In contemporary artificial intelligence, the challenge is making intricate connectionist systems—comprising millions of parameters—more comprehensible, defensible, and rationally grounded. Two prevailing methodologies address this complexity. The inaugural approach amalgamates symbolic methodologies with connectionist paradigms, culminating in a hybrid system. This strategy systematizes extensive parameters within a limited framework of formal, symbolic rules. Conversely, the latter strategy remains staunchly connectionist, eschewing hybridity. Instead of internal transparency, it fabricates an external, transparent proxy system. This ancillary system’s mandate is elucidating the principal system’s (...) decisions, essentially approximating its outcomes. Leveraging natural language processing as our analytical lens, this paper elucidates both methodologies: the hybrid method is underscored by the compositional vector semantics, whereas the purely connectionist method evolves as a derivative of neural semantic parsers. This discourse extols the merits of the purely connectionist approach for its inherent flexibility and for a pivotal delineation: segregating the explanatory apparatus from the operational core, thereby rendering artificial intelligence systems reminiscent of human cognition. (shrink)

In demonstration, speakers use real-world activity both for its practical effects and to help make their points. The demonstrations of origami mathematics, for example, reconfigure pieces of paper by folding, while simultaneously allowing their author to signal geometric inferences. Demonstration challenges us to explain how practical actions can get such precise significance and how this meaning compares with that of other representations. In this paper, we propose an explanation inspired by David Lewis’s characterizations of coordination and scorekeeping in conversation. In (...) particular, we argue that words, gestures, diagrams and demonstrations can function together as integrated ensembles that contribute to conversation, because interlocutors use them in parallel ways to coordinate updates to the conversational record. (shrink)

Understanding novel phrases (e.g. upside‐down daisy) and classifying objects in categories named by phrases ought to have common properties, but you'd never know it from current theories. The best candidate for both jobs is the Theory Theory, but it faces difficulties when theories are impoverished. A potential solution is a dual approach that couples theories (representations‐about categories) with fixed mentalese expressions (representations‐of categories). Both representations combine information in parallel when understanding phrases. Although there are objections to the notion that theories (...) can combine, some of them lose their force when applied to a suitable dual model. (shrink)

Imagine advanced computers that could, by virtue merely of being programmed in the right ways, act, react, communicate, and otherwise behave like humans. Might such computers be capable of understanding, thinking, believing, and the like? The framework developed in this paper for tackling challenging questions of concept application (in any realm of discourse) answers in the affirmative, contrary to Searle’s famous ‘Chinese Room’ thought experiment, which purports to prove that ascribing such mental processes to computers like these would be necessarily (...) incorrect. The paper begins by arguing that the core issue concerns language, specifically the discourse-community-guided mapping of phenomena onto linguistic categories. It then offers a model of how people adapt language to deal with novel states of affairs and thereby lend generality to their words, employing processes of assimilation, lexemic creation, and accommodation (in intersense and intrasense varieties). Attributions of understanding to some computers lie in the middle range on a spectrum of acceptability and are thus reasonable. Possible objections deriving from Searle’s writings require supplementing the model with distinctions between present and future acceptability, and between contemplated and uncontemplated word uses, as well as a literal-figurative distinction that is more sensitive than Searle’s to actual linguistic practice and the multiplicity of subsenses possible within a single literal sense. The paper then critiques two misleading rhetorical features of Searle’s Chinese Room presentation, and addresses a contemporary defense of Searle that seems to confront the sociolinguistic issue, but fails to allow for intrasense accommodation. It concludes with a brief consideration of the proper course for productive future discussion. (shrink)

The language we use over the course of conversation changes as we establish common ground and learn what our partner finds meaningful. Here we draw upon recent advances in natural language processing to provide a finer‐grained characterization of the dynamics of this learning process. We release an open corpus (>15,000 utterances) of extended dyadic interactions in a classic repeated reference game task where pairs of participants had to coordinate on how to refer to initially difficult‐to‐describe tangram stimuli. We find that (...) different pairs discover a wide variety of idiosyncratic but efficient and stable solutions to the problem of reference. Furthermore, these conventions are shaped by the communicative context: words that are more discriminative in the initial context (i.e., that are used for one target more than others) are more likely to persist through the final repetition. Finally, we find systematic structure in how a speaker's referring expressions become more efficient over time: Syntactic units drop out in clusters following positive feedback from the listener, eventually leaving short labels containing open‐class parts of speech. These findings provide a higher resolution look at the quantitative dynamics of ad hoc convention formation and support further development of computational models of learning in communication. (shrink)

This article presents a detailed formal approach to concepts and concept combination. Sense generation is a competence‐level theory that attempts to respect constraints from the various cognitive sciences, and postulates “quasi‐classical” conceptual structures where attributes receive only one value (but are defeasible and so do not represent necessary and sufficient conditions on category membership) and where classification is binary (but explicitly context‐sensitive). It is also argued that any general theory of concepts must account for “privative” combinations (e.g., stone lion, fake (...) gun, apparent friend) as extreme test‐cases of representational and classificatory flexibility. The approach presented therefore provides a treatment of these combinations. The approach differentiates between the “lexical concept” (the stable information represented in a mental lexicon) which acts as a base from which the various “senses” (flexible contents associated with words and phrases in context, and used in classification) are “generated.” Generation allows nonmonotonicity, so that in different circumstances, different attributes may be defeated or modified. Classification is treated as relative to the perspective adopted, so that a classification acceptable from one perspective may be unacceptable from another, without contradiction. The result is a view that assumes bottom‐up priority in concept combination, where the range of senses generated by bottom‐up rules of combination is tempered by pragmatic‐communicative constraints on classification. An account of the representational and classification behavior of privative combinations is outlined, and the article concludes with a discussion of some of the implications of the approach. (shrink)

This paper discusses some requirements on a folk-psychological, computational account of concepts. Although most psychological views take the folk-psychological stance that concept-possession requires capacities of both representation and classification, such views lack a philosophical context. In contrast, philosophically motivated views stress one of these capacities at the expense of the other. This paper seeks to provide some philosophical motivation for the (folk-) psychological stance. Philosophical and psychological constraints on a computational level account provide the context for evaluating two theses. The (...) first, the Classificatory View, is that concept-possession is constituted by the ability to classify states of the world. I argue, against this view, that to be able to classify, a thinker must also be able to represent the world. The second thesis, the Representational View, is that to possess a concept is constituted by the ability to represent the world. I argue that ascribing this ability is incoherent without ascribing an ability to classify. Hence, a detailed computational specification of concept-possession suggests that the folk-psychological stance is accurate. Philosophical views of concepts, (e.g. Fodor, 1987), adhering to one of the strong theses, whilst adverting to folk-psychological motivations, are thus both insufficiently complex and incoherent. (shrink)

Traditional theories of sarcasm treat it as a case of a speaker's meaning the opposite of what she says. Recently, 'expressivists' have argued that sarcasm is not a type of speaker meaning at all, but merely the expression of a dissociative attitude toward an evoked thought or perspective. I argue that we should analyze sarcasm in terms of meaning inversion, as the traditional theory does; but that we need to construe 'meaning' more broadly, to include illocutionary force and evaluative attitudes (...) as well as propositional content. I distinguish four subclasses of sarcasm, individuated in terms of the target of inversion. Three of these classes raise serious challenges for a standard implicature analysis. (shrink)

On a familiar and prima facie plausible view of metaphor, speakers who speak metaphorically say one thing in order to mean another. A variety of theorists have recently challenged this view; they offer criteria for distinguishing what is said from what is merely meant, and argue that these support classifying metaphor within 'what is said'. I consider four such criteria, and argue that when properly understood, they support the traditional classification instead. I conclude by sketching how we might extract a (...) workable notion of 'what is said' from ordinary intuitions about saying. (shrink)

The nature of complex concepts has important implications for the computational modelling of the mind, as well as for the cognitive science of concepts. This paper outlines the way in which RVC – a Relational View of Concepts – accommodates a range of complex concepts, cases which have been argued to be non-compositional. RVC attempts to integrate a number of psychological, linguistic and psycholinguistic considerations with the situation-theoretic view that information-carrying relations hold only relative to background situations. The central tenet (...) of RVC is that the content of concepts varies systematically with perspective. The analysis of complex concepts indicates that compositionality too should be considered to be sensitive to perspective. Such a view accords with concepts and mental states being situated and the implications for theories of concepts and for computational models of the mind are discussed. (shrink)

A widely accepted thesis in the philosophy of language is that natural language proper names are rigid designators, and that they are so de jure, or as a matter of the “semantic rules of the language.” This paper questions this claim, arguing that rigidity cannot be plausibly construed as a property of name types and that the alternative, rigidity construed as a property of tokens, means that they cannot be considered rigid de jure; rigidity in this case must be viewed (...) as a pragmatic and not a semantic property. (shrink)

This article is a survey of the literature on context sensitivity in the philosophy of language.

The standard approach to model how human beings understand natural languages is the symbolic, compositional approach according to which the meaning of a complex expression is a function of the meanings of its constituents. In other words, meaning plays a fundamental role in the model. In this work, because of the polysemous, flexible, dynamic, and contextual structure of natural languages, this approach is rejected. Instead, a connectionist model which eliminates the concept of meaning is proposed.

Contribution au numéro spécial de la Revue Internationale de Philosophie sur John Searle.

In the first half of the paper I critically review some previous attempts to deal with metonymy. I focus in particular on the classical approach, the associationist approach and the Gricean approach. The main point of my criticisms is that the notion of empirical associations among objects is in itself inadequate for a complete descriptive and explanatory account of metonymy.

Traditional approaches tend to regard figuration (and by extension, deference in general) as an essentially marked or playful use of language, which is associated with a pronounced stylistic effect. For linguistic purposes, however, there is no reason for assigning a special place to deferred uses that are stylistically notable — the sorts of usages that people sometimes qualify with a phrase like "figuratively speaking." There is no important linguistic difference between using redcoat to refer to a British soldier and using (...) suit to refer to a corporate executive (as in "A couple of suits stopped by to talk about the new products"). What creates the stylistic effect of the latter is not the mechanism that generates it, but the marked background assumptions that license it — here, the playful presupposition that.. (shrink)