Andrew Clark

Mindware: An Introduction to the Philosophy of Cognitive Science invites readers to join in up-to-the-minute conceptual discussions of the fundamental issues, problems, and opportunities in cognitive science. Written by one of the most renowned scholars in the field, this vivid and engaging introductory text relates the story of the search for a cognitive scientific understanding of mind. This search is presented as a no-holds-barred journey from early work in artificial intelligence, through connectionist (artificial neural network) counter-visions, and on to neuroscience, artificial life, dynamics, and robotics. The journey ends with some wide-ranging and provocative speculation about the complex coadaptive dance between mind, culture, and technology. Each chapter opens with a brief sketch of a major research tradition or perspective, followed by short yet substantial critical discussions dealing with key topics and problems. Ranging across both standard philosophical territory and the landscape of cutting-edge cognitive science, Clark highlights challenging issues in an effort to engage readers in active debate. Topics covered include mental causation; machine intelligence; the nature and status of folk psychology; the hardware/software distinction; emergence; relations between life and mind; the nature of perception, cognition, and action; and the continuity (or otherwise) of high-level human intelligence with other forms of adaptive response. Numerous illustrations, text boxes, and extensive suggestions for further reading enhance the text's utility. Helpful appendices provide background information on dualism, behaviorism, identity theory, consciousness, and more. An exceptional text for introductory and more advanced courses in cognitive science and the philosophy of mind, Mindware is also essential reading for anyone interested in these fascinating and ever-changing fields

O'Regan and Noe present a wonderfully detailed and comprehensive defense of a position whose broad outline we absolutely and unreservedly endorse. They are right, it seems to us, to stress the intimacy of conscious content and embodied action, and to counter the idea of a Grand Illusion with the image of an agent genuinely in touch, via active exploration, with the rich and varied visual scene. This is an enormously impressive achievement, and we hope that the comments that follow will be taken in a spirit of constructive questioning. Overall, we have two main reservations.

Can what we know change what we see? Does language affect cognition and perception? The last few years have seen increased attention to these seemingly disparate questions, but with little theoretical advance. We argue that substantial clarity can be gained by considering these questions through the lens of predictive processing, a framework in which mental representations—from the perceptual to the cognitive—reflect an interplay between downward-flowing predictions and upward-flowing sensory signals. This framework provides a parsimonious account of how what we know ought to change what we see and helps us understand how a putatively high-level trait such as language can impact putatively low-level processes such as perception. Within this framework, language begins to take on a surprisingly central role in cognition by providing a uniquely focused and flexible means of constructing predictions against which sensory signals can be evaluated. Predictive processing thus provides a plausible mechanism for many of the reported effects of language on perception, thought, and action, and new insights on how and when speakers of different languages construct the same “reality” in alternate ways.

Over the last 30 years, representationalist and dynamicist positions in the philosophy of cognitive science have argued over whether neurocognitive processes should be viewed as representational or not. Major scientific and technological developments over the years have furnished both parties with ever more sophisticated conceptual weaponry. In recent years, an enactive generalization of predictive processing – known as active inference – has been proposed as a unifying theory of brain functions. Since then, active inference has fueled both representationalist and dynamicist campaigns. However, we believe that when diving into the formal details of active inference, one should be able to find a solution to the war; if not a peace treaty, surely an armistice of a sort. Based on an analysis of these formal details, this paper shows how both representationalist and dynamicist sensibilities can peacefully coexist within the new territory of active inference.

There is a school of thought which links connectionist models of cognition to eliminativism-the thesis that the constructs of commonsense psychology do not exist. This way of construing the impact of connectionist modelling is, I argue, deeply mistaken and depends crucially on a shallow analysis of the notion of explanation. I argue that good, higher level descriptions may group together physically heterogenous mechanisms, and that the constructs of folk psychology may fulfil such a grouping function even if they fail to pick out discrete states of individual processing. More speculatively, the paper goes on to suggest that the virtues which recommend such constructs from a theorist’s third person perspective may nonetheless have close analogues in individual processing. Various difficulties for what I term a pure distributed connectionism may be solved by systems which utilise both distributed and classical symbolic representations-the latter exhibiting the discreteness and semantic interpretability which the eliminativist requires to vindicate common-sense psychology. If human beings turned out to be such mixed systems, then the eliminativist claim would be doubly misguided. It would be false as a conditional since even if pure distributed connectionism were a complete and accurate formal model of individual processing, it would not follow that the other, higher level constructs were not accurate and essential tools for a different kind of explanation. And it would involve a false antecedent since in a mixed system the symbolic descriptions may indeed be incarnate in the system’s own individual processing. The paper is in four main sections. Section 1 attempts to systematise a certain received picture of the relative status and accuracy of various levels of description of a connectionist system. Section 2 then introduces a general model of explanations which aim to group systems into equivalence classes defined for various purposes. Each grouping requires a special vocabulary and the constructs of any given vocabulary are legitimate just insofar as the grouping is interesting and useful. Section 3 goes on to show that, relative to such a model of explanation, the constructs of both symbolic AI and commonsense psychology may have a legitimate role to play in giving psychological explanations. This role is not just that of a useful approximation. The paper ends with a speculative section in which the argument for the theoretical usefulness of such symbolic constructs is extended, in a very natural way, to the domain of individual processing. Here the cognizer, in the process of regulating, debugging and understanding her own representations, creates symbols to stand for sets of distributed activity patterns. The section points out some difficulties for a pure distributed approach which may be eased by the addition of such symbolic constructs and relates the speculations to the current debate over the ’correct’architecture of cognition. If the speculations are on target, this whole debate turns out to be fundamentally illposed.

After offering a brief account of how we understand the shared circuits model (SCM), we divide our response into four sections. First, in section R1, we assess to what extent SCM is committed to an account of the ontogeny and phylogeny of shared circuits. In section R2, we examine doubts raised by several commentators as to whether SCM might be expanded so as to accommodate the mirroring of emotions, sensations, and intransitive actions more generally. Section R3 responds to various criticisms that relate to the account of social-learning Hurley proposes in the target article. We conclude in section R4 by responding to a number of commentators who argued for the limitation of control theory as a framework for studying social cognition

Ranging across both standard philosophical territory and the landscape of cutting-edge cognitive science, Mindware: An Introduction to the Philosophy of Cognitive Science, Second Edition, is a vivid and engaging introduction to key issues, research, and opportunities in the field.Starting with the vision of mindware as software and debates between realists, instrumentalists, and eliminativists, Andy Clark takes students on a no-holds-barred journey through connectionism, dynamical systems, and real-world robotics before moving on to the frontiers of cognitive technologies, enactivism, predictive coding, and the extended mind. Throughout, he highlights challenging issues in an effort to engage students in active debate. Each chapter opens with a brief sketch of a major research tradition or perspective, followed by concise critical discussions dealing with key topics and problems.

Recent studies such as Thelen and Smith, Kelso, Van Gelder, Beer, and others have presented a forceful case for a dynamical systems approach to understanding cognition and adaptive behavior. These studies call into question some foundational assumptions concerning the nature of cognitive scientific explanation and the role of notions such as internal representation and computation. These are exciting and important challenges. But they must be handled with care. It is all to easy in this debate to lose sight of the explanatorily important issues and to talk at cross purposes, courtesy of the various ways in which different theorists often concieve key terms.

Selinger and Engstrom, A moratorium on cyborgs: Computation, cognition and commerce, 2008 (this issue) urge upon us a moratorium on ‘cyborg discourse’. But the argument underestimates the richness and complexity of our ongoing communal explorations. It leans on a somewhat outdated version of the machine metaphor (exemplified perhaps by a frozen 1970’s Cyborg). The modern cyborg, informed by an evolving computational model of mind, can play a positive role in the critical discussions that Selinger and Engstrom seek.

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.

Some regularities enjoy only an attenuated existence in a body of training data. These are regularities whose statistical visibility depends on some systematic recoding of the data. The space of possible recodings is, however, infinitely large – it is the space of applicable Turing machines. As a result, mappings that pivot on such attenuated regularities cannot, in general, be found by brute-force search. The class of problems that present such mappings we call the class of “type-2 problems.” Type-1 problems, by contrast, present tractable problems of search insofar as the relevant regularities can be found by sampling the input data as originally coded. Type-2 problems, we suggest, present neither rare nor pathological cases. They are rife in biologically realistic settings and in domains ranging from simple animat behaviors to language acquisition. Not only are such problems rife – they are standardly solved! This presents a puzzle. How, given the statistical intractability of these type-2 cases, does nature turn the trick? One answer, which we do not pursue, is to suppose that evolution gifts us with exactly the right set of recoding biases so as to reduce specific type-2 problems to type-1 mappings. Such a heavy-duty nativism is no doubt sometimes plausible. But we believe there are other, more general mechanisms also at work. Such mechanisms provide general strategies for managing problems of type-2 complexity. Several such mechanisms are investigated. At the heart of each is a fundamental ploy – namely, the maximal exploitation of states of representation already achieved by prior, simpler learning so as to reduce the amount of subsequent computational search. Such exploitation both characterizes and helps make unitary sense of a diverse range of mechanisms. These include simple incremental learning, modular connectionism, and the developmental hypothesis of “representational redescription”. In addition, the most distinctive features of human cognition – language and culture – may themselves be viewed as adaptations enabling this representation/computation trade-off to be pursued on an even grander scale.

  Marinov''s critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called subsymbolic distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms — as in the case of ID3 — the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed

How is it that thoroughly physical material beings such as ourselves can think, dream, feel, create and understand ideas, theories and concepts? How does mere matter give rise to all these non-material mental states, including consciousness itself? An answer to this central question of our existence is emerging at the busy intersection of neuroscience, psychology, artificial intelligence, and robotics.In this groundbreaking work, philosopher and cognitive scientist Andy Clark explores exciting new theories from these fields that reveal minds like ours to be prediction machines - devices that have evolved to anticipate the incoming streams of sensory stimulation before they arrive. These predictions then initiate actions that structure our worlds and alter the very things we need to engage and predict. Clark takes us on a journey in discovering the circular causal flows and the self-structuring of the environment that define "the predictive brain." What emerges is a bold, new, cutting-edge vision that reveals the brain as our driving force in the daily surf through the waves of sensory stimulation.

Like those famous nations divided by a single tongue, my paper and Professor P.M. Churchland's deep and engaging reply offer different spins on a common heritage. The common heritage is, of course, a connectionist vision of the inner neural economy- a vision which depicts that economy in terms of supra-sentential state spaces, vector-to-vector transformations, and the kinds of skillful pattern-recognition routine we share with the bulk of terrestrial intelligent life-forms. That which divides us is, as ever, much harder to isolate and name. Clearly, it has something to do with the role of moral talk and exchange, and something to do with the conception of morality itself. Most of this Reply will be devoted to clarifying the nature of the disputed territory. First, though I shall rehearse some points of agreement concerning moral talk and progress.

How should we characterize the functional role of conscious visual experience? In particular, how do the conscious contents of visual experience guide, bear upon, or otherwise inform our ongoing motor activities? According to an intuitive and philosophically influential conception, the links are often quite direct. The contents of conscious visual experience, according to this conception, are typically active in the control and guidance of our fine-tuned, real-time engagements with the surrounding three-dimensional world. But this idea is hostage to empirical fortune. It is a hostage, moreover, whose safety is in serious doubt. Thus Milner and Goodale argue for a deep and abiding dissociation between the contents of conscious seeing, on the one hand, and the resources used for the on-line guidance of visuo-motor action, on the other. This ‘dual visual systems’ hypothesis, which finds many echoes in various other bodies of cognitive scientific research, poses a prima facie challenge to the Assumption of Experience-Based Control. More importantly, it provides fuel for an alternative and philosophically suggestive account of the functional role of conscious visual experience.

Predictive processing accounts of neural function view the brain as a kind of prediction machine that forms models of its environment in order to anticipate the upcoming stream of sensory stimulation. These models are then continuously updated in light of incoming error signals. Predictive processing has offered a powerful new perspective on cognition, action, and perception. In this chapter we apply the insights from predictive processing to the study of emotions. The upshot is a picture of emotion as inseparable from perception and cognition, and a key feature of the embodied mind.

The use of forward models is well established in cognitive and computational neuroscience. We compare and contrast two recent, but interestingly divergent, accounts of the place of forward models in the human cognitive architecture. On the Auxiliary Forward Model account, forward models are special-purpose prediction mechanisms implemented by additional circuitry distinct from core mechanisms of perception and action. On the Integral Forward Model account, forward models lie at the heart of all forms of perception and action. We compare these neighbouring but importantly different visions and consider their implications for the cognitive sciences. We end by asking what kinds of empirical research might offer evidence favouring one or the other of these approaches.