and time. Within the organism, memory can be seen as the sending of messages over time, communication between stages as opposed to spatial parts. Psychological memory and genetic memory are compared with respect to their relations to a sender-receiver model. Some puzzles about “genetic information” can be resolved by seeing the (...) genome as a cell-level memory with no sender. (shrink)

This paper centers around the notion that internal, mental representations are grounded in structural similarity, i.e., that they are so-called S-representations. We show how S-representations may be causally relevant and argue that they are distinct from mere detectors. First, using the neomechanist theory of explanation and the interventionist account of causal relevance, we provide a precise interpretation of the claim that in S-representations, structural similarity serves as a “fuel of success”, i.e., a relation that is exploitable for the representation using (...) system. Then, we discuss crucial differences between S-representations and indicators or detectors, showing that—contrary to claims made in the literature—there is an important theoretical distinction to be drawn between the two. (shrink)

According to the predictive coding theory of cognition , brains are predictive machines that use perception and action to minimize prediction error, i.e. the discrepancy between bottom–up, externally-generated sensory signals and top–down, internally-generated sensory predictions. Many consider PCT to have an explanatory scope that is unparalleled in contemporary cognitive science and see in it a framework that could potentially provide us with a unified account of cognition. It is also commonly assumed that PCT is a representational theory of sorts, in (...) the sense that it postulates that our cognitive contact with the world is mediated by internal representations. However, the exact sense in which PCT is representational remains unclear; neither is it clear that it deserves such status—that is, whether it really invokes structures that are truly and nontrivially representational in nature. In the present article, I argue that the representational pretensions of PCT are completely justified. This is because the theory postulates cognitive structures—namely action-guiding, detachable, structural models that afford representational error detection—that play genuinely representational functions within the cognitive system. (shrink)

If one formulates Helmholtz's ideas about perception in terms of modern-day theories one arrives at a model of perceptual inference and learning that can explain a remarkable range of neurobiological facts. Using constructs from statistical physics it can be shown that the problems of inferring what cause our sensory inputs and learning causal regularities in the sensorium can be resolved using exactly the same principles. Furthermore, inference and learning can proceed in a biologically plausible fashion. The ensuing scheme rests on (...) Empirical Bayes and hierarchical models of how sensory information is generated. The use of hierarchical models enables the brain to construct prior expectations in a dynamic and context-sensitive fashion. This scheme provides a principled way to understand many aspects of the brain's organisation and responses. In this paper, we suggest that these perceptual processes are just one emergent property of systems that conform to a free-energy principle. The free-energy considered here represents a bound on the surprise inherent in any exchange with the environment, under expectations encoded by its state or configuration. A system can minimise free-energy by changing its configuration to change the way it samples the environment, or to change its expectations. These changes correspond to action and perception, respectively, and lead to an adaptive exchange with the environment that is characteristic of biological systems. This treatment implies that the system's state and structure encode an implicit and probabilistic model of the environment. We will look at models entailed by the brain and how minimisation of free-energy can explain its dynamics and structure. (shrink)

If one formulates Helmholtz’s ideas about perception in terms of modern-day theories one arrives at a model of perceptual inference and learning that can explain a remarkable range of neurobiological facts. Using constructs from statistical physics it can be shown that the problems of inferring what cause our sensory inputs and learning causal regularities in the sensorium can be resolved using exactly the same principles. Furthermore, inference and learning can proceed in a biologically plausible fashion. The ensuing scheme rests on (...) Empirical Bayes and hierarchical models of how sensory information is generated. The use of hierarchical models enables the brain to construct prior expectations in a dynamic and context-sensitive fashion. This scheme provides a principled way to understand many aspects of the brain’s organisation and responses. In this paper, we suggest that these perceptual processes are just one emergent property of systems that conform to a free-energy principle. The free-energy considered here represents a bound on the surprise inherent in any exchange with the environment, under expectations encoded by its state or configuration. A system can minimise free-energy by changing its configuration to change the way it samples the environment, or to change its expectations. These changes correspond to action and perception, respectively, and lead to an adaptive exchange with the environment that is characteristic of biological systems. This treatment implies that the system’s state and structure encode an implicit and probabilistic model of the environment. We will look at models entailed by the brain and how minimisation of free-energy can explain its dynamics and structure. (shrink)