This paper advocates explicitness about the type of entity to be considered as content- bearing in connectionist systems; it makes a positive proposal about how vehicles of content should be individuated; and it deploys that proposal to argue in favour of representation in connectionist systems. The proposal is that the vehicles of content in some connectionist systems are clusters in the state space of a hidden layer. Attributing content to such vehicles is required to vindicate the standard explanation for some (...) classificatory networks’ ability to generalise to novel samples their correct classification of the samples on which they were trained. (shrink)

Jerry Fodor’s arguments against the possibility of concept learning, and the responses that have been offered in defense of the coherence of concept learning, have both by and large assumed that concept learning is a descriptive process. I offer an alternative, ostensive approach to concept learning and explain how descriptive concept learning can be explained as a version of ostensive concept learning. I argue that an ostensive view of concept learning offers an empirically plausible and philosophically adequate account of concept (...) learning that is explanatorily superior to traditional descriptive views. (shrink)

forthcoming in R. Menary (ed.), The Extended Mind.

This paper evaluates the Natural-Kinds Argument for cognitive extension, which purports to show that the kinds presupposed by our best cognitive science have instances external to human organism. Various interpretations of the argument are articulated and evaluated, using the overarching categories of memory and cognition as test cases. Particular emphasis is placed on criteria for the scientific legitimacy of generic kinds, that is, kinds characterized in very broad terms rather than in terms of their fine-grained causal roles. Given the current (...) state of cognitive science, I conclude that we have no reason to think memory or cognition are generic natural kinds that can ground an argument for cognitive extension. (shrink)

The possibility of group minds or group mental states has been considered by a number of authors addressing issues in social epistemology and related areas (Goldman 2004, Pettit 2003, Gilbert 2004, Hutchins 1995). An appeal to group minds might, in the end, do indispensable explanatory work in the social or cognitive sciences. I am skeptical, though, and this essay lays out some of the reasons for my skepticism. The concerns raised herein constitute challenges to the advocates of group minds (or (...) group mental states), challenges that might be overcome as theoretical and empirical work proceeds. Nevertheless, these hurdles are, I think, genuine and substantive, so much so that my tentative conclusion will not be optimistic. If a group mind is supposed to be a single mental system having two or more minds as proper parts,¹ the prospects for group minds seem dim. (shrink)

The recent literature on mental causation has not been kind to nonreductive, materialist functionalism (‘functionalism’, hereafter, except where that term is otherwise qualified). The exclusion problem2 has done much of the damage, but the epiphenomenalist threat has taken other forms. Functionalism also faces what I will call the ‘problem of metaphysically necessary effects’ (Block, 1990, pp. 157-60, Antony and Levine, 1997, pp. 91-92, Pereboom, 2002, p. 515, Millikan, 1999, p. 47, Jackson, 1998, pp. 660-61). Functionalist mental properties are individuated partly (...) by their relation to the very effects those properties’ instantiations are thought to cause. Consequently, functionalist causal generalizations would seem to have the following problematical structure: The state of being, among other things, a cause of e (under such-andsuch conditions) causes e (under those conditions).3 The connection asserted lacks the contingency one would expect of a causal generalization. Mental states of the kind in question are, by metaphysical necessity, causes of e; any state that does not cause e is thereby a different kind of state. Yet, a mental state’s being the sort of state it is must play some causal role if functionalism is to account for mental causation.4 In what follows, I first articulate more fully the problem of metaphysically necessary effects. I then criticize three functionalist attempts to solve the problem directly. Given the failure of functionalist efforts to meet the problem head-on, I consider less direct strategies: these involve formulating functionalism or its causal claims in such a way that they appear not to generate the problem of metaphysically necessary effects. I argue against these indirect solutions, in each case concluding either that the problem still arises or that avoiding it requires the adoption of an unorthodox form of functionalism (itself a surprising result). In the final.. (shrink)

