Stevan Harnad

According to "computationalism" (Newell, 1980; Pylyshyn 1984; Dietrich 1990), mental states are computational states, so if one wishes to build a mind, one is actually looking for the right program to run on a digital computer. A computer program is a semantically interpretable formal symbol system consisting of rules for manipulating symbols on the basis of their shapes, which are arbitrary in relation to what they can be systematically interpreted as meaning. According to computationalism, every physical implementation of the right symbol system will have mental states.

Steels & Belpaeme's (S&B's) simulations contain all the right components, but they are put together wrongly. Color categories are unrepresentative of categories in general and language is not merely naming. Language evolved because it provided a powerful new way to acquire categories (through instruction, rather than just the old way of other species, through trial-and-error experience). It did not evolve so that multiple agents looking at the same objects could let one another know which of the objects they had in mind, co-coining names for them on the fly.

1.1 The predominant approach to cognitive modeling is still what has come to be called "computationalism" (Dietrich 1990, Harnad 1990b), the hypothesis that cognition is computation. The more recent rival approach is "connectionism" (Hanson & Burr 1990, McClelland & Rumelhart 1986), the hypothesis that cognition is a dynamic pattern of connections and activations in a "neural net." Are computationalism and connectionism really deeply different from one another, and if so, should they compete for cognitive hegemony, or should they collaborate? These questions will be addressed here, in the context of an obstacle that is faced by computationalism (as well as by connectionism if it is either computational or seeks cognitive hegemony on its own): The symbol grounding problem (Harnad 1990).

The problem seems apparent even in Glasgow's term ``depict'', which is used by way of contrast with ``describe''. Now ``describe'' refers relatively unproblematically to strings of symbols, such as those in this written sentence, that are systematically interpretable as propositions describing objects, events, or states of affairs. But what does ``depict'' mean? In the case of a picture -- whether a photo or a diagram -- it is clear what depict means. A picture is an object (I will argue below that it is an analog object, relative to what it is a picture of) and it DEPICTS yet another object: the object it is a picture OF. But in the case of an array, whether described formally, with numerical coordinates, or stored in a machine, or ``depicted'' diagrammatically by way of a secondary illustration, it is not at all clear whether the entity in question is indeed a picture, or merely yet another set of symbols that is INTERPRETABLE as referring to a picture, which picture in turn depicts an object! It is clear that we are dealing with many layers of interpretation here already, and so far we are still talking only about external objects (such as pictures, symbols and objects simpliciter). We still have not gotten to MENTAL objects, such as mental ``images''

Peer Review and Copyright each have a double role: Formal refereeing protects (R1) the author from publishing and (R2) the reader from reading papers that are not of sufficient quality. Copyright protects the author from (C1) theft of text and (C2) theft of authorship. It has been suggested that in the electronic medium we can dispense with peer review, "publish" everything, and let browsing and commentary do the quality control. It has also been suggested that special safeguards and laws may be needed to enforce copyright on the Net. I will argue, based on 20 years of editing Behavioral and Brain Sciences, a refereed (paper) journal of peer commentary, 8 years of editing Psycoloquy, a refereed electronic journal of peer commentary, and 1 year of implementing CogPrints, an electronic archive of unrefereed preprints and refereed reprints in the cognitive sciences modeled on the Los Alamos Physics Eprint Archive, that (i) peer commentary is a supplement, not a substitute, for peer review, (ii) the authors of refereed papers, who get and seek no royalties from the sale of their texts, only want protection from theft of authorship on the Net, not from theft of text, which is a victimless crime, and hence (iii) the trade model (subscription, site license or pay- per-view) should be replaced by author page-charges to cover the much reduced cost of implementing peer review, editing and archiving on the Net, in exchange for making the learned serial corpus available for free for all forever.

