6.2.2. Subcictemac acuíferoc
6.2.2.7. Subcistema Plana de Sagunto
“If the seventeenth and the early eighteenth centuries are the age of clocks, the later eighteenth and nineteenth centuries constitute the age of steam engines, then the present time is the age of
communication and control.”
— Norbert Wiener (1961, p. 39)
In a historical overview of artificial intelligence research that he describes as more “grist for the historian’s mill” than a fleshed-out history in itself, Allan Newell notes that neither theories nor research methodologies are particularly satisfying ways to structure a historical account of his field.
In this discipline which he pioneered along with contemporaries like Herbert Simon and John McCarthy, the theories put forth, especially when successful, were often inextricably “embedded in computer systems,” and he felt that in such cases the systems – and by implication what they could actually do – always seemed to “speak louder than the commentary” (Newell, 1982, p. 2). ‘Paradigms’
or ‘research programmes’ were concepts too coarse-grained for framing the history of AI, he argued, and the field has developed and maintained a single paradigm, or at most two.1 Instead, we might better track the “history of AI as a whole… in terms of the geography of tasks successfully
1 Here he was referring specifically to the split between symbolic, representational AI of the type he and Simon originally pursued, and alternative, connectionist, previously cybernetic conceptions of cognition – a divide which I cover in greater detail below. Yet he also rhetorically implies that they are perhaps not quite distinct paradigms, given that both have broadly conceived cognition in physicalistic, mechanistic terms, and share a number of core researchers and concepts.
performed by AI systems” (ibid., p.6). In this chapter I begin to explore what such a history might look like, highlighting the crucial importance of computational tools for theorization in cognitive science. This is what the human actors throughout its history have always tried to affirm: that their ideas were constantly being guided and shaped by interaction with nonhumans, with their programs and their robots, as they have either cooperated with or resisted their theories of mind. 2
Against this premise that ‘systems speak louder than commentary,’ however, I wish to juxtapose the potency of metaphor, a feature which some might consign to the realm of mere commentary. As much as it is a history of system-building, the history of cognitive science is the history of one generative metaphor, of mind as computation. Across this history the sense of the metaphor has shifted, as has that of computation itself. But the case always remains, as Kurt Danziger contends, that “the identifying relationships that occur in psychological discourse are of two kinds: There are explicit literal definitions, and there are relationships of metaphorical analogy.
On the whole, the latter tend to be more pervasive than the former” (Danziger, 1990, p. 337). After examining some other takes on metaphor in science and psychology more specifically, I devote much of this chapter to some specific examples of technological metaphors in the discourse of mind which predate cognitive science proper: first from an era when computation was something done by specially trained humans, and then when the most fertile ground for analogy-making was the new domain of telecommunications. My analysis here is informed by the work of Michael Arbib and Mary Hesse, the most persuasive exponents of the view that “scientific revolutions are, in fact, metaphorical revolutions” (Arbib & Hesse, 1986, p. 156). Another touchstone is David Edge’s
2 Already it should be noted that this aspect of cognitive science undermines the conventional split between ‘natural’ and
‘constructed,’ since what these tools reveal is only indirectly a matter of nature – and only if the analogy of mind with machine can be made to stick. The reality which cognitive science practically and directly reveals is a wholly constructed one, an abstract and functional description of the cognitive as instantiated in artificial systems. Yet as Newell’s longtime collaborator Herbert Simon and others have argued, this may be able to tell us a great deal about the architecture of natural minds (Simon, 1996b).
argument that in scientific research, “the successful metaphor does not merely provide answers to pre-existing questions: rather, by radically restructuring our perception of the situation, it creates new questions, and in so doing, largely determines the nature of the answers” (Edge, 1974, p. 136). As he goes on to note, such metaphors often derive from the ‘hardware’ of ‘control technology’ – this is particularly true in the cases of cybernetics and cognitive science.
Equally, however, Edge’s framing highlights some of the conceptual confusion which can result from an emphasis on metaphor. This is no minor problem, in that statements like the above often raise the ire of scientists, who again wish to privilege instead the observable actions of their nonhumans: for such minds, science is not about the superficial rhetoric one uses to promote or explain one’s work, but about what one’s programs can do. I draw upon some other influential past accounts of metaphor to argue in this section that – as in many other, if not all scientific fields – the role of metaphor in cognitive science is far from superficial. Metaphors have a ‘generativity,’ and are in some sense constitutive of theory (Danziger, 1990). They precede and shape the development of formal models. They allow for communication across scientific fields, particularly relevant in the development of ‘interdisciplines’ like cognitive science. Yet they also leave open precisely which elements of the source domain should be mapped on to which of the target. While they clearly do structure scientific questions, as Edge suggests, they do not tell us everything about scientific practice, and it is troublingly ambiguous to simply state that they “determine the nature of the answers” in science.
In discussions of metaphor in the cognitive sciences, we confront two quasi-misunderstandings, both with some merit and certain inadequacies:
1. “There is no reason why computation ought to be treated as merely a metaphor for cognition, as opposed to a hypothesis about the literal nature of cognition” (Pylyshyn, 1980, p. 114).
