Mike Anderson1 has given us a thoughtful and useful field guide: Not in the genre of a bird-watcher’s guide which is carried in the field and which contains detailed descriptions of possible sightings, but in the sense of a guide to a field (in this case embodied cognition) which aims to identify that field’s general principles and properties. I’d like to make some comments that will hopefully complement Anderson’s work, highlighting points of agreement and disagreement between his view of the (...) field and my own, and acting as a devil’s advocate in places where further discussion seems to be required. Given the venue for this guide, we can safely restrict the discussion to embodied artificial intelligence (EAI), even if such work draws on notions of embodied cognition.. (shrink)

This article addresses a classical question: Can a machine use language meaningfully and if so, how can this be achieved? The first part of the paper is mainly philosophical. Since meaning implies intentionality on the part of the language user, artificial systems which obviously lack intentionality will be `meaningless'. There is, however, no good reason to assume that intentionality is an exclusively biological property and thus a robot with bodily structures, interaction patterns and development similar to those of human beings (...) would constitute a system possibly capable of meaning – a conjecture supported through a Wittgenstein-inspired thought experiment. The second part of the paper focuses on the empirical and constructive questions. Departing from the principle of epigenesis stating that during every state of development new structure arises on the basis of existing structure plus various sorts of interaction, a model of human cognitive and linguistic development is proposed according to which physical, social and linguistic interactions between the individual and the environment have their respective peaks in three consecutive stages of development: episodic, mimetic and symbolic. The transitions between these stages are qualitative, and bear a similarity to the stages in phylogenesis proposed by Donald and Deacon. Following the principle of epigenetic development, robotogenesis could possibly recapitulate ontogenesis, leading to the emergence of intentionality, consciousness and meaning. (shrink)

When the various disciplines participating in cognitive science are listed, philosophy almost always gets a guernsey. Yet, a couple of years ago at the conference of the Cognitive Science Society in Boulder (USA), there was no philosophy or philosopher with any prominence on the program. When queried on this point, the organizer (one of the "superstars" of the field) claimed it was partly an accident, but partly also due to an impression among members of the committee that philosophy is basically (...) a waste of time. Philosophy, they thought, is mostly obscure bullshit that does little to help, and much to hinder, real progress in cognitive science. (shrink)

It is argued that, based on Kant's descriptive metaphysics, one can prescribe the necessary metaphysical underpinnings for the possibility of conscious experience in an artificial system. This project is developed by giving an account of the a priori concepts of the understanding in such a system. A specification and implementation of the nomological conditions for a conscious system allows one to know a priori that any system possessing this structure will be conscious; thus enabling us to avoid possible false-indicators of (...) consciousness like that offered in a behaviouristic analysis. This is an alternative approach to the bottom-up or top-down approaches adopted by, for example CYC (Lenat and Feigenbaum 1992) and COG (Brooks 1994; Brooks and Stein 1993), neither of which, alone, or in some hybrid form, have proved productive. (shrink)

This essay continues my investigation of `syntactic semantics': the theory that, pace Searle's Chinese-Room Argument, syntax does suffice for semantics (in particular, for the semantics needed for a computational cognitive theory of natural-language understanding). Here, I argue that syntactic semantics (which is internal and first-person) is what has been called a conceptual-role semantics: The meaning of any expression is the role that it plays in the complete system of expressions. Such a `narrow', conceptual-role semantics is the appropriate sort of semantics (...) to account (from an `internal', or first-person perspective) for how a cognitive agent understands language. Some have argued for the primacy of external, or `wide', semantics, while others have argued for a two-factor analysis. But, although two factors can be specified–-one internal and first-person, the other only specifiable in an external, third-person way–-only the internal, first-person one is needed for understanding how someone understands. A truth-conditional semantics can still be provided, but only from a third-person perspective. (shrink)

Hauser argues that his pocket calculator (Cal) has certain arithmetical abilities: it seems Cal calculates. That calculating is thinking seems equally untendentious. Yet these two claims together provide premises for a seemingly valid syllogism whose conclusion - Cal thinks - most would deny. He considers several ways to avoid this conclusion, and finds them mostly wanting. Either we ourselves can't be said to think or calculate if our calculation-like performances are judged by the standards proposed to rule out Cal; or (...) the standards- e.g., autonomy and self-consciousness- make it impossible to verify whether anything or anyone (save oneself) meets them. While appeals to the intentionality of thought or the unity of minds provide more credible lines of resistance, available accounts of intentionality and mental unity are insufficiently clear and warranted to provide very substantial arguments against Cal's title to be called a thinking thing. Indeed, considerations favoring granting that title are more formidable than is generally appreciated. Rapaport's comments suggest that, on a strong view of thinking, mere calculating is not thinking (and pocket calculators don't think), but on a weak, but unexciting, sense of thinking, pocket calculators do think. He closes with some observations on the implications of this conclusion. (shrink)

