Michael Dummett famously maintained that analytic philosophy was simply philosophy that followed Frege in treating the philosophy of language as the basis for all other philosophy (1978, 441). But one important insight to emerge from computer science is how difficult it is to animate the linguistic artifacts that the analysis of thought produces. Yet, modeling the effects of thought requires a new skill that goes beyond analysis: procedural literacy. Some of the most promising research in philosophy makes use of a (...) variety of modeling techniques that go beyond basic logic and elementary probability theory. What unifies this approach is a focus on what Alan Perlis called procedural literacy. This essay argues that the future spoils in philosophical research will disproportionally go to those who are procedurally literate. (shrink)

The Turing Test is routinely understood as a behaviourist test for machine intelligence. Diane Proudfoot has argued for an alternative interpretation. According to Proudfoot, Turing’s claim that intelligence is what he calls ‘an emotional concept’ indicates that he conceived of intelligence in response-dependence terms. As she puts it: ‘Turing’s criterion for “thinking” is…: x is intelligent if in the actual world, in an unrestricted computer-imitates-human game, x appears intelligent to an average interrogator’. The role of the famous test is thus (...) to provide the conditions in which to examine the average interrogator’s responses. I shall argue that Proudfoot’s analysis falls short. The philosophical literature contains two main models of response-dependence, what I shall call the transparency model and the reference-fixing model. Proudfoot resists the thought that Turing might have endorsed one of these models to the exclusion of the other. But the details of her own analysis indicate that she is, in fact, committed to the claim that Turing’s account of intelligence is grounded in a transparency model, rather than a reference-fixing one. By contrast, I shall argue that while Turing did indeed conceive of intelligence in response-dependence terms, his account is grounded in a reference-fixing model, rather than a transparency one. This is fortunate, because, as an account of intelligence, the transparency model is arguably problematic in a way that the reference-fixing model isn’t. (shrink)

Cognitive scientists have long used distributional semantic representations of categories. The predominant approach uses distributional representations of category‐denoting nouns, such as “city” for the category city. We propose a novel scheme that represents categories as prototypes over representations of names of its members, such as “Barcelona,” “Mumbai,” and “Wuhan” for the category city. This name‐based representation empirically outperforms the noun‐based representation on two experiments (modeling human judgments of category relatedness and predicting category membership) with particular improvements for ambiguous nouns. We (...) discuss the model complexity of both classes of models and argue that the name‐based model has superior explanatory potential with regard to concept acquisition. (shrink)

We propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealised explanation game in which players collaborate to find the best explanation for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal (...) patterns of variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions. (shrink)

We propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealisedexplanation gamein which players collaborate to find the best explanation(s) for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal patterns of (...) variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a (conditionally) optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions. (shrink)

Inferences are familiar movements of thought, but despite important recent work on the topic, we do not yet have a fully satisfying theory of inference. Here I provide a functionalist theory of inference. I argue that the functionalist framework allows us the flexibility to meet various demands on a theory of inference that have been proposed (such as that it must explain inferential Moorean phenomena and epistemological ‘taking’). While also allowing us to compare, contrast, adapt, and combine features of extant (...) theories of inference into one unified theory. In fleshing out the inference role, I also criticize the common assumption that inference requires rule-following. (shrink)

Under what circumstances if ever ought we to grant that Artificial Intelligences (AI) are persons? The question of whether AI could have the high degree of moral status that is attributed to human persons has received little attention. What little work there is employs western conceptions of personhood, while non-western approaches are neglected. In this article, I discuss African conceptions of personhood and their implications for the possibility of AI persons. I focus on an African account of personhood that is (...) prima facie inimical to the idea that AI could ever be ‘persons’ in the sense typically attributed to humans. I argue that despite its apparent anthropocentrism, this African account could admit AI as persons. (shrink)

