A key component of scientific inquiry, especially inquiry devoted to developing mechanistic explanations, is delineating the phenomenon to be explained. The task of delineating phenomena, however, has not been sufficiently analyzed, even by the new mechanistic philosophers of science. We contend that Marr’s characterization of what he called the computational level provides a valuable resource for understanding what is involved in delineating phenomena. Unfortunately, the distinctive feature of Marr’s computational level, his dual emphasis on both what is computed and why (...) it is computed, has not been appreciated in philosophical discussions of Marr. Accordingly we offer a distinctive account of CL. This then allows us to develop two important points about delineating phenomena. First, the accounts of phenomena that figure in explanatory practice are typically not qualitative but precise, formal or mathematical, representations. Second, delineating phenomena requires consideration of the demands the environment places on the mechanism—identifying, as Marr put it, the basis of the computed function in the world. As valuable as Marr’s account of CL is in characterizing phenomena, we contend that ultimately he did not go far enough. Determining the relevant demands of the environment on the mechanism often requires detailed empirical investigation. Moreover, often phenomena are reconstituted in the course of inquiry on the mechanism itself. (shrink)

It has been suggested that Marr took the three levels he famously identifies to be independent. In this paper, we argue that Marr's view is more nuanced. Specifically, we show that the view explicitly articulated in his work attempts to integrate the levels, and in doing so results in Marr attacking both reductionism and vagueness. The result is a perspective in which both high-level information-processing constraints and low-level implementational constraints play mutually reinforcing and constraining roles. We discuss our recent work (...) on Spaun—currently the world's largest functional brain model—that demonstrates the positive impact of this kind of unifying integration of Marr's levels. We argue that this kind of integration avoids his concerns with both reductionism and vagueness. In short, we suggest that the methods behind Spaun can be used to satisfy Marr's explicit interest in combining high-level functional and detailed mechanistic explanations. (shrink)

An interesting feature of some sets of representations is that their structure mirrors the structure of the items they represent. Founding an account of representational content on isomorphism, homomorphism or structural resemblance has proven elusive, however, largely because these relations are too liberal when the candidate structure over representational vehicles is unconstrained. Furthermore, in many cases where there is a clear isomorphism, it is not relied on in the way the representations are used. That points to a potential resolution: that (...) an isomorphism must be used, hence usable, if it is to be an ingredient in a theory of content. This paper argues that the class of exploitable isomorphisms can indeed play a content-constituting role. (shrink)

This paper concerns how motor actions are neurally represented and coded. Action planning and motor preparation can be studied using a specific type of representational activity, motor imagery. A close functional equivalence between motor imagery and motor preparation is suggested by the positive effects of imagining movements on motor learning, the similarity between the neural structures involved, and the similar physiological correlates observed in both imaging and preparing. The content of motor representations can be inferred from motor images at a (...) macroscopic level, based on global aspects of the action and the motor rules and constraints which predict the spatial path and kinematics of movements. A more microscopic neural account calls for a representation of object-oriented action. Object attributes are processed in different neural pathways depending on the kind of task the subject is performing. During object-oriented action, a pragmatic representation is activated in which object affordances are transformed into specific motor schemas. Animal as well as human clinical data implicate the posterior parietal and premotor cortical areas in schema instantiation. A mechanism is proposed that is able to encode the desired goal of the action and is applicable to different levels of representational organization. (shrink)

Time underlies many interesting human behaviors. Thus, the question of how to represent time in connectionist models is very important. One approach is to represent time implicitly by its effects on processing rather than explicitly (as in a spatial representation). The current report develops a proposal along these lines first described by Jordan (1986) which involves the use of recurrent links in order to provide networks with a dynamic memory. In this approach, hidden unit patterns are fed back to themselves: (...) the internal representations which develop thus reflect task demands in the context of prior internal states. A set of simulations is reported which range from relatively simple problems (temporal version of XOR) to discovering syntactic/semantic features for words. The networks are able to learn interesting internal representations which incorporate task demands with memory demands: indeed, in this approach the notion of memory is inextricably bound up with task processing. These representations reveal a rich structure, which allows them to be highly context‐dependent, while also expressing generalizations across classes of items. These representations suggest a method for representing lexical categories and the type/token distinction. (shrink)

It is not widely realised that Turing was probably the first person to consider building computing machines out of simple, neuron-like elements connected together into networks in a largely random manner. Turing called his networks 'unorganised machines'. By the application of what he described as 'appropriate interference, mimicking education' an unorganised machine can be trained to perform any task that a Turing machine can carry out, provided the number of 'neurons' is sufficient. Turing proposed simulating both the behaviour of the (...) network and the training process by means of a computer program. We outline Turing's connectionist project of 1948. (shrink)

