Being Where? Andy Clark and the Problem of Advanced Cognition1
    Michael Meadon
    School of Philosophy and Ethics University of KwaZulu-Natal Durban, 4041 Michael.meadom@gmail.com Abstract Andy Clark (1989, 1993, 1997) is a leading philosophical exponent of a view of mind as an ‘associative engine’, or connectionist pattern-completer, composed of multiple special-purpose modules that communicate in only limited ways and eschew detailed forms of internal representation. The modules, Clark and his allies argue, are both coordinated and integrated by the environment, whilst ‘off-loading’ onto it by calling on external computational resources (‘external scaffolds’) to reduce cognitive load. Defenders of this position further maintain that even examples of sophisticated and distinctively human cognition, such as long-term planning or running a multi-national company, emerge from connectionist pattern-completing brains in the ‘constraining presence of public language, culture and institutions’ (1997: 33). This constellation of ideas, Clark argues, amounts to a completely new science of mind that radically reforms ‘our whole way of thinking about intelligent behaviour’. Unfortunately, this rhetoric far outstrips the evidence: while a reasonable case can be made that external scaffolds are necessary for many types of cognition, the assertion that pattern-completion plus external scaffolding is a sufficient explanation of all human cognition has not been demonstrated. The insufficiency of the Clarkian view is particularly evident in the case of advanced cognition in the economic sphere.
    
    Andy Clark’s Being There: Putting Brain, Body and World Together Again is a rich, wide-ranging and thought-provoking tour through several important recent developments in cognitive science. The aim of the book, according to Clark, is to ‘display and examine’ (1997: xviii) and to ‘trace some of [the] origins, display [the] flavor, and confront some of [the] problems’ (1997: xiii) of a new view of mind as a ‘controller of embodied action’ (1997: 7) and thus ‘inextricably interwoven with body, world and action’ (1997: xvii). This view, Clark insists, has (at the time of writing) matured into a ‘more rounded, compelling and integrative’ (1997: xviii) account, promising a
    1 I thank David Spurrett and John McCoy for comments and corrections on this paper. © 2008 Michael Meadon; licensee South African Journal of Philosophy. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.5), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
    
    S. Afr. J. Philos. 2008, 27(2)
    
    121
    
    ‘sweeping reform [of] our whole way of thinking about intelligent behaviour’ and constituted ‘nothing less that a new science of mind’ (1997: xii-xiii). The model this new view was meant to overthrow is what Clark elsewhere (1989: 9-24) calls ‘classical cognitivism’: the ‘disembodied, atemporal intellectualist vision of mind’ (1997: 7) involving a physical symbol system à la Newell and Simon (1976). In Being There, the CYC project (Lenat and Feigenbaum 1992) – which aimed to handcraft a massive database of commonsense knowledge, thereby instilling ‘genuine intelligence’ in a computer – is classical cognitivism’s representative, which Clark wholly condemns as ‘absolutely, fundamentally, and fatally flawed’ (1997: 4).2 The book is not, however, a sustained critique of classical cognitivism; it is an introduction, exposition and exploration of Clark’s favoured position. The core of Clark’s view is that the brain is an ‘associative engine’ (1997: 53; 1993: 189) or connectionist pattern-completer, composed of ‘multiple special-purpose problem solvers’ (1997: 14) utilising only ‘limited communication [and] exploiting partial and action-oriented forms of internal representation’ (1997: 190-191). The internal systems are ‘orchestrated by environmental inputs’ (1997: 14) and ‘lean on the environment’ (1997.: 59) by calling on external ‘computational and informational’ resources to ‘transform inputs, to simplify search, to aid recognition, to prompt associative recall [and] to offload memory’ (1997: 68). Finally, even ‘the most advanced and distinctive aspects of human cognition’ (1997: 179) emerge from our connectionist pattern-completing brain in the ‘constraining presence of public language, culture and institutions’ (1997: 33). To rephrase, Clark claims that: (1) the mind’s on-line capabilities are restricted to simple iterated pattern-completion (as in connectionist models), (2) the environment in the form of scaffolding plays a computational role and induces overall behavioural coherence, (3) the brain consists of multiple, independent, special-purpose and parallel computational processes, (4) internal representation is both partial and action-orientated, 3 and (5) advanced cognition can be accounted for in terms of pattern-completing computation and external resources or scaffolds. It is important to reiterate that the five propositions above are – or seem to be – the main positive claims of Being There; they do not exhaust the list of issues Clark takes a stand on. Furthermore, there will not be enough space in this paper to cover nearly as much ground as Clark does in Being There – the focus will be on Proposition 5, and specifically on Clark’s argument that advanced cognition in the economic sphere can adequately be accounted for in terms of his view of mind.
    2 This is not to suggest that Clark thinks either that a refutation of CYC is a sufficient reason to conclude that classical cognitivism as a whole is flawed, or that classical cognitivism generally is as deeply flawed as CYC. Clark bills CYC as ‘an extreme example of the opposite view’ (1997: 2); not as a representative case suitable for generalisation. See Clark (1989: 12-13, 127-136 & passim) for an argument against a specific version of the physical symbol system hypothesis. This is not quite right because it implies that Clark thinks internal representations are necessary for human cognition, whereas he rather seems to be agnostic about the issue. The point is that, if there are representations, they better pay their way by serving some computational function. As Clark puts it: ‘[t]o the extent that the biological brain does trade in anything usefully described as “internal representation,” a large body of those representations will be local and action-orientated rather than objective and action-independent.’ (1997: 149; emphasis added).
    