Let us simplify the problem of “consciousness” or “visual consciousness”: Seeing is feeling. The difference between an optical transducer/effector that merely interacts with optical input, and a conscious system that sees, is that there is something it feels like for that conscious system to see, and that system feels that feeling. All talk about “internal representations” and internal or external difference registration or detection, and so on, is beside the point. The point is that what is seen is felt, not merely registered, processed, and acted upon. To explain consciousness in terms of sensorimotor action, one has to explain why and how any of that processing is felt; otherwise one is merely giving an optokinetic explanation of I/O (Input/Ouput) capacities (and of whether those capacities are actually or optimally generated by sensorimotor contingency processors, analog representations, symbolic representations, or other forms of internal structure/process), not of the fact that they are felt. Nor will it do to say “qualia are illusions.” Qualia are feelings. Am I under the illusion that I am seeing (i.e., feeling) something right now? What is the truth then? That I am not feeling, but merely acting? No, I'm afraid Descartes had it right. Certain things are not open to doubt. They either need to be explained, or passed over in silence, in favor of the unfelt correlated functions that we can explain (Harnad 1995; 2000; 2001).

Cognitive science is a form of "reverse engineering" (as Dennett has dubbed it). We are trying to explain the mind by building (or explaining the functional principles of) systems that have minds. A "Turing" hierarchy of empirical constraints can be applied to this task, from t1, toy models that capture only an arbitrary fragment of our performance capacity, to T2, the standard "pen-pal" Turing Test (total symbolic capacity), to T3, the Total Turing Test (total symbolic plus robotic capacity), to T4 (T3 plus internal [neuromolecular] indistinguishability). All scientific theories are underdetermined by data. What is the right level of empirical constraint for cognitive theory? I will argue that T2 is underconstrained (because of the Symbol Grounding Problem and Searle's Chinese Room Argument) and that T4 is overconstrained (because we don't know what neural data, if any, are relevant). T3 is the level at which we solve the "other minds" problem in everyday life, the one at which evolution operates (the Blind Watchmaker is no mind-reader either) and the one at which symbol systems can be grounded in the robotic capacity to name and manipulate the objects their symbols are about. I will illustrate this with a toy model for an important component of T3 -- categorization -- using neural nets that learn category invariance by "warping" similarity space the way it is warped in human categorical perception: within-category similarities are amplified and between-category similarities are attenuated. This analog "shape" constraint is the grounding inherited by the arbitrarily shaped symbol that names the category and by all the symbol combinations it enters into. No matter how tightly one constrains any such model, however, it will always be more underdetermined than normal scientific and engineering theory. This will remain the ineliminable legacy of the mind/body problem

Harnad accepts the picture of computation as formalism, so that any implementation of a program - thats any implementation - is as good as any other; in fact, in considering claims about the properties of computations, the nature of the implementing system - the interpreter - is invisible. Let me refer to this idea as 'Computationalism'. Almost all the criticism, claimed refutation by Searle's argument, and sharp contrasting of this idea with others, rests on the absoluteness of this separation between a computational system and its implementation.

Cognition is thinking; it feels like something to think, and only those who can feel can think. There are also things that thinkers can do. We know neither how thinkers can think nor how they are able to do what they can do. We are waiting for cognitive science to discover how. Cognitive science does this by testing hypotheses about what processes can generate what doing.This is called the Turing Test. It cannot test whether a process can generate feeling, hence thinking — only whether it can generate doing. The processes that generate thinking and know-how are “distributed” within the heads of thinkers, but not across thinkers’ heads. Hence there is no such thing as distributed cognition, only collaborative cognition. Email and the Web have spawned a new form of collaborative cognition that draws upon individual brains’ real-time interactive potential in ways that were not possible in oral, written or print interactions.

We utilize higher order automated deduction technologies for the logical analysis of natural-language arguments. Our approach, termed computational hermeneutics, is grounded on recent progress in the area of automated theorem proving for classical and nonclassical higher order logics, and it integrates techniques from argumentation theory. It has been inspired by ideas in the philosophy of language, especially semantic holism and Donald Davidson’s radical interpretation; a systematic approach to interpretation that does justice to the inherent circularity of understanding: the whole is understood compositionally on the basis of its parts, while each part is understood only in the context of the whole (hermeneutic circle). Computational hermeneutics is a holistic, iterative approach where we evaluate the adequacy of some candidate formalization of a sentence by computing the logical validity of (i) the whole argument it appears in and (ii) the dialectic role the argument plays in some piece of discourse.