We begin by distinguishing computationalism from a number of other theses that are sometimes conflated with it. We also distinguish between several important kinds of computation: computation in a generic sense, digital computation, and analog computation. Then, we defend a weak version of computationalism—neural processes are computations in the generic sense. After that, we reject on empirical grounds the common assimilation of neural computation to either analog or digital computation, concluding that neural computation is sui generis. Analog computation requires continuous (...) signals; digital computation requires strings of digits. But current neuroscientific evidence indicates that typical neural signals, such as spike trains, are graded like continuous signals but are constituted by discrete functional elements (spikes); thus, typical neural signals are neither continuous signals nor strings of digits. It follows that neural computation is sui generis. Finally, we highlight three important consequences of a proper understanding of neural computation for the theory of cognition. First, understanding neural computation requires a specially designed mathematical theory (or theories) rather than the mathematical theories of analog or digital computation. Second, several popular views about neural computation turn out to be incorrect. Third, computational theories of cognition that rely on non-neural notions of computation ought to be replaced or reinterpreted in terms of neural computation. (shrink)

Realists about animal cognition confront a puzzle. If animals have real, contentful cognitive states, why can’t anyone say precisely what the contents of those states are? I consider several possible resolutions to this puzzle that are open to realists, and argue that the best of these is likely to appeal to differences in the format of animal cognition and human language.

The two contrasting theoretical approaches to visual perception, the constructivist and the ecological, are briefly presented and illustrated through their analyses of space and size perception. Earlier calls for their reconciliation and unification are reviewed. Neurophysiological, neuropsychological, and psychophysical evidence for the existence of two quite distinct visual systems, the ventral and the dorsal, is presented. These two perceptual systems differ in their functions; the ventral system's central function is that of identification, while the dorsal system is mainly engaged in (...) the visual control of motor behavior. The strong parallels between the ecological approach and the functioning of the dorsal system, and between the constructivist approach and the functioning of the ventral system are noted. It is also shown that the experimental paradigms used by the proponents of these two approaches match the functions of the respective visual systems. A dual-process approach to visual perception emerges from this analysis, with the ecological-dorsal process transpiring mainly without conscious awareness, while the constructivist-ventral process is normally conscious. Some implications of this dual-process approach to visual-perceptual phenomena are presented, with emphasis on space perception. Key Words: constructivist; dual-process approach; ecological; size perception; space perception; two visual systems; visual perception theories. (shrink)

It is argued that a number of important, and seemingly disparate, types of representation are species of a single relation, here called structural representation, that can be described in detail and studied in a way that is of considerable philosophical interest. A structural representation depends on the existence of a common structure between a representation and that which it represents, and it is important because it allows us to reason directly about the representation in order to draw conclusions about the (...) phenomenon that it depicts. The present goal is to give a general and precise account of structural representation, then to use that account to illuminate several problems of current philosophical interest — including some that do not initially seem to involve representation at all. In particular, it is argued that ontological reductions (like that of the natural numbers to sets), compositional accounts of semantics, several important sorts of mental representation, and (perhaps) possible worlds semantics for intensional logics are all species of structural representation and are fruitfully studied in the framework developed here. (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)

The received view is that computational states are individuated at least in part by their semantic properties. I offer an alternative, according to which computational states are individuated by their functional properties. Functional properties are specified by a mechanistic explanation without appealing to any semantic properties. The primary purpose of this paper is to formulate the alternative view of computational individuation, point out that it supports a robust notion of computational explanation, and defend it on the grounds of how computational (...) states are individuated within computability theory and computer science. A secondary purpose is to show that existing arguments for the semantic view are defective. (shrink)

It is not widely realised that Turing was probably the first person to consider building computing machines out of simple, neuron-like elements connected together into networks in a largely random manner. Turing called his networks unorganised machines. By the application of what he described as appropriate interference, mimicking education an unorganised machine can be trained to perform any task that a Turing machine can carry out, provided the number of neurons is sufficient. Turing proposed simulating both the behaviour of the (...) network and the training process by means of a computer program. We outline Turing's connectionist project of 1948. (shrink)

This paper explores the difference between Connectionist proposals for cognitive a r c h i t e c t u r e a n d t h e s o r t s o f m o d e l s t hat have traditionally been assum e d i n c o g n i t i v e s c i e n c e . W e c l a i m t h a t t h (...) e m a j o r d i s t i n c t i o n i s t h a t , w h i l e b o t h Connectionist and Classical architectures postulate representational mental states, the latter but not the former are committed to a symbol-level of representation, or to a ‘language of thought’: i.e., to representational states that have combinatorial syntactic and semantic structure. Several arguments for combinatorial structure in mental representations are then reviewed. These include arguments based on the ‘systematicity’ of mental representation: i.e., on the fact that cognitive capacities always exhibit certain symmetries, so that the ability to entertain a given thought implies the ability to entertain thoughts with semantically related contents. We claim that such arguments make a powerful case that mind/brain architecture is not Connectionist at the cognitive level. We then consider the possibility that Connectionism may provide an account of the neural (or ‘abstract neurological’) structures in which Classical cognitive architecture is implemented. We survey a n u m b e r o f t h e s t a n d a r d a r g u m e n t s t h a t h a v e b e e n o f f e r e d i n f a v o r o f Connectionism, and conclude that they are coherent only on this interpretation. (shrink)

On the occasion of a first conference on Cognitive Science, it seems appropriate to review the basis of common understanding between the various disciplines. In my estimate, the most fundamental contribution so far of artificial intelligence and computer science to the joint enterprise of cognitive science has been the notion of a physical symbol system, i.e., the concept of a broad class of systems capable of having and manipulating symbols, yet realizable in the physical universe. The notion of symbol so (...) defined is internal to this concept, so it becomes a hypothesis that this notion of symbols includes the symbols that we humans use every day of our lives. In this paper we attempt systematically, but plainly, to lay out the nature of physical symbol systems. Such a review is in ways familiar, but not thereby useless. Restatement of fundamentals is an important exercise. (shrink)

Many current neurophysiological, psychophysical, and psychological approaches to vision rest on the idea that when we see, the brain produces an internal representation of the world. The activation of this internal representation is assumed to give rise to the experience of seeing. The problem with this kind of approach is that it leaves unexplained how the existence of such a detailed internal representation might produce visual consciousness. An alternative proposal is made here. We propose that seeing is a way of (...) acting. It is a particular way of exploring the environment. Activity in internal representations does not generate the experience of seeing. The out- side world serves as its own, external, representation. The experience of seeing occurs when the organism masters what we call the gov- erning laws of sensorimotor contingency. The advantage of this approach is that it provides a natural and principled way of accounting for visual consciousness, and for the differences in the perceived quality of sensory experience in the different sensory modalities. Sev- eral lines of empirical evidence are brought forward in support of the theory, in particular: evidence from experiments in sensorimotor adaptation, visual “filling in,” visual stability despite eye movements, change blindness, sensory substitution, and color perception. (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)

This paper analyzes the rapid and unexpected rise of deep learning within Artificial Intelligence and its applications. It tackles the possible reasons for this remarkable success, providing candidate paths towards a satisfactory explanation of why it works so well, at least in some domains. A historical account is given for the ups and downs, which have characterized neural networks research and its evolution from “shallow” to “deep” learning architectures. A precise account of “success” is given, in order to sieve out (...) aspects pertaining to marketing or sociology of research, and the remaining aspects seem to certify a genuine value of deep learning, calling for explanation. The alleged two main propelling factors for deep learning, namely computing hardware performance and neuroscience findings, are scrutinized, and evaluated as relevant but insufficient for a comprehensive explanation. We review various attempts that have been made to provide mathematical foundations able to justify the efficiency of deep learning, and we deem this is the most promising road to follow, even if the current achievements are too scattered and relevant for very limited classes of deep neural models. The authors’ take is that most of what can explain the very nature of why deep learning works at all and even very well across so many domains of application is still to be understood and further research, which addresses the theoretical foundation of artificial learning, is still very much needed. (shrink)

Although machine learning has been successful in recent years and is increasingly being deployed in the sciences, enterprises or administrations, it has rarely been discussed in philosophy beyond the philosophy of mathematics and machine learning. The present contribution addresses the resulting lack of conceptual tools for an epistemological discussion of machine learning by conceiving of deep learning networks as ‘judging machines’ and using the Kantian analysis of judgments for specifying the type of judgment they are capable of. At the center (...) of the argument is the fact that the functionality of deep learning networks is established by training and cannot be explained and justified by reference to a predefined rule-based procedure. Instead, the computational process of a deep learning network is barely explainable and needs further justification, as is shown in reference to the current research literature. Thus, it requires a new form of justification, that is to be specified with the help of Kant’s epistemology. (shrink)

