By Sari Pietikäinen, Alexandra Jaffe, Helen Kelly-Holmes, Nik Coupland
Sociolinguistics from the Periphery "presents a fascinating book about change: shifting political, economic and cultural conditions; ephemeral, sometimes even seasonal, multilingualism; and altered imaginaries for minority and indigenous languages and their users"
Review of The Mind Doesn't Work That Way: The Scope and Limits of Computational Psychology
Date: Wed, 11 Sep 2002 03:49:12 -0400 From: Kornel R. BANGHA Subject: Psycholinguistics; Fodor, Jerry A. (2001) The Mind Doesn't Work That Way
Fodor, Jerry A. (2001) The Mind Doesn't Work That Way: The Scope and Limits of Computational Psychology, Cambridge, Massachusetts, The MIT Press
Reviewed by: Kornel Bangha, University of Montreal
This book is about what could be called the New Synthesis: the Computational Theory of Mind combined with nativism and (neo-)Darwinism. It is a response to Pinker (1997) and Plotkin (1997) who defend the New Synthesis. Fodor's aim is to show that what our cognitive science has found out about the mind so far is mostly that we don't know how it works.
Fodor states that CTM is by far the best theory of cognition that we've got; indeed, the only one we've got that's worth the bother of a serious discussion. There are facts about the mind that it accounts for and we would be utterly at a loss to explain without it; and its central idea - that intentional processes are syntactic operations defined on mental representations - is strikingly elegant. There is, in short, every reason to suppose that CTM is part of the truth about cognition. However, there is no reason to think that this is the whole story about how the mind works or even a large part of the truth.
In the first chapter, Fodor begins with making a distinction between Chomsky's story about innateness and the New Synthesis. The author states that the present phase of nativistic theorising about the cognitive mind began with two suggestions of Noam Chomsky's: that there are substantive, universal constraints on the kind of grammars that natural language can have; and that these constrains express correspondingly substantive and universal properties of human psychology (determined, presumably, by the characteristic genetic endowment of our species). The central problem of language acquisition arises from the poverty of the "primary linguistic data" from which the child construct a theory of the language; and the proposed solution of the problem is that much of the knowledge that linguistic competence depends on is available to the child a priori (i.e., prior to learning).
Chomsky's rationalism consist primarily in nativism about the knowledge that cognitive capacities manifest while New Synthesis rationalism consists primarily in nativism about the computational mechanisms that exploit such a knowledge for the purposes of cognition. The New Synthesis shares with traditional rationalism its emphasis on innate content; but it has added Turing's idea that mental architecture is computational.
Computation starts, for instance, with the remarkable fact that you can tell, just by looking at it, that any (declarative) sentence of the syntactic form P and Q is true if and only if P and Q are themselves both true: you don't have to know anything about the nonlinguistic word. This really is remarkable since it's what they mean, together with fact about the non-linguistic world, that decide whether P or Q are true. This line of thought is often summarized by saying that some inferences are "formally valid", which is in turn to say that they hold just in virtue of the "syntax" of the sentences that enter into them. It was Turing's great discovery that machines can be designed to evaluate any inference that is formally valid in that sense. The basic thesis of the new psychological synthesis is that cognitive mental processes are (perhaps exhaustively) constituted by the kinds of operations that such machines perform.
Fodor considers that rationalists are nativists practically by definition. The main idea of rationalist psychology is that beliefs, desires, thoughts, and the like have logical forms, and that their logical forms are among the determinants of the roles they play in mental process. What connects rationalist psychologies and the thesis that mental processes are computations is primarily the idea that the logical form of a thought might be reconstructed by the syntax of a mental representation that expresses it. A rationalist psychology says that typical laws about the mind specify ways in which the logical form of a mental state determines its role in mental processes. So a rationalist is in need of a theory about how a mental process could be sensitive to the logical form of mental states. Turing's notion of computation provides exactly what a rationalist cognitive scientist needs to fill this gap.
In the second chapter, Fodor argues that there are some very deep problems with viewing cognition as computational, but these problems emerge primarily in respect of mental processes that aren't modular. Indeed, Turing's idea that mental processes are computations (i.e., that they are syntactically driven), together with Chomsky's idea that poverty of the stimulus arguments set a lower bound to the information a mind must have innately, are half of the New Synthesis. The rest is the "massive modularity" thesis and the claim that cognitive architecture is a Darwinian adaptation.
These problems are presented as follows. Mental processes are sensitive solely to the syntax of mental representations (because mental processes are computations). Syntactic properties of mental representations are essential (because the syntactic properties of any representation are essential). Conclusion: Mental processes are insensitive to context dependent properties of mental representation. And this is where the trouble starts. For it would seem that, as a matter of fact, this conclusion isn't true. Fodor proposes some counter-examples: simplicity, abductive inference and conservativism.
Simplicity is the first example of a context-dependent property of mental representation to which cognitive processes are responsive. The complexity of a thought is not intrinsic; it depends on the context. But the syntax of a representation is one of its essential properties and so doesn't change when the representation is transported from one context to another. So how could the simplicity of a thought supervene on its syntax, as CTM requires it?
The author argues that abductive inferences could be computations only at the price of a ruinous holism; that is, by assuming that the units of thoughts are much bigger than in fact they could possibly be.
Finally, estimates of which beliefs count for a lot and which ones count for a little when one is reckoning the conservativism of a theory change have to be context sensitive. But the syntactic properties of representations aren't theory sensitive and can't change with the context.
The third chapter considers some ways in which computational nativists have tried to evade the limitations mentioned in chapter 2. Fodor believes that abduction is a terrible problem for cognitive science, one that is unlikely to be solved by any kind of theory we have heard of so far. However, cognitive scientists might hope two kids of solution. Some psychologists think that even if they are unable to model the global determination of ideally rational inference, they can produce heuristic approximations good enough to account for the cognitive capacities that people actually have. Other psychologists often prefer a connectionist model of cognitive architecture, which they think has no principled difficulty with holistic effects in cognition.
The main problem with heuristics is simply that it doesn't work. This is because reliable abduction may require that the whole background of epistemic commitments be somehow brought to bear in planning and belief fixation. But feasible abduction requires, in practice, that no more than a small subset of even the relevant background beliefs is actually considered. How to make abductive inferences that are both reliable and feasible is what they call in AI the frame probleme. The failure of AI is considered by the author as the failure of the Classical CTM to perform well in practice.
Fodor argues that there is something fundamentally wrong with connectionist networks. Since nodes in different networks, or at different positions at the same network, are ipso facto different types of nodes, it follows that its position in its network is among a node's essential features and that nodes can't be "transported" from one network to another. Thus, network architectures haven't any way to say that representations can have recurrent parts; for example, that "John loves Mary" and "Mary loves John" do.
Fodor presents in the fourth chapter what modularity is as well as what his main problem with it is. The author calls the idea that most or all of cognition is modular the "massive modularity" thesis (MM) and considers that the likelihood of New Synthesis Psychology will turn out to be a reasonably general theory of the cognitive mind is hostage to MM. Classical computations are sensitive, at most, to the local context; and so too are the computations that modular mechanisms perform. Modular cognition is the kind of processing of which the Classical computation story is the most likely to be true.
It is sometime claimed that there are very general, adaptationist considerations that militate in favour of massively modular cognitive architecture over domain-general architectures, or "mixed" ones that acknowledge computational mechanism of both kinds. According to Cosmides and Tooby there are three reasons why it's "impossible in principle" that the human mind consist of nothing but domain-general mechanisms. The first is that definition of error is domain-dependent. The second is the poverty of stimulus. The third reason is that combinatorial explosion paralyzes any system that is truly domain-general.
Fodor examines each of these three arguments and tells us why he doesn't consider them convincing. At the end of the chapter, he presents an argument against Massive Modularity: The Input Problem (i.e. the problem of identifying representations in its proprietary domain). Really massive modularity is a coherent account of cognitive architecture only if the input problem for each module can be solved by inferences that aren't abductive (or otherwise holistic); that is, by domain-specific mechanisms. There isn't, however, any reason to think that it can.
The New Synthesis is widely commited to the thesis that "cognitive architecture is an evolutionary adaptation" and one might wonder how this claim fits with the other two. That's what the last chapter is about.
Fodor offers a disapproving survey of the main standard arguments for adaptationism about cognition: consistency, teleology and complexity. Consistency claims that psychology should be constrained by the theory of evolution since valid scientific knowledge - whether from the same or different field - should be mutually consistent. Teleology argues that functional explication is essential in the biological sciences and in cognitive science too. The argument of complexity says that there is no way except evolutionary selection for Nature to build a complex mind.
At the end of the chapter, Fodor tells us why he thinks there is an intrinsic connection between adaptationism and the particular kind of cognitive nativism that New Synthesis psychologists endorse.
The book is very small (about 100 pages) but the reading is far from being easy. There is little in it about linguistics and much about cognitive sciences. Indeed, it concerns several fields: psychology, logics, Artificial Intelligence, linguistics, and even biology, philosophy, etc. I would consider it as a merit and a defect at the same time. It is good to have a large view, and most probably even necessary for Fodor's topics, but it might be quite hard to follow for people less familiar with cognitive psychology.
I believe that once you understand the points Fodor writes about, you get also convinced. Abductive inference remains a problem, not only in Classical CTM but also in the New Synthesis. This leads us to the author's conclusion that we don't know much about the mind.
The only point which remains unclear for me in the book is the following: Why does Fodor think that Plato would have understood Chomsky well enough, and Turing not at all?
Pinker, S. (1997) How the Mind Works. New York: Norton Plotkin, H. (1997) Evolution in Mind. London: Alan Lane Tooby, J. and Cosmides, L. (1992) The Adapted Mind. Oxford: Oxford University Press ?
ABOUT THE REVIEWER:
About the reviewer: Kornel Bangha prepares a Ph. D. of Linguistics and Artificial Intelligence at the University of Montreal. His research is about how the process of interpretation of linguistic units in discourse is influenced by lexical knowledge, by context and by knowledge about the world. µ