    3
    
    122
    
    S. Afr. J. Philos. 2008, 27(2)
    
    In Section I of Being There, Clark argues that the brain is a pattern-completion device composed of multiple, only partially communicating, neural networks. Such a model may seem inadequate to account for several types of cognition, but, Clark claims, external scaffolds come to the ‘naked’ brain’s rescue by simplifying problem-solving tasks,thereby lightening the computational load substantially. While Clark presents several examples (scrabble, arithmetic, Tetris, etc.) of how these scaffolds allow an associative engine to achieve relatively simple tasks that would otherwise be beyond it, the ‘pressing problem of advanced cognition’ looms: can this approach explain sophisticated cogitations such as long-term planning, deciding how to vote, running a country, and so on? (1997: 82, 179). Section III of Being There is an attempt to show that the approach can indeed explain advanced cognition: Ch. 9 (Minds and Markets) takes on economic behaviours, while Ch. 10 (Language: The Ultimate Artifact) is an argument that language serves as an important scaffold. This paper will cover only the former extension and, regrettably, we will see that Clark’s brief foray into economic theory – interesting and novel though it is – fails to achieve his purpose. Before attempting to justify the above claim, it is necessary to outline Clark’s argument in some detail. He begins with the important observation that there seems to be a definite pattern in the successes and failures of economic theorising using the assumption of substantive rationality.4 Successes include the theory of the firm, theories about the behaviour of political parties, and specific types of microeconomic experimentation; among the failures are models of large-scale economic change over time, consumer behaviour, public policy-making, voter behaviour and, more generally, most decisions involving uncertainty (1997: 181).5 It is surprising, maintains Clark, that there are successes, given the ‘gross… irrealism’ (1997: 181) of substantive rationality’s ‘theory of the psychological mechanisms of daily individual choice’ (1997: 181).6 What, then, accounts for the successes? And, importantly, what accounts for the pattern of hits and misses? Clark provides two answers that he seems to think are equivalent; firstly, that ‘the best explanation of the pattern appears to involve a dissociation between cases of what may be termed highly scaffolded choice and cases of more weakly constrained individual cogitation’ (1997: 181-182), and secondly, he endorses
    4 Substantive rationality assumes that actors make decisions, suitably delimited by side-constrains, in perfect knowledge, by means-ends analysis, in terms of some decision-rule, to satisfy a complete and consistent set of preferences. It is therefore a broader notion in two different senses than Clark (1997: 181-182) suggests. Firstly, expected utility maximising models are only a subset of the decision rules posited for substantive rationality. (See Machina (1987) for several alternatives). Secondly, Clark conflates neo-classical economics and substantive rationality when the former in fact makes several additional assumptions – more on this below. In economics and rational choice theory, it is assumed that agents make choices in one of three distinct situations: (1) under conditions of certainty, (2) under conditions of risk, and (3) under conditions of uncertainty. (1) involves certainty of the options available in a particular choice-situation and infallible knowledge of consequences. (2) assumes certain knowledge about alternatives and the ability to assign non-arbitrary numerical probabilities to each outcome of every possible course of action. Under uncertainty (3), the actor is again assumed to know all the relevant options, but it is not able to assign non-arbitrary probabilities to the alternatives. See Watkins (1970: 179-193). It is worth noting that the majority of the proponents of substantive rationality deny that their view is a model of the psychological mechanisms of decision making. A standard gloss is that agents are not really substantively rational, but behave ‘as if’ they are. Indeed, Friedman (1953, esp. § III) – who Clark cites – is famous for denying that substantive rationality entails any claims about individual psychology. While Clark notes this in the article Ch. 9 is based on (1996: 270), he either misunderstands the point or disagrees with Friedman (et al.) for unspecified reasons.
    