Building artificial moral agents (AMAs) underscores the fragmentary character of presently available models of human ethical behavior. It is a distinctly different enterprise from either the attempt by moral philosophers to illuminate the “ought” of ethics or the research by cognitive scientists directed at revealing the mechanisms that influence moral psychology, and yet it draws on both. Philosophers and cognitive scientists have tended to stress the importance of particular cognitive mechanisms, e.g., reasoning, moral sentiments, heuristics, intuitions, or a moral grammar, (...) in the making of moral decisions. However, assembling a system from the bottom-up which is capable of accommodating moral considerations draws attention to the importance of a much wider array of mechanisms in honing moral intelligence. Moral machines need not emulate human cognitive faculties in order to function satisfactorily in responding to morally significant situations. But working through methods for building AMAs will have a profound effect in deepening an appreciation for the many mechanisms that contribute to a moral acumen, and the manner in which these mechanisms work together. Building AMAs highlights the need for a comprehensive model of how humans arrive at satisfactory moral judgments. (shrink)

I address the problem of indefiniteness in quantum mechanics: the problem that the theory, without changes to its formalism, seems to predict that macroscopic quantities have no definite values. The Everett interpretation is often criticised along these lines, and I shall argue that much of this criticism rests on a false dichotomy: that the macroworld must either be written directly into the formalism or be regarded as somehow illusory. By means of analogy with other areas of physics, I develop the (...) view that the macroworld is instead to be understood in terms of certain structures and patterns which emerge from quantum theory (given appropriate dynamics, in particular decoherence). I extend this view to the observer, and in doing so make contact with functionalist theories of mind. (shrink)

Analogies come in several forms that serve distinct functions. Inductive analogy is a common type of analogical argument, but critical thinking texts sometimes treat all analogies as inductive. Such an analysis ignores figurative analogies, which may elucidate but do not argue; and also neglects a priori arguments by analogy, a type of analogical argument prominent in law and ethics. A priori arguments by analogy are distinctive, but--contrary to the claims of Govier and Sunstein-they are best understood as deductive, rather than (...) a special form of non deductive reasoning. (shrink)

I argue that John Searle's (1980) influential Chinese room argument (CRA) against computationalism and strong AI survives existing objections, including Block's (1998) internalized systems reply, Fodor's (1991b) deviant causal chain reply, and Hauser's (1997) unconscious content reply. However, a new ``essentialist'' reply I construct shows that the CRA as presented by Searle is an unsound argument that relies on a question-begging appeal to intuition. My diagnosis of the CRA relies on an interpretation of computationalism as a scientific theory about the (...) essential nature of intentional content; such theories often yield non-intuitive results in non-standard cases, and so cannot be judged by such intuitions. However, I further argue that the CRA can be transformed into a potentially valid argument against computationalism simply by reinterpreting it as an indeterminacy argument that shows that computationalism cannot explain the ordinary distinction between semantic content and sheer syntactic manipulation, and thus cannot be an adequate account of content. This conclusion admittedly rests on the arguable but plausible assumption that thought content is interestingly determinate. I conclude that the viability of computationalism and strong AI depends on their addressing the indeterminacy objection, but that it is currently unclear how this objection can be successfully addressed. (shrink)

This paper approaches the interaction of a health professional with an AI system for diagnostic purposes as a hybrid decision making process and conceptualizes epistemo-ethical constraints on this process. We argue for the importance of the understanding of the underlying machine epistemology in order to raise awareness of and facilitate realistic expectations from AI as a decision support system, both among healthcare professionals and the potential benefiters. Understanding the epistemic abilities and limitations of such systems is essential if we are (...) to integrate AI into the decision making processes in a way that takes into account its applicability boundaries. This will help to mitigate potential harm due to misjudgments and, as a result, to raise the trust—understood here as a belief in reliability of—in the AI system. We aim at a minimal requirement for AI meta-explanation which should distinguish machine epistemic processes from similar processes in human epistemology in order to avoid confusion and error in judgment and application. An informed approach to the integration of AI systems into the decision making for diagnostic purposes is crucial given its high impact on health and well-being of patients. (shrink)

Vehicle externalism maintains that the vehicles of our mental representations can be located outside of the head, that is, they need not be instantiated by neurons located inside the brain of the cogniser. But some disagree, insisting that ‘non-derived’, or ‘original’, content is the mark of the cognitive and that only biologically instantiated representational vehicles can have non-derived content, while the contents of all extra-neural representational vehicles are derived and thus lie outside the scope of the cognitive. In this paper (...) we develop one aspect of Menary’s vehicle externalist theory of cognitive integration—the process of enculturation—to respond to this longstanding objection. We offer examples of how expert mathematicians introduce new symbols to represent new mathematical possibilities that are not yet understood, and we argue that these new symbols have genuine non-derived content, that is, content that is not dependent on an act of interpretation by a cognitive agent and that does not derive from conventional associations, as many linguistic representations do. (shrink)

In philosophy of mind, zombies are imaginary creatures that are exact physical duplicates of conscious subjects but for whom there is no first-personal experience. Zombies are meant to show that physicalism—the theory that the universe is made up entirely out of physical components—is false. In this paper, I apply the zombie thought experiment to the realm of morality to assess whether moral agency is something independent from sentience. Algorithms, I argue, are a kind of functional moral zombie, such that thinking (...) about the latter can help us better understand and regulate the former. I contend that the main reason why algorithms can be neither autonomous nor accountable is that they lack sentience. Moral zombies and algorithms are incoherent as moral agents because they lack the necessary moral understanding to be morally responsible. To understand what it means to inflict pain on someone, it is necessary to have experiential knowledge of pain. At most, for an algorithm that feels nothing, ‘values’ will be items on a list, possibly prioritised in a certain way according to a number that represents weightiness. But entities that do not feel cannot value, and beings that do not value cannot act for moral reasons. (shrink)

Of all contemporary social theorists, Luhmann has best understood the centrality of the concept of meaning to social theory and has most extensively worked out the notion's implications. However, despite the power of his theory, the theory suffers from difficulties impeding its reception. This article attempts to remedy this situation with some critical arguments and proposals for revision. First, the theory Luhmann adopted from biology as the basis of his own theory was a poor choice since that theory has no (...) explanatory power, being purely descriptive; furthermore, that theory is fundamentally flawed since it implies that viruses are impossible. Second, Luhmann's theory of meaning cannot coherently make the social domain autonomous as he desires since Luhmann does not take into account the distinction between syntax and semantics. By introducing this distinction, making clear that social systems consist of rules, not just communications, and raising the rule concept to the same prominence in social theory as those of actor and system, autonomy can be maintained while avoiding the counterintuitive aspects of Luhmann's theory. (shrink)

Glenberg's account falls short in several respects. Besides requiring clearer explication of basic concepts, his account fails to recognize the autonomous nature of perception. His account of what is remembered, and its description, is too static. His strictures against connectionist modeling might be overcome by combining the notions of psychological space and principled learning in an embodied and situated network.

Explanations of cognitive processes provided by traditional artificial intelligence were based on the notion of the knowledge level. This perspective has been challenged by new AI that proposes an approach based on embodied systems that interact with the real‐world. We demonstrate that these two views can be unified. Our argument is based on the assumption that knowledge level explanations can be defined in the context of Bayesian theory while the goals of new AI are captured by using a well established (...) robot based model of learning and problem solving, called Distributed Adaptive Control (DAC). In our analysis we consider random foraging and we prove that minor modifications of the DAC architecture renders a model that is equivalent to a Bayesian analysis of this task. Subsequently, we compare this enhanced, “rational,” model to its “non‐rational” predecessor and a further control condition using both simulated and real robots, in a variety of environments. Our results show that the changes made to the DAC architecture, in order to unify the perspectives of old and new AI, also lead to a significant improvement in random foraging. (shrink)

Explanations of cognitive processes provided by traditional artificial intelligence were based on the notion of the knowledge level. This perspective has been challenged by new AI that proposes an approach based on embodied systems that interact with the real‐world. We demonstrate that these two views can be unified. Our argument is based on the assumption that knowledge level explanations can be defined in the context of Bayesian theory while the goals of new AI are captured by using a well established (...) robot based model of learning and problem solving, called Distributed Adaptive Control (DAC). In our analysis we consider random foraging and we prove that minor modifications of the DAC architecture renders a model that is equivalent to a Bayesian analysis of this task. Subsequently, we compare this enhanced, “rational,” model to its “non‐rational” predecessor and a further control condition using both simulated and real robots, in a variety of environments. Our results show that the changes made to the DAC architecture, in order to unify the perspectives of old and new AI, also lead to a significant improvement in random foraging. (shrink)

This commentary elaborates on Gray's conclusion that his neurophysiological model of consciousness might explain how consciousness arises from the brain, but does not address how consciousness evolved, affects behaviour or confers survival value. The commentary argues that such limitations apply to all neurophysiological or other third-person perspective models. To approach such questions the first-person nature of consciousness needs to be taken seriously in combination with third-person models of the brain.