Darwin differs from Newton and Einstein in that his ideas do not require a complicated or deep mind to understand them, and perhaps did not even require such a mind in order to generate them in the first place. It can be explained to any school-child (as Newtonian mechanics and Einsteinian relativity cannot) that living creatures are just Darwinian survival/reproduction machines. They have whatever structure they have through a combination of chance and its consequences: Chance causes changes in the genetic blueprint from which organisms' bodies are built, and if those changes are more successful in helping their owners survive and reproduce than their predecessors or their rivals, then, by definition, those changes are reproduced, and thereby become more prevalent in succeeding generations: Whatever survives/reproduces better survives/reproduces better. That is the tautological force that shaped us

Some of the features of animal and human categorical perception (CP) for color, pitch and speech are exhibited by neural net simulations of CP with one-dimensional inputs: When a backprop net is trained to discriminate and then categorize a set of stimuli, the second task is accomplished by "warping" the similarity space (compressing within-category distances and expanding between-category distances). This natural side-effect also occurs in humans and animals. Such CP categories, consisting of named, bounded regions of similarity space, may be the ground level out of which higher-order categories are constructed; nets are one possible candidate for the mechanism that learns the sensorimotor invariants that connect arbitrary names (elementary symbols?) to the nonarbitrary shapes of objects. This paper examines how and why such compression/expansion effects occur in neural nets.

  Computation is interpretable symbol manipulation. Symbols are objects that are manipulated on the basis of rules operating only on theirshapes, which are arbitrary in relation to what they can be interpreted as meaning. Even if one accepts the Church/Turing Thesis that computation is unique, universal and very near omnipotent, not everything is a computer, because not everything can be given a systematic interpretation; and certainly everything can''t be givenevery systematic interpretation. But even after computers and computation have been successfully distinguished from other kinds of things, mental states will not just be the implementations of the right symbol systems, because of the symbol grounding problem: The interpretation of a symbol system is not intrinsic to the system; it is projected onto it by the interpreter. This is not true of our thoughts. We must accordingly be more than just computers. My guess is that the meanings of our symbols are grounded in the substrate of our robotic capacity to interact with that real world of objects, events and states of affairs that our symbols are systematically interpretable as being about

Cognition is thinking; it feels like something to think, and only those who can feel can think. There are also things that thinkers can do. We know neither how thinkers can think nor how they are able do what they can do. We are waiting for cognitive science to discover how. Cognitive science does this by testing hypotheses about what processes can generate what doing (“know-how”) This is called the Turing Test. It cannot test whether a process can generate feeling, hence thinking -- only whether it can generate doing. The processes that generate thinking and know-how are “distributed” within the heads of thinkers, but not across thinkers’ heads. Hence there is no such thing as distributed cognition, only collaborative cognition. Email and the Web have spawned a new form of collaborative cognition that draws upon individual brains’ real-time interactive potential in ways that were not possible in oral, written or print interactions.

Creativity may be a trait, a state or just a process defined by its products. It can be contrasted with certain cognitive activities that are not ordinarily creative, such as problem solving, deduction, induction, learning, imitation, trial and error, heuristics and "abduction," however, all of these can be done creatively too. There are four kinds of theories, attributing creativity respectively to (1) method, (2) "memory" (innate structure), (3) magic or (4) mutation. These theories variously emphasize the role of an unconscious mind, innate constraints, analogy, aesthetics, anomalies, formal constraints, serendipity, mental analogs, heuristic strategies, improvisatory performance and cumulative collaboration. There is some virtue in each, but the best model is still the one implicit in Pasteur's dictum: "Chance favors the prepared mind." And because the exercise and even the definition of creativity requires constraints, it is unlikely that "creativity training" or an emphasis on freedom in education can play a productive role in this preparation.