    5
    
    6
    
    S. Afr. J. Philos. 2008, 27(2)
    
    123
    
    Satz and Ferejohn’s suggestion that a ‘structurally determined theory of interests’ which ‘allow[s] us to impute “preferences” on the basis of the constrains on success’ (1997: 183) accounts for the pattern. Although the relationship between these two answers is somewhat unclear, it seems Clark believes a ‘theory of interests’ to be available in all and only those cases where choices are highly scaffolded. Substantive rationality thus produces accurate models, on this view, only when our pattern-completing brain is embedded in (and thus augmented and constrained by) external scaffolds such as institutions, firms and organisations; whenever these scaffolds are absent, substantive rationality provides no purchase on choices or behavioural dynamics. Clark is reiterating, therefore, a long-time favourite point: what constitutes the rational mind is the fast pattern-completing abilities of the brain plus the problem-simplifying role of external scaffolds.7 It is when the brain and the scaffolds collaborate that behaviours predicted by neo-classical theory emerge and when behaviour deviates from economists’ predictions it is because the unaugmented brain does not have comparable cognitive wherewithal. In the highly-scaffolded case the constrains on behaviour can be so strong that individual rationality and individual worldviews become relatively unimportant and hence people become ‘interchangeable cogs in a larger machine’ and thus ‘the external structuring provided by institutions and organizations bears much of the explanatory burden for explaining current economic patterns’ (1997: 182, 184). Clark goes on to point out that not just any set of institutions, norms, policies, etc. induces behaviours predicted by neo-classical theory so he extends the above picture by introducing a process of cultural/institutional evolution. In this view, the institutions and other scaffolds have ‘themselves evolved as a result of selective pressure to mazimize rewards of a certain kind’ (1997: 191). The fact that brains can only complete patterns is consequently unimportant because the scaffolds augment the brain’s cognitive capacity and the scaffolds, in their turn, evolved via stringent selection. As a result, the brain-scaffold complex produces behaviour in line with economists’ predictions. In order to evaluate Clark’s position properly, it is vital to be clear about its exact nature and overall role in Being There. The proposition at stake in Chapter 9 is that pattern-completion and external scaffolding are individually necessary and jointly sufficient to account for all actual human cognitive successes in the economic sphere. Or conversely, it is not necessary to posit any computational capacities beyond simple iterated pattern-completion to explain advanced human cognition in the form of economic behaviour, because, as usual, scaffolds come to the brain’s rescue. Clark explicitly acknowledges that this is his position (§ 9.1: 179-180) but unfortunately, he goes on to defend a similar but importantly different view starting in § 9.2 (1997:180). As a consequence of this confusion, his argument is evidentially much more narrowly targeted than it ought to be, and is concerned primarily with the necessity of scaffolds. This is regrettable, because demonstrating individual necessity is one thing, but establishing joint sufficiency is quite another: indeed, for the class of behaviours under consideration, the individual necessity of scaffolding and pattern-completion is trivial. As shorthand for ‘the cognitive abilities of the unaugmented brain,’ pattern-completion is obviously necessary, because only those creatures with the relevant cognitive abilities produce the relevant economic behaviours. (No matter how powerful the scaffolds,
    7 Compare: ‘it makes sense to see Von Neumann architecture as mistakenly modeling in-the-head computation on computation that in humans consists of both an in-the-head component and (to begin with) an in-the-world component’ (Clark 1989: 135) and ‘it is the human brain plus these chunks of external scaffolding that finally constitutes the smart, rational inference engine we call mind’ (Clark 1997: 180).
    