What, in essence, characterizes the mind? According to Searle, the potential to be conscious provides the only definitive criterion. Thus, conscious states are unquestionably "mental"; "shallow unconscious" states are also "mental" by virtue of their capacity to be conscious (at least in principle); but there are no "deep unconscious mental states" - i.e. those rules and procedures without access to consciousness, inferred by cognitive science to characterize the operations of the unconscious mind are not mental at all. Indeed, according to (...) Searle, they have no ontological status - they are simply ways of describing some interesting facets of purely physiological phenomena. (shrink)

A commentary on a target article by Alison Gopnik (1993) How we know our minds: the illusion of first-person knowledge of intentionality. Focusing on evidence of how children acquire a theory of mind, this commentary argues that there are internal inconsistencies in theories that both argue for the functional role of conscious experiences and the irreducibility of those experiences to third-person viewable information processing.

According to the dominant computational approach in cognitive science, cognitive agents are digital computers; according to the alternative approach, they are dynamical systems. This target article attempts to articulate and support the dynamical hypothesis. The dynamical hypothesis has two major components: the nature hypothesis (cognitive agents are dynamical systems) and the knowledge hypothesis (cognitive agents can be understood dynamically). A wide range of objections to this hypothesis can be rebutted. The conclusion is that cognitive systems may well be dynamical systems, (...) and only sustained empirical research in cognitive science will determine the extent to which that is true. (shrink)

In “Absent Qualia and the Mind-Body Problem,” Michael Tye (2006) presents an argument by which he claims to show the inconceivability of beings that are functionally equivalent to phenomenally conscious beings but lack any qualia. On that basis, he concludes that qualia can be fully defined in functional terms. The argument does not suffice to establish the claimed results. In particular it does not show that such absent qualia cases are inconceivable. Tye’s argument relies on a principle P according to (...) which the exchange of isomorphic states between functionally equivalent systems will preserve their equivalence. If they were functionally equivalent before the exchange, they will also be so after. Consideration of the contextual nature of realization shows that Principle P is not a general truth as Tye claims, and his argument against the possibility of absent qualia thus fails. (shrink)

This introduction presents an overview of the articles in this special issue, within the framework of an argument for the conclusion that there are various roads leading from imagination to knowledge.

Conceivability arguments constitute a serious threat against reductive physicalism. Recently, a number of authors have proven and characterized a devastating logical truth, centered on these arguments: namely, that their soundness entails the inconceivability of reductive physicalism. In this paper, I demonstrate that is only a logical truth when reductive physicalism is interpreted in its stronger, intrinsic sense, as opposed to its weaker—yet considerably more popular—extrinsic sense. The basic idea generalizes: perhaps surprisingly, stronger forms of reduction are uniquely resistant to the (...) conceivability arguments opposing them. So far as the modal epistemology of reduction is concerned, therefore, it pays to go intrinsic. (shrink)

Tato má esej vychází z četby knihy Marca de Meye Kognitivní paradigma. Na základě jeho výkladu kognitivního paradigmatu charakterizuji kognitivní pohled, který de Mey zastává. Dále stručně charakterizuji Searlovo pojetí silné umělé inteligence, které vidím jako hlavní cíl, o jehož dosažení by disciplína umělé inteligence měla usilovat. Konečně pak zkoumám, jakým způsobem by se dala vytyčená kognitivní pozice uplatnit při snaze o vytvoření silné umělé inteligence. Průniky, které zde vidím, spočívají především v samotném modelování reprezentací světa, v mnohosti takových modelů (...) schopných spojit se do jednotného celku a v propojení stratifikovaného modelu vnímání s akcemi, jež vede k vytvoření nejprve implicitní a posléze i explicitní podoby znalostí. (shrink)