Were we just the Darwinian adaptive survival/reproduction machines von Hippel & Trivers invoke to explain us, the self-deception problem would not only be simpler, but also nonexistent. Why would unconscious robots bother to misinform themselves so as to misinform others more effectively? But as we are indeed conscious rather than unconscious robots, the problem is explaining the causal role of consciousness itself, not just its supererogatory tendency to misinform itself so as to misinform (or perform) better

A provisional model is presented in which categorical perception (CP) provides our basic or elementary categories. In acquiring a category we learn to label or identify positive and negative instances from a sample of confusable alternatives. Two kinds of internal representation are built up in this learning by "acquaintance": (1) an iconic representation that subserves our similarity judgments and (2) an analog/digital feature-filter that picks out the invariant information allowing us to categorize the instances correctly. This second, categorical representation is associated with the category name. Category names then serve as the atomic symbols for a third representational system, the (3) symbolic representations that underlie language and that make it possible for us to learn by "description." Connectionism is one possible mechainsm for learning the sensory invariants underlying categorization and naming. Among the implications of the model are (a) the "cognitive identity of (current) indiscriminables": Categories and their representations can only be provisional and approximate, relative to the alternatives encountered to date, rather than "exact." There is also (b) no such thing as an absolute "feature," only those features that are invariant within a particular context of confusable alternatives. Contrary to prevailing "prototype" views, however, (c) such provisionally invariant features must underlie successful categorization, and must be "sufficient" (at least in the "satisficing" sense) to subserve reliable performance with all-or-none, bounded categories, as in CP. Finally, the model brings out some basic limitations of the "symbol-manipulative" approach to modeling cognition, showing how (d) symbol meanings must be functionally grounded in nonsymbolic, "shape-preserving" representations -- iconic and categorical ones. Otherwise, all symbol interpretations are ungrounded and indeterminate. This amounts to a principled call for a psychophysical (rather than a neural) "bottom-up" approach to cognition.

Connectionism and computationalism are currently vying for hegemony in cognitive modeling. At first glance the opposition seems incoherent, because connectionism is itself computational, but the form of computationalism that has been the prime candidate for encoding the "language of thought" has been symbolic computationalism (Dietrich 1990, Fodor 1975, Harnad 1990c; Newell 1980; Pylyshyn 1984), whereas connectionism is nonsymbolic (Fodor & Pylyshyn 1988, or, as some have hopefully dubbed it, "subsymbolic" Smolensky 1988). This paper will examine what is and is not a symbol system. A hybrid nonsymbolic/symbolic system will be sketched in which the meanings of the symbols are grounded bottom-up in the system's capacity to discriminate and identify the objects they refer to. Neural nets are one possible mechanism for learning the invariants in the analog sensory projection on which successful categorization is based. "Categorical perception" (Harnad 1987a), in which similarity space is "warped" in the service of categorization, turns out to be exhibited by both people and nets, and may mediate the constraints exerted by the analog world of objects on the formal world of symbols

The experimental analysis of naming behavior can tell us exactly the kinds of things Horne & Lowe (H & L) report here: (1) the conditions under which people and animals succeed or fail in naming things and (2) the conditions under which bidirectional associations are formed between inputs (objects, pictures of objects, seen or heard names of objects) and outputs (spoken names of objects, multimodal operations on objects). The "stimulus equivalence" that H & L single out is really just the reflexive, symmetric and transitive property of pairwise associations among the above. This is real and of some interest, but it unfortunately casts very little light on symbolization and language in general, and naming capacity in particular. The associative equivalence between name and object is trivial in relation to the real question, which is: How do we (or any system that can do it) manage to connect names to things correctly (Harnad 1987, 1990, 1992)? The experimental analysis of naming behavior begs this question entirely, simply taking it for granted that the connection is somehow successfully accomplished.

Do scientists agree? It is not only unrealistic to suppose that they do, but probably just as unrealistic to think that they ought to. Agreement is for what is already established scientific history. The current and vital ongoing aspect of science consists of an active and often heated interaction of data, ideas and minds, in a process one might call "creative disagreement." The "scientific method" is largely derived from a reconstruction based on selective hindsight. What actually goes on has much less the flavor of a systematic method than of trial and error, conjecture, chance, competition and even dialectic.