Eliminative materialism is a popular view of the mind which holds that propositional attitudes, the typical units of our traditional understanding, are unsupported by modern connectionist psychology and neuroscience, and consequently that propositional attitudes are a poor scientific postulate, and do not exist. Since our traditional folk psychology employs propositional attitudes, the usual argument runs, it too represents a poor theory, and may in the future be replaced by a more successful neurologically grounded theory, resulting in a drastic improvement in (...) our interpersonal relationships. I contend that these eliminativist arguments typically run together two distinct capacities: the folk psychological mechanisms which we use to understand one another, and scientific and philosophical guesses about the structure of those understandings. Both capacities are ontologically committed and therefore empirical. However, the commitments whose prospects look so dismal to the eliminativist, in particular the causal and logical image of propositional attitudes, belong to the guesses, and not necessarily to the underlying mechanisms. It is the commitments of traditional philosophical perspectives about the operation of our folk psychology which are contradicted by␣new evidence and modeling methods in connectionist psychology. Our actual folk psychology was not clearly committed to causal, sentential propositional attitudes, and thus is not directly threatened by connectionist psychology. (shrink)

The objective of research in the foundations of Al is to explore such basic questions as: What is a theory in Al? What are the most abstract assumptions underlying the competing visions of intelligence? What are the basic arguments for and against each assumption? In this essay I discuss five foundational issues: (1) Core Al is the study of conceptualization and should begin with knowledge level theories. (2) Cognition can be studied as a disembodied process without solving the symbol grounding (...) problem. (3) Cognition is nicely described in propositional terms. (4) We can study cognition separately from learning. (5) There is a single architecture underlying virtually all cognition. I explain what each of these implies and present arguments from both outside and inside Al why each has been seen as right or wrong. (shrink)

Experts in medical informatics have argued for the incorporation of ever more machine-learning algorithms into medical care. As artificial intelligence research advances, such technologies raise the possibility of an “iDoctor,” a machine theoretically capable of replacing the judgment of primary care physicians. In this article, I draw on Martin Heidegger’s critique of technology to show how an algorithmic approach to medicine distorts the physician–patient relationship. Among other problems, AI cannot adapt guidelines according to the individual patient’s needs. In response to (...) the objection that AI could develop this capacity, I use Hubert Dreyfus’s analysis of AI to argue that attention to the needs of each patient requires the physician to attune his or her perception to the patient’s history and physical exam, an ability that seems uniquely human. Human physician judgment will remain better suited to the practice of primary care despite anticipated advances in AI technology. (shrink)

Economists tend to represent learning as a procedure for estimating the parameters of the "correct" econometric model. We extend this approach by assuming that agents specify as well as estimate models. Learning thus takes the form of a dynamic process of developing models using an internal language of representation where expectations are formed by forecasting with the best current model. This introduces a distinction between the form and content of the internal models which is particularly relevant for boundedly rational agents. (...) We propose a framework for such model development which use a combination of measures: the error with respect to past data, the complexity of the model, the cost of finding the model and a measure of the model's specificity The agent has to make various trade-offs between them. A utility learning agent is given as an example. (shrink)

In this thesis I present a virtual laboratory which implements five different models for controlling animats: a rule-based system, a behaviour-based system, a concept-based system, a neural network, and a Braitenberg architecture. Through different experiments, I compare the performance of the models and conclude that there is no best model, since different models are better for different things in different contexts. The models I chose, although quite simple, represent different approaches for studying cognition. Using the results as an empirical philosophical (...) aid, I note that there is no best approach for studying cognition, since different approaches have all advantages and disadvantages, because they study different aspects of cognition from different contexts. This has implications for current debates on proper approaches for cognition: all approaches are a bit proper, but none will be proper enough. I draw remarks on the notion of cognition abstracting from all the approaches used to study it, and propose a simple classification for different types of cognition. (shrink)