This paper addresses a fundamental line of research in neuroscience: the identification of a putative neural processing core of the cerebral cortex, often claimed to be “canonical”. This “canonical” core would be shared by the entire cortex, and would explain why it is so powerful and diversified in tasks and functions, yet so uniform in architecture. The purpose of this paper is to analyze the search for canonical explanations over the past 40 years, discussing the theoretical frameworks informing this research. (...) It will highlight a bias that, in my opinion, has limited the success of this research project, that of overlooking the dimension of cortical development. The earliest explanation of the cerebral cortex as canonical was attempted by David Marr, deriving putative cortical circuits from general mathematical laws, loosely following a deductive-nomological account. Although Marr’s theory turned out to be incorrect, one of its merits was to have put the issue of cortical circuit development at the top of his agenda. This aspect has been largely neglected in much of the research on canonical models that has followed. Models proposed in the 1980s were conceived as mechanistic. They identified a small number of components that interacted as a basic circuit, with each component defined as a function. More recent models have been presented as idealized canonical computations, distinct from mechanistic explanations, due to the lack of identifiable cortical components. Currently, the entire enterprise of coming up with a single canonical explanation has been criticized as being misguided, and the premise of the uniformity of the cortex has been strongly challenged. This debate is analyzed here. The legacy of the canonical circuit concept is reflected in both positive and negative ways in recent large-scale brain projects, such as the Human Brain Project. One positive aspect is that these projects might achieve the aim of producing detailed simulations of cortical electrical activity, a negative one regards whether they will be able to find ways of simulating how circuits actually develop. (shrink)

In this paper we defend structural representations, more specifically neural structural representation. We are not alone in this, many are currently engaged in this endeavor. The direction we take, however, diverges from the main road, a road paved by the mathematical theory of measure that, in the 1970s, established homomorphism as the way to map empirical domains of things in the world to the codomain of numbers. By adopting the mind as codomain, this mapping became a boon for all those (...) convinced that a representation system should bear similarities with what was being represented, but struggled to find a precise account of what such similarities mean. The euforia was brief, however, and soon homomorphism revealed itself to be affected by serious weaknesses, the primary one being that it included systems embarrassingly alien to representations. We find that the defense attempts that have followed, adopt strategies that share a common format: valid structural representations come as “homomorphism plus X”, with various “X”, provided in descriptive format only. Our alternative direction stems from the observation of the overlooked departure from homomorphism as used in the theory of measure and its later use in mental representations. In the former case, the codomain or the realm of numbers, is the most suited for developing theorems detailing the existence and uniqueness of homomorphism for a wide range of empirical domains. In the latter case, the codomain is the realm of the mind, possibly more vague and more ill-defined than the empirical domain itself. The time is ripe for articulating the mapping between represented domains and the mind in formal terms, by exploiting what is currently known about coding mechanisms in the brain. We provide a sketch of a possible development in this direction, one that adopts the theory of neural population coding as codomain. We will show that our framework is not only not in disagreement with the “plus X” proposals, but can lead to natural derivation of several of the “X”. (shrink)

Since the cognitive revolution, it has become commonplace that cognition involves both computation and information processing. Is this one claim or two? Is computation the same as information processing? The two terms are often used interchangeably, but this usage masks important differences. In this paper, we distinguish information processing from computation and examine some of their mutual relations, shedding light on the role each can play in a theory of cognition. We recommend that theorists of cognition be explicit and careful (...) in choosing notions of computation and information and connecting them together.Keywords: Computation; Information processing; Computationalism; Computational theory of mind; Cognitivism. (shrink)

Deep learning is a kind of machine learning which happens in a certain type of artificial neural networks called deep networks. Artificial deep networks, which exhibit many similarities with biological ones, have consistently shown human-like performance in many intelligent tasks. This poses the question whether this performance is caused by such similarities. After reviewing the structure and learning processes of artificial and biological neural networks, we outline two important reasons for the success of deep learning, namely the extraction of successively (...) higher level features and the multiple layer structure, which are closely related to each other. Then some indications about the framing of this heated debate are given. After that, an assessment of the value of artificial deep networks as models of the human brain is given from the similarity perspective of model representation. Finally, a new version of computational functionalism is proposed which addresses the specificity of deep neural computation better than classic, program based computational functionalism. (shrink)