Causal theories of mental content (CTs) ground certain aspects of a concept's meaning in the causal relations a concept bears to what it represents. Section 1 explains the problems CTs are meant to solve and introduces terminology commonly used to discuss these problems. Section 2 specifies criteria that any acceptable CT must satisfy. Sections 3, 4, and 5 critically survey various CTs, including those proposed by Fred Dretske, Jerry Fodor, Ruth Garrett Millikan, David Papineau, Dennis Stampe, Dan Ryder, and the (...) author himself. The final section considers general objections to the causal approach. (shrink)

Observations from philosophy and psychology heavily favor the Empiricist tenet that many lexical concepts are learned. However, many observations also heavily favor the Nativist tenet that such concepts are representationally atomic. Fodor Representations: Philosophical essays on the foundations of cognitive science, 1981, LOT2: The language of thought revisited, 2008) has famously argued that representationally atomic concepts cannot be learned, at least not learned by hypothesis formation and testing. Concept theorists who want to preserve observations about concept learning have developed acquisition (...) models on which the acquired concepts are either non-atomic or are acquired by a process that doesn’t involve hypothesis formation and testing. I offer a model, Baptizing Meanings for Concepts, in which representationally atomic concepts are learned by hypothesis formation and testing. The concepts are learned by the agent’s hypothesizing the existence of a latent/hidden/imperceptible property in objects to explain the objects’ perceptible similarities. Once a hidden property is hypothesized, a new atomic mental name is assigned to it, and this atomic name becomes the concept. Any connections between the name and the representations involved in linking the name to its referent are stored as contingent. Further experience may give the agent reason to revise its hypotheses about latent properties as explanations for its observations. I discuss a software robot implementation of the BMC process that uses a Bayesian learning network. The implementation provides an existence proof of the possibility of learning representationally atomic concepts by hypothesis formation and testing. (shrink)

Observations from philosophy and psychology heavily favor the Empiricist tenet that many lexical concepts are learned. However, many observations also heavily favor the Nativist tenet that such concepts are representationally atomic. Fodor Representations: Philosophical essays on the foundations of cognitive science, 1981, LOT2: The language of thought revisited, 2008) has famously argued that representationally atomic concepts cannot be learned, at least not learned by hypothesis formation and testing. Concept theorists who want to preserve observations about concept learning have developed acquisition (...) models on which the acquired concepts are either non-atomic or are acquired by a process that doesn’t involve hypothesis formation and testing. I offer a model, Baptizing Meanings for Concepts, in which representationally atomic concepts are learned by hypothesis formation and testing. The concepts are learned by the agent’s hypothesizing the existence of a latent/hidden/imperceptible property in objects to explain the objects’ perceptible similarities. Once a hidden property is hypothesized, a new atomic mental name is assigned to it, and this atomic name becomes the concept. Any connections between the name and the representations involved in linking the name to its referent are stored as contingent. Further experience may give the agent reason to revise its hypotheses about latent properties as explanations for its observations. I discuss a software robot implementation of the BMC process that uses a Bayesian learning network. The implementation provides an existence proof of the possibility of learning representationally atomic concepts by hypothesis formation and testing. (shrink)

Quantities like mass and temperature are properties that come in degrees. And those degrees (e.g. 5 kg) are properties that are called the magnitudes of the quantities. Some philosophers (e.g., Byrne 2003; Byrne & Hilbert 2003; Schroer 2010) talk about magnitudes of phenomenal qualities as if some of our phenomenal qualities are quantities. The goal of this essay is to explore the anti-physicalist implication of this apparently innocent way of conceptualizing phenomenal quantities. I will first argue for a metaphysical thesis (...) about the nature of magnitudes based on Yablo’s proportionality requirement of causation. Then, I will show that, if some phenomenal qualities are indeed quantities, there can be no demonstrative concepts about some of our phenomenal feelings. That presents a significant restriction on the way physicalists can account for the epistemic gap between the phenomenal and the physical. I’ll illustrate the restriction by showing how that rules out a popular physicalist response to the Knowledge Argument. (shrink)

