The central aim of this paper is to shed light on the nature of explanation in computational neuroscience. I argue that computational models in this domain possess explanatory force to the extent that they describe the mechanisms responsible for producing a given phenomenon—paralleling how other mechanistic models explain. Conceiving computational explanation as a species of mechanistic explanation affords an important distinction between computational models that play genuine explanatory roles and those that merely provide accurate descriptions or predictions of phenomena. It (...) also serves to clarify the pattern of model refinement and elaboration undertaken by computational neuroscientists. (shrink)

In artificial intelligence, recent research has demonstrated the remarkable potential of Deep Convolutional Neural Networks (DCNNs), which seem to exceed state-of-the-art performance in new domains weekly, especially on the sorts of very difficult perceptual discrimination tasks that skeptics thought would remain beyond the reach of artificial intelligence. However, it has proven difficult to explain why DCNNs perform so well. In philosophy of mind, empiricists have long suggested that complex cognition is based on information derived from sensory experience, often appealing to (...) a faculty of abstraction. Rationalists have frequently complained, however, that empiricists never adequately explained how this faculty of abstraction actually works. In this paper, I tie these two questions together, to the mutual benefit of both disciplines. I argue that the architectural features that distinguish DCNNs from earlier neural networks allow them to implement a form of hierarchical processing that I call “transformational abstraction”. Transformational abstraction iteratively converts sensory-based representations of category exemplars into new formats that are increasingly tolerant to “nuisance variation” in input. Reflecting upon the way that DCNNs leverage a combination of linear and non-linear processing to efficiently accomplish this feat allows us to understand how the brain is capable of bi-directional travel between exemplars and abstractions, addressing longstanding problems in empiricist philosophy of mind. I end by considering the prospects for future research on DCNNs, arguing that rather than simply implementing 80s connectionism with more brute-force computation, transformational abstraction counts as a qualitatively distinct form of processing ripe with philosophical and psychological significance, because it is significantly better suited to depict the generic mechanism responsible for this important kind of psychological processing in the brain. (shrink)

Completeness is an important but misunderstood norm of explanation. It has recently been argued that mechanistic accounts of scientific explanation are committed to the thesis that models are complete only if they describe everything about a mechanism and, as a corollary, that incomplete models are always improved by adding more details. If so, mechanistic accounts are at odds with the obvious and important role of abstraction in scientific modelling. We respond to this characterization of the mechanist’s views about abstraction and (...) articulate norms of completeness for mechanistic explanations that have no such unwanted implications. _1_ Introduction _2_ A Balancing Act: When Do Details Matter? _3_ The Norms of Causal Explanation _4_ The Norms of Constitutive Explanation _5_ Salmon-Completeness _6_ From More Details to More Relevant Details _7_ Non-explanatory Virtues of Abstraction _8_ From Explanatory Models to Explanatory Knowledge _9_ Mechanistic Completeness Reconsidered _10_ Conclusion. (shrink)

There have been recent disagreements in the philosophy of neuroscience regarding which sorts of scientific models provide mechanistic explanations, and which do not. These disagreements often hinge on two commonly adopted, but conflicting, ways of understanding mechanistic explanations: what I call the “representation-as” account, and the “representation-of” account. In this paper, I argue that neither account does justice to neuroscientific practice. In their place, I offer a new alternative that can defuse some of these disagreements. I argue that individual models (...) do not provide mechanistic explanations by themselves. Instead, individual models are always used to complement a huge body of background information and pre-existing models about the target system. With this in mind, I argue that mechanistic explanations are distributed across sets of different, and sometimes contradictory, scientific models. Each of these models contributes limited, but essential, information to the same mechanistic explanation, but none can be considered a mechanistic explanation in isolation of the others. (shrink)

In a recent paper, Kaplan (Synthese 183:339–373, 2011) takes up the task of extending Craver’s (Explaining the brain, 2007) mechanistic account of explanation in neuroscience to the new territory of computational neuroscience. He presents the model to mechanism mapping (3M) criterion as a condition for a model’s explanatory adequacy. This mechanistic approach is intended to replace earlier accounts which posited a level of computational analysis conceived as distinct and autonomous from underlying mechanistic details. In this paper I discuss work in (...) computational neuroscience that creates difficulties for the mechanist project. Carandini and Heeger (Nat Rev Neurosci 13:51–62, 2012) propose that many neural response properties can be understood in terms of canonical neural computations. These are “standard computational modules that apply the same fundamental operations in a variety of contexts.” Importantly, these computations can have numerous biophysical realisations, and so straightforward examination of the mechanisms underlying these computations carries little explanatory weight. Through a comparison between this modelling approach and minimal models in other branches of science, I argue that computational neuroscience frequently employs a distinct explanatory style, namely, efficient coding explanation. Such explanations cannot be assimilated into the mechanistic framework but do bear interesting similarities with evolutionary and optimality explanations elsewhere in biology. (shrink)

The view that the brain is a sort of computer has functioned as a theoretical guideline both in cognitive science and, more recently, in neuroscience. But since we can view every physical system as a computer, it has been less than clear what this view amounts to. By considering in some detail a seminal study in computational neuroscience, I first suggest that neuroscientists invoke the computational outlook to explain regularities that are formulated in terms of the information content of electrical (...) signals. I then indicate why computational theories have explanatory force with respect to these regularities:in a nutshell, they underscore correspondence relations between formal/mathematical properties of the electrical signals and formal/mathematical properties of the represented objects. I finally link my proposal to the philosophical thesis that content plays an essential role in computational taxonomy. (shrink)

An increasing number of philosophers have promoted the idea that mechanism provides a fruitful framework for thinking about the explanatory contributions of computational approaches in cognitive neuroscience. For instance, Piccinini and Bahar :453–488, 2013) have recently argued that neural computation constitutes a sui generis category of physical computation which can play a genuine explanatory role in the context of investigating neural and cognitive processes. The core of their proposal is to conceive of computational explanations in cognitive neuroscience as a subspecies (...) of mechanistic explanations. This paper identifies several challenges facing their mechanistic account and sketches an alternative way of thinking about the epistemic roles of computational approaches used in the study of brain and cognition. Drawing on examples from both low-level and systems-level computational neuroscience, I argue that at least some computational explanations of neural and cognitive processes are partially independent from mechanistic constraints. (shrink)

This paper addresses the questions concerningthe relationship between scientific andcognitive processes. The fact that both,science and cognition, aim at acquiring somekind of knowledge or representationabout the world is the key for establishing alink between these two domains. It turns outthat the constructivist frameworkrepresents an adequate epistemologicalfoundation for this undertaking, as its focusof interest is on the (constructive)relationship between the world and itsrepresentation. More specifically, it will beshown how cognitive processes and their primaryconcern to construct a representation of theenvironment and to (...) generate functionallyfitting behavior can act as the basis forembedding the activities and dynamics of theprocess of science in them by making use ofconstructivist concepts, such as functionalfitness, structure determinedness, etc.Cognitive science and artificiallife provide the conceptual framework of representational spaces and their interactionbetween each other and with the environmentenabling us to establish this link betweencognitive processes and thedevelopment/dynamics of scientific theories.The concepts of activation, synaptic weight,and genetic (representational) spaces arepowerful tools which can be used asexplanatory vehiclesfor a cognitivefoundation of science, more specifically forthe context of discovery (i.e., thedevelopment, construction, and dynamics ofscientific theories and paradigms).Representational spaces do not only offer us abetter understanding of embedding science incognition, but also show, how theconstructivist framework, both, can act as anadequate epistemological foundation for theseprocesses and can be instantiated by theserepresentational concepts from cognitivescience. The final part of this paper addresses somemore fundamental questions concerning thepositivistic and constructivist understandingof science and human cognition. Among otherthings it is asked, whether a purelyfunctionalist and quantitative view of theworld aiming almost exclusively at itsprediction and control is really satisfying forour intellect (having the goal of achieving aprofound understanding of reality). (shrink)

How should biological behaviour be modelled? A relatively new approach is to investigate problems in neuroethology by building physical robot models of biological sensorimotor systems. The explication and justification of this approach are here placed within a framework for describing and comparing models in the behavioural and biological sciences. First, simulation models – the representation of a hypothesis about a target system – are distinguished from several other relationships also termed “modelling” in discussions of scientific explanation. Seven dimensions on which (...) simulation models can differ are defined and distinctions between them discussed: 1. Relevance: whether the model tests and generates hypotheses applicable to biology. 2. Level: the elemental units of the model in the hierarchy from atoms to societies. 3. Generality: the range of biological systems the model can represent. 4. Abstraction: the complexity, relative to the target, or amount of detail included in the model. 5. Structural accuracy: how well the model represents the actual mechanisms underlying the behaviour. 6. Performance match: to what extent the model behaviour matches the target behaviour. 7. Medium: the physical basis by which the model is implemented. No specific position in the space of models thus defined is the only correct one, but a good modelling methodology should be explicit about its position and the justification for that position. It is argued that in building robot models biological relevance is more effective than loose biological inspiration; multiple levels can be integrated; that generality cannot be assumed but might emerge from studying specific instances; abstraction is better done by simplification than idealisation; accuracy can be approached through iterations of complete systems; that the model should be able to match and predict target behaviour; and that a physical medium can have significant advantages. These arguments reflect the view that biological behaviour needs to be studied and modelled in context, that is, in terms of the real problems faced by real animals in real environments. Key Words: animal behaviour; levels; models; neuroethology; realism; robotics; simulation. (shrink)

The paper is an examination of the ways and extent to which neuroscience places constraints on cognitive science. In Part I, I clarify the issue, as well as the notion of levels in cognitive inquiry. I then present and address, in Part II, two arguments designed to show that facts from neuroscience are at a level too low to constrain cognitive theory in any important sense. I argue, to the contrary, that there are several respects in which facts from neurophysiology (...) will constrain cognitive theory. Part III then turns to an examination of Connectionism and Classical Cognitivism to determine which, if either, is in a better position to accomodate neural constraints in the ways suggested in Part II. (shrink)

In this paper we review some problems with traditional approaches for acquiring and representing knowledge in the context of developing user interfaces. Methodological implications for knowledge engineering and for human-computer interaction are studied. It turns out that in order to achieve the goal of developing human-oriented (in contrast to technology-oriented) human-computer interfaces developers have to develop sound knowledge of the structure and the representational dynamics of the cognitive system which is interacting with the computer.We show that in a first step (...) it is necessary to study and investigate the different levels and forms of representation that are involved in the interaction processes between computers and human cognitive systems. Only if designers have achieved some understanding about these representational mechanisms, user interfaces enabling individual experiences and skill development can be designed. In this paper we review mechanisms and processes for knowledge representation on a conceptual, epistemological, and methodologieal level, and sketch some ways out of the identified dilemmas for cognitive modeling in the domain of human-computer interaction. (shrink)