    124
    
    S. Afr. J. Philos. 2008, 27(2)
    
    chairs never collaborate to build jumbo jets). Scaffolds too are trivially necessary: the sheer complexity of many actual economic endeavours would quickly overwhelm individual reason in the absence of environmental off-loading in the form of standard operating procedures, paper, email messages, a well-defined organisation structure, etc. (No matter how clever the person, individually discovering and then remembering the sequence of actions required to build a jumbo jet is impossible).8 The joint sufficiency assertion, however, is by no means trivial: it holds that, whenever a (suitably complex) pattern-completer is embedded in a richly scaffolded environment, outcomes of complexity comparable to those evinced by advanced human society results. Despite the need for substantial evidence to support this proposition, Clark’s defence is meagre and wholly unconvincing. To speculate, what seems to have gone wrong is simple enough: the paper (Clark 1996) on which Ch. 9 is based was unsatisfactorily adapted to fulfil its evidential role in Being There. The paper in question (‘Economic Reason: The Interplay of Individual Learning and External Structure’) was aimed at economists, and as a result understandably contained an extended argument against the implicit ‘cognitive internalism’ (Cowley & Spurrett 2003) of neo-classical economics. That is, an important concern of Clark’s (1996) is to establish the insufficiency of unaided reason and the necessity of external scaffolds.9 The first eight chapters of Being There, however, already contain extensive evidence for the necessity of scaffolds and, moreover, are aimed at a completely different audience with significantly different preconceptions. The lacuna in Clark’s argument, as we have seen, is the ‘problem of advanced cognition’ – i.e. the seeming insufficiency of pattern-completion plus scaffolding as the explanation for the particularly advanced tokens of human-level intelligence. Chapter 9 therefore ought to have been an argument for the sufficiency of Clark’s account for understanding complex economic behaviours. (Which in turn would have supported the more general conclusion that pattern-completion and scaffolding can account for advanced cognition.) What we get, instead, is a confused muddle: Clark sets out (§ 9.1) to provide the requisite positive account, but then ends up arguing against substantive rationality (§ 9.2 onwards) and, to boot, stretches the notion of scaffolding beyond breaking point and conflates a number of separate issues. Clark (1996)’s contention that substantive rationality is deeply flawed as a theory of the psychology of choice, is vulnerable to the observation that, at least some of the time, it works. That is, economic theories employing the assumption of substantive rationality are occasionally successful and predictive so, runs the thought, perhaps the assumption is not so unrealistic after all. What is required to rebut this challenge is an account that (1) does not assume substantive rationality but (2) can account for the
    8 Despite economists’ use of the substantive rationality assumption, there has long been consensus that scaffolds play an explanatory role for some economic behaviour, particularly collective ones. (Although economists of course do not generally call such problem-simplification ‘scaffolding’). Communist central planning, for example, is standardly criticised as unviable on precisely the grounds that scaffolds are necessary for successful collective decisions in particularly complicated cases. Market forces such as prices and profits, for example, are thought to carry information, thereby reducing the cognitive load. (A ubiquitous firm-level heuristic is to alter strategy by inspecting profits – i.e. to know whether a new strategy is working, a CEO need only consult the CFO; she does not necessarily need complete knowledge of the conditions prevalent in her firm’s industry.) Clark probably overestimates the importance of such an argument; he seems not to appreciate that, despite the extraordinary extent of trained incapacity in economics, few economists believe substantive rationality is an accurate psychological model.
    
    9
    
    S. Afr. J. Philos. 2008, 27(2)
    
    125
    
    other theories’ successes, and additionally (3) explains why there were occasional successes in the first place. Providing this rebuttal is Clark’s (1996) main concern and Being There inherits this project, whence the concern with explaining the pattern of successes and failures of classical economics. As we have seen, what neatly divides the successful sheep from the failed goats, according to Clark, is scaffolding: where significant scaffolding is absent, economic theory fails and where it is present, economic theory succeeds. Clark, however, seems to have misread his sources because he conflates neo-classical economics and substantive rationality, and thus individual and collective rationality.10 The notion of substantive rationality is in fact considerably broader than neo-classical economics, because the latter does, and the former does not assume that transaction costs are zero (Coase 1960; North 1993). Or, almost equivalently, traditional economics assumes the existence of a large number of (transaction-cost reducing) constraining institutions, particularly a state that protects a substantial body of property rights (Olson 1982 & 2000; Tilly 1985; North & Thomas 1973).11 The neo-classical result that markets are Pareto optimal12 depends on this assumption, and since transaction costs are in fact hardly ever zero (Wallis & North 1986; North 1993), the predictions of neo-classical theories are accurate in only a highly restricted number of cases. The class of economic behaviours that evince advanced cognition, however, is far larger than the set of accurate neo-classical models, so in ignoring the former and focusing on the latter, Clark lost sight of his explanatory target. Clark needs to demonstrate that his model of human cognition can account for advanced behaviours and cogitations and clearly such behaviours are not restricted to Pareto optimal markets. Monopolies, the Soviet Union’s economic system, public policy-making etc. are (or were) all less than efficient and certainly imperfect, but they still count as the products of advanced cognition. and thus Clark must show that pattern-completion and scaffolding are sufficient to account for them. Ironically, he explicitly cites ‘voting [and] consumer choice’ in § 9.1 as examples of the advanced cognition he needs to explain (Clark 1997: 179), but then in § 9.2 admits that, in those two cases, external scaffolding is weak (1997: 183). If so, how does pattern-completion etc. account for such cogitations? Clark never even attempts an answer – he is evidently confused about what it is he ought to be explaining quite apart from the unhelpful focus on the necessity of scaffolds. Even the meagre reasons for the joint sufficiency of scaffolds and pattern-completion that can be extracted from Clark’s account are misdirected: he attempts a defence of sufficiency for only a tiny minority of the class of relevant advanced cogitations, i.e. those explicable in terms of neo-classical economics. Moreover, Clark’s defence of sufficiency is wholly unconvincing, because he helps himself, without argument or real justification, to an evolutionary process that is dubious at best. That is, Clark unfortunately makes a habit of moving between two different senses of “evolution” (biological and cultural) without comment or acknowledging that the first sense refers to a fully worked out consensus account, and the latter to a controversial, problematic and quite possibly dubious extension thereof. A
    10 See especially p. 182: Clark repeatedly uses ‘substantive rationality’ and ‘neoclassical economic theory’ interchangeably. Clark (1996: 270) notes there is a difference, but still goes on to make the same mistake. 11 The literature just referenced is sometimes called neo-institutional economics. 12 A distribution is said to be Pareto optimal just in case no individual’s utility can be increased without reducing somebody else’s utility. It is a standard (but much criticised) measure of efficiency and, speaking very roughly, collectively rationality.
    