It is a widely accepted thesis in the cognitive sciences and in naturalistic philosophy of mind that the contents of at least some mental representations are innate. A question that has popped up in discussions concerning innate mental representations is this. Are externalist theories of mental content applicable to the content of innate representations? Views on the matter vary and sometimes conflict. To date, there has been no comprehensive assessment of the relationship between content externalism and content innateness. The aim (...) of this paper is to provide such an assessment. I focus on the notions of innateness that are employed in innateness hypotheses within the cognitive sciences and adjacent fields of philosophy, and on causal externalist theories of content. I distinguish between three accounts of what being innate might amount to in innateness hypotheses within the cognitive sciences, and between three types of causal externalism. I explain what the possibility of innate externalistically individuated representations depends upon given all nine combinations. I explain why causal externalism can be true of innate mental representations, given but one of these combinations. (shrink)

The predominance of machine learning based techniques in cognitive neuroscience raises a host of philosophical and methodological concerns. Given the messiness of neural activity, modellers must make choices about how to structure their raw data to make inferences about encoded representations. This leads to a set of standard methodological assumptions about when abstraction is appropriate in neuroscientific practice. Yet, when made uncritically these choices threaten to bias conclusions about phenomena drawn from data. Contact between the practices of multivariate pattern analysis (...) and philosophy of science can help to illuminate the conditions under which we can use artificial neural networks to better understand neural mechanisms. This paper considers a specific technique for MVPA called representational similarity analysis. I develop a theoretically-informed account of RSA that draws on early connectionist research and work on idealization in the philosophy of science. By bringing a philosophical account of cognitive modelling in conversation with RSA, this paper clarifies the practices of neuroscientists and provides a generalizable framework for using artificial neural networks to study neural mechanisms in the brain. (shrink)

Causal theories of mental content attempt to explain how thoughts can be about things. They attempt to explain how one can think about, for example, dogs. These theories begin with the idea that there are mental representations and that thoughts are meaningful in virtue of a causal connection between a mental representation and some part of the world that is represented. In other words, the point of departure for these theories is that thoughts of dogs are about dogs because dogs (...) cause the mental representations of dogs. (shrink)

Most people find it obvious that concepts like APPLE, DOG, WATER, CACTUS, SWIM, CHIRP, FURRY, and SMOOTH are acquired from perceptual experiences along with some kind of inferential procedure. Models of how these concepts are inferentially acquired, however, force the acquired concepts to be representationally complex, built from, and composed by, the more primitive representations. Since at least the time of Plato, philosophers and psychologists have struggled to find complex sets of representations that have the same meanings, definitionally or probabilistically, (...) as these concepts. For example, to think about the property-kind being gold is not the same as to think about the complex property-kind being yellowish & shiny & malleable.... I call this Fodor's Challenge: Find an acquisition process that is genuinely inferential and yields a concept that genuinely is one of these lexical concepts. Rather than continue the pursuit of a complex representation that has the same meaning as our concept GOLD. I offer a model on which many lexical concepts are acquired from perception and inference, without being built up from, and composed by, the representations involved. The model, Baptizing Meanings for Concepts, is inspired in part by Saul Kripke's baptism process for assigning meanings to linguistic terms. Many lexical concepts, according to the BMC, are acquired by inferring the presence a new property-kind, picking out the property-kind in terms of those perceptible features, and then assigning a simple mental term, a concept, for that property-kind. (shrink)

There is no doubt that social interaction plays an important role in language-learning, as well as in concept acquisition. In surprising contrast, social interaction makes only passing appearance in our most promising naturalistic theories of content. This is particularly true in the case of mental content (e.g., Cummins, 1996; Dretske, 1981, 1988; Fodor, 1987, 1990a; Millikan, 1984); and insofar as linguistic content derives from mental content (Grice, 1957), social interaction seems missing from our best naturalistic theories of both.1 In this (...) paper, I explore the ways in which even the most individualistic of theories of mental content can, and should, accommodate social effects. I focus especially on the way in which inferential relations, including those that are socially taught, influence language-learning and concept acquisition. I argue that these factors affect the way subjects conceive of mental and linguistic content. Such effects have a dark side: the social and inferential processes in question give rise to misleading intuitions about content itself. They create the illusion that content and inferential relations are more deeply intertwined than they actually are. This illusion confounds an otherwise attractive solution to what is known as ‘Frege’s puzzle’ (Salmon, 1986). I.. (shrink)