Recent progress in artificial intelligence has renewed interest in building systems that learn and think like people. Many advances have come from using deep neural networks trained end-to-end in tasks such as object recognition, video games, and board games, achieving performance that equals or even beats that of humans in some respects. Despite their biological inspiration and performance achievements, these systems differ from human intelligence in crucial ways. We review progress in cognitive science suggesting that truly human-like learning and thinking (...) machines will have to reach beyond current engineering trends in both what they learn and how they learn it. Specifically, we argue that these machines should build causal models of the world that support explanation and understanding, rather than merely solving pattern recognition problems; ground learning in intuitive theories of physics and psychology to support and enrich the knowledge that is learned; and harness compositionality and learning-to-learn to rapidly acquire and generalize knowledge to new tasks and situations. We suggest concrete challenges and promising routes toward these goals that can combine the strengths of recent neural network advances with more structured cognitive models. (shrink)

I examine the following question: Do actions require representations that are intrinsic to the action itself? Recent work by Mark Rowlands, Michael Wheeler, and Andy Clark suggests that actions may require a minimal form of representation. I argue that the various concepts of minimal representation on offer do not apply to action per se and that a non‐representationalist account that focuses on dynamic systems of self‐organizing continuous reciprocal causation at the sub‐personal level is superior. I further recommend a scientific pragmatism (...) regarding the concept of representation. (shrink)

We distinguish between three philosophical views on the neuroscience of predictive models: predictive coding, predictive processing and predictive engagement. We examine the concept of active inference under each model and then ask how this concept informs discussions of social cognition. In this context we consider Frith and Friston’s proposal for a neural hermeneutics, and we explore the alternative model of enactivist hermeneutics.

In this paper we defend structural representations, more specifically neural structural representation. We are not alone in this, many are currently engaged in this endeavor. The direction we take, however, diverges from the main road, a road paved by the mathematical theory of measure that, in the 1970s, established homomorphism as the way to map empirical domains of things in the world to the codomain of numbers. By adopting the mind as codomain, this mapping became a boon for all those (...) convinced that a representation system should bear similarities with what was being represented, but struggled to find a precise account of what such similarities mean. The euforia was brief, however, and soon homomorphism revealed itself to be affected by serious weaknesses, the primary one being that it included systems embarrassingly alien to representations. We find that the defense attempts that have followed, adopt strategies that share a common format: valid structural representations come as “homomorphism plus X”, with various “X”, provided in descriptive format only. Our alternative direction stems from the observation of the overlooked departure from homomorphism as used in the theory of measure and its later use in mental representations. In the former case, the codomain or the realm of numbers, is the most suited for developing theorems detailing the existence and uniqueness of homomorphism for a wide range of empirical domains. In the latter case, the codomain is the realm of the mind, possibly more vague and more ill-defined than the empirical domain itself. The time is ripe for articulating the mapping between represented domains and the mind in formal terms, by exploiting what is currently known about coding mechanisms in the brain. We provide a sketch of a possible development in this direction, one that adopts the theory of neural population coding as codomain. We will show that our framework is not only not in disagreement with the “plus X” proposals, but can lead to natural derivation of several of the “X”. (shrink)

Cognitive niche construction is the process whereby organisms create and maintain cause–effect models of their niche as guides for fitness influencing behavior. Extended mind theory claims that cognitive processes extend beyond the brain to include predictable states of the world. Active inference and predictive processing in cognitive science assume that organisms embody predictive (i.e., generative) models of the world optimized by standard cognitive functions (e.g., perception, action, learning). This paper presents an active inference formulation that views cognitive niche construction as (...) a cognitive function aimed at optimizing organisms' generative models. We call that process of optimization extended active inference. (shrink)

Predictive processing approaches to the mind are increasingly popular in the cognitive sciences. This surge of interest is accompanied by a proliferation of philosophical arguments, which seek to either extend or oppose various aspects of the emerging framework. In particular, the question of how to position predictive processing with respect to enactive and embodied cognition has become a topic of intense debate. While these arguments are certainly of valuable scientific and philosophical merit, they risk underestimating the variety of approaches gathered (...) under the predictive label. Here, we first present a basic review of neuroscientific, cognitive, and philosophical approaches to PP, to illustrate how these range from solidly cognitivist applications—with a firm commitment to modular, internalistic mental representation—to more moderate views emphasizing the importance of ‘body-representations’, and finally to those which fit comfortably with radically enactive, embodied, and dynamic theories of mind. Any nascent predictive processing theory must take into account this continuum of views, and associated theoretical commitments. As a final point, we illustrate how the Free Energy Principle attempts to dissolve tension between internalist and externalist accounts of cognition, by providing a formal synthetic account of how internal ‘representations’ arise from autopoietic self-organization. The FEP thus furnishes empirically productive process theories by which to guide discovery through the formal modelling of the embodied mind. (shrink)