    126
    
    S. Afr. J. Philos. 2008, 27(2)
    
    representative case of this habit occurs on p. 78: having argued that there is a ‘delicate harmony’ between brains, bodies and world that explains termite nest building and successful ship navigation, Clark asks how this harmonisation comes about. The answer, avers Clark, is that [f]or the nervous systems of individual termites, an important part of the answer is clearly ‘through evolution’. Hutchins suggests that a kind of quasi-evolutionary process may be at work in navigation a navigation team too. The key feature is simply that small changes occur without prior design activity, and these changes tend to be preserved according to the degree to which they enhance biological success. (1997: 78, emphasis added) The first and third sentences describe biological evolution, and the second a kind of cultural evolution (‘memetics’). Without being eagle-eyed enough to spot the ‘quasi’ and knowledgeable enough to realise its significance, one would have failed to notice the transition. What is worse, in what follows the quotation, Clark allows Hutchins’s account to parasitise the explanatory success of biological evolution when cultural evolution is in fact far less widely accepted, and in several important ways deeply disanalogous to its biological form (Pinker 1997: 208-210). To understand why, consider that an extremely common error made when evaluating scientific positions is mistaking the descriptive propositions (the dependent variables or explanandum) for the theory (the logically integrated set of independent variables or explanans). This error is rooted in vernacular parlance, which uses the word ‘theory’ as a synonym for ‘proposition’ (i.e. a statement that is either true or false). But on the assumption that we have no need for two words representing exactly concept and taking cognizance of the tradition stemming from at least John Stuart Mill, running through Karl Popper onwards, such usage is both unnecessary and misleading. A better and more useful terminology is to say that, although all theories are propositions, not all propositions are theories and an important class of propositions are those that are non-theoretical and observable. Such a theory/observation distinction helps resolve otherwise misdiagnosed disagreements about the status of cultural versus biological evolution. The term (biological) ‘evolution’ is unhelpfully used to refer to both the observation that organisms change over time (‘descent with modification’, in Charles Darwin’s phrase) and the theory that explains why this change occurs, i.e. natural selection et al. (Gould 1981). This observation helps us because it becomes clear that cultural evolution is analogous to biological evolution on a descriptive/observational, but not explanatory/theoretical level. That is, it may be useful and suggestive to redescribe certain cultural phenomena in similar ways to the observed facts of biological evolution – or a kind of epidemiology of ideas. However, there simply does not exist a theory of cultural evolution similar in explanatory power to the theory of natural selection. It is therefore clearly unacceptable for Clark to conflate cultural and biological evolution as he does.13 At a minimum, therefore, Clark has insufficiently discharged his onus, and
    13 A consequence of this is that the ‘Paleyian’ question with which Clark started remains unanswered: whence the delicate balance between brain, body and world, if we rule out cultural evolution? That is, if scaffolds solve the problem of coherence (recall proposition 2, above, or see Clark 1997: 32, 191), where does the fine balance between mind and world come from? Paley (2006), let us not forget, asked the right question (see Dawkins 1986) with respect to systems that are highly complex and non-random, and thereby demonstrated that only a narrow set of theories could account for complex contrivance. Since there is no acceptable cultural evolutionary explanation (and a biological one is out of question,
    
    S. Afr. J. Philos. 2008, 27(2)
    
    127
    
    scepticism is thus in order. More substantively, if we agree cultural evolution is dubious, the sufficiency argument is undermined further. A final difficulty bears commenting on. Clark’s use of scaffolds in Ch. 9 is deeply problematic because he provides an impossibly vague description of their computational role. It is notable that the simple examples of scaffolding (i.e. most of those mentioned in Chapter 3) were culled, in the vast majority of cases, from published academic sources. That is, many of Clark’s early examples of scaffolds are subject to actual empirical research. The computational off-loading role of seeming over-rotation of zoids in tetris, for example, was not merely hypothesised from the armchair, but established in the laboratory. (See, inter alia, Maglio & Wenger 2000). All we have about scaffolds in Ch. 9, however, is hand-waving speculation. The full extent of Clark’s discussion of the scaffolding meant to allow advanced cognition in the economic sphere (which I quote at length to avoid the charge that I am erecting a straw man), is the following: firms and organizations provide an external resource in which individuals behave in ways dictated by norms, policies, and practices… [P]roblem solving, in these arenas, often involves locally effective pattern-recognition strategies which are invoked as a result of some externally originating prompt (such as a green slip in the ‘in’ tray, discharged in a preset manner) and which leave their marks on further traces (slips of paper, e-mail messages, whatever) which then are available for future manipulations within the overarching machinery of the firm. (1997: 185). While it is intuitively plausible that part of the above allows environmental off-loading (particularly mnemonically), until rigorous empirical research is conducted, it is impossible know whether any such strategies are computationally effective or actually employed in the real world. The fields of management studies and organisational psychology would no doubt have much to contribute to such research, and some may already exist. Clark, however, seems not to have investigated – he is perfectly happy to speculate from his armchair without so much as a single reference to relevant empirical studies or even a footnote conceding how speculative his account is in the absence of such studies. In summary, Clark’s attempt to account for advanced cognition in terms of pattern-completion and scaffolding is, in the economic sphere at least, a wholesale failure. While it is certain that scaffolds have a role to play in explaining economic behaviours, there is no real argument for us to evaluate in Ch. 9 that pattern-completion and scaffolding are jointly sufficient. The reason for this is that Clark becomes confused – in two different ways – about the position he is supposed to defend. First he misdirects his energies into a contextually unnecessary attack on substantive rationality, and then attempts a defence of sufficiency for only a tiny minority of the relevant cases. Moreover, even the narrow sufficiency argument that is made suffers seriously because it depends on a dubious evolutionary process. It is reasonable to conclude, therefore, that Clark has failed to address the problem of advanced cognition for a large class of economic and social behaviours.
    as most of the scaffold-brain-world complexes are evolutionarily too novel), what accounts for the delicate balance required for them to function at all? Clark has no satisfactory answer for us, so this remains a serious problem for his account.
    
    128
    
    S. Afr. J. Philos. 2008, 27(2)
    
    Bibliography Clark, A. 1989. Microcognition: Philosophy, Cognitive Science, and Parallel Distributed Processing. Cambridge, MA: MIT Press. – 1993. Associative Engines: Connectionism, Concepts, and Representational Change. Cambridge, MA: MIT Press. – 1996. ‘Economic Reason: The Interplay of Individual Learning and External Structure’ in Drobak, J. and Nye, J. (eds.) 1996. The Frontiers Of The New Institutional Economics. San Diego, CA: Academic Press. – 1997. Being There: Putting Brain, Body, and World Together Again. Cambridge, MA: MIT Press. Coase, R. 1960. ‘The Problem of Social Cost,’ Journal of Law and Economics, 3(1): pp. 1-44. Cowley, S. & Spurrett, D. 2003. ‘Putting apes (body and language) together again’, Language Sciences. 25(3): pp 289-318. Dawkins, R. 1986. The Blind Watchmaker. New York: Norton. Friedman, M. 1953. ‘The Methodology of Positive Economics,’ in Essays in Positive Economics. Chicago: University of Chicago Press. Gould, S. J. 1981. ‘Evolution as Fact and Theory,’ www.stephenjaygould.org/library/gould_fact-and-theory.html (accessed 3 May, 2007). Lenat, D. & Feigenbaum, E. 1992. ‘On the thresholds of knowledge,’ in Kirsh, D. (ed.) Foundations of Artificial Intelligence. Cambridge, MA: MIT Press. Machina, M. 1987. ‘Choice Under Uncertainty: Problems Solved and Unsolved,’ Journal of Economic Perspectives, 1(1): pp. 121-154. Maglio, P. P. & Wenger, M. J. 2000. ‘Two Views are Better than One: Epistemic Actions May Prime,’ http:/www.ibm.com/cs/people/pmaglio/pubs/psychotetris2000.pdf (Accessed 1 May, 2007). Newell, A. & Simon, H. 1976. ‘Computer science as empirical enquiry,’ in Haugeland, J. (ed.). 1976. Mind Design Cambridge, MA: MIT Press. North, D. C. 1993. ‘Economic Performance Through Time,’ Text of Prize Lecture in Economic Science in Memory of Alfred Nobel. North, D. C. & Thomas, R. 1973. The Rise of the Western World: A New Economic History Cambridge: Cambridge University Press. Olson, M. 1982. The Rise and Decline of Nations: Economic Growth, Stagflation, and Social Rigidities New Haven: Yale University Press. – 2000. Power and Prosperity: Outgrowing Communist and Capitalist Dictatorships New York: Basic Books. Paley, W. 2006. Natural Theology Oxford: Oxford University Press, 1805. Pinker, S. 1997. How the Mind Works London: Penguin. Tilly, C. 1985. ‘War Making and State Making as Organized Crime,’ in Evans, P. et. al. (ed.). Brining the State Back In Cambridge: Cambridge University Press.
    
    S. Afr. J. Philos. 2008, 27(2)
    
    129
    
    Wallis, J. J. & North, D. C. 1986. ‘Measuring the Transaction Sector in the American Economy,’ in Engerman, S.L. & Gallman, R.E. (eds.). Long Term Factors in American Economic Growth Chicago: University of Chicago Press. Watkins, J. 1970. ‘Imperfect Rationality,’ in Borger, R. & Cioffi, F. (eds.). Explanation in the Behavioural Sciences Cambridge: Cambridge University Press.

Add Comment

To add a comment, you need to login or signup (you can signup without leaving this page)