2. “Metaphors do not ‘express’ scientists’ ideas; they are the ideas. Metaphors suggest new visions, images, and models; they inspire scientists to approach problems in new ways”
(Otis, 2001, p. 59).
It does little good to deny or suppress the generative role of metaphor in cognitive science, and it is simply false to imply that researchers have developed some complete ‘literal’ mapping of human cognitive processes onto artificial, computational ones. Despite some local successes replicating cognition-like phenomena with computer programs, the idea that mind is computation remains in many ways just a suggestive (and contentious) metaphor. Equally, however, it is a dubious rhetorical gesture to suggest that ‘scientists’ ideas’ are wholly metaphorical, or that metaphors shape scientific answers in a deterministic fashion. The metaphor of cognition-as-computation is indeed a working hypothesis within cognitive science, one which researchers are labouring on an ongoing basis to translate from the metaphorical and abstract into the literal and concrete. The success of this labour is by no means assured. The history of the field though is a story of successive groups each
recruiting some significant actors – be they politicians, funding agencies, journalists or computational systems – to their particular vision of how this literalization should proceed.
To say that metaphors determine the ‘nature of the answers’ can be read in quite different ways, and with quite incorrect implications. Computational metaphors determine the nature of answers in cognitive science only in the sense that answers are to be framed in terms of information-processing and in principle translatable into algorithms. The content of those algorithms, however, is wholly underdetermined by metaphor. We can still accept a charitable reading of the idea: the fundamental generative metaphor of cognition-as-computation is what grants unity to the discourse and to a set of research problems across its constituent disciplines. It is integral to its ‘creole’
vocabulary (Galison, 1997), and implies a definitively computational character for its tools and
theories alike. In this chapter I expand on this idea of reciprocal influence between tools and
theories (Gigerenzer, 1996), while arguing that both metaphors and tools can always be developed in a heterogeneous multiplicity of ways. Metaphors have enormous implications for scientific theories, but they are not the sole or even the predominant influence in the processes by which they are constructed. Rather, theories are formed at the intersections of generative metaphors with experimental setups and simulations, as supported by a whole coterie of propositions: some intended ‘literally,’ some ‘figuratively;’ some testable, some axiomatic, some merely suggestive; all often at odds with one another.
To some extent – an extent which I can only partially explore here – what I am describing is characteristic of theorization and knowledge in general. As David Leary contends, “All knowledge is ultimately rooted in metaphorical (or analogical) modes of perception and thought. Thus, metaphor necessarily plays a fundamental role in psychology, as in any other domain” (Leary, 1990, p. 2). He too is aware that he is ‘far from the first’ to propose that language and thought are fundamentally metaphorical, dating the idea back to Aristotle. Metaphor, in this tradition, is really a shorthand for any redescription of one semantic domain in terms of another.3 To use a well-entrenched conceptual metaphor, it is a matter of ‘seeing as.’ Or we may conceive of it as applying to something a name – and corresponding description – which belongs by convention to another thing. The somewhat circular nature of these attempts indicates the difficulty in constructing a ‘literal’ definition of metaphor. There are any number of other metaphors for metaphor, and a whole body of relevant work on the structure of metaphor, both in general (Arbib & Hesse, 1986; Black, 1963; Ricœur, 1993) and specifically in relation to cognition (Arbib, 1972; Gentner, Holyoak, & Kokinov, 2001;
3 This means the term is typically used to encompass other tropes such as simile and analogy, though others following Douglas Hofstadter prefer to use ‘analogy’ in the general case.
Hofstadter, 1979, 1996, G. Lakoff & Johnson, 1999, 2003; Rosch & Lloyd, 1978). All of these studies, in different ways, lend credence to the thesis that our capacity for metaphorization, broadly conceived, lies at the root of our cognitive capacities. Within language we can fruitfully understand phenomena in terms of things we know or believe about others; language and symbolic mediation in themselves can be understood as metaphorical in a more fundamental sense, inasmuch as words stand in for objects, and mathematical symbols can be manipulated in place of them.4 There are many layers of mediation at work in cognition. This chapter explores how metaphor runs through them and connects them to one another.
When we view metaphor as pervasive within and essential to all human thought, it of course follows that the discourse of cognitive science, like any other field of endeavour, is highly
metaphorical. But this is a rather empty proposition without further specification. What kinds of metaphors are used in cognitive science? What do they do for cognitive scientists? How do they link together various actors and help them to cohere? What kinds of resistances and transformations do they generate as they are disseminated out into the world? These questions underpin all that follows in this work. The character of metaphors in cognitive science is quite clearly computational, and yet we shall see that this is a far from homogeneous domain. Here, though, I begin by reviewing some other relevant work on metaphor, and then some antecedent technological metaphors of mind:
clocks, railways, telephones, and telegraphs. McLuhan contended that “just as a metaphor
transforms and transmits experience, so do the media” (McLuhan, 1994, p. 59). Some of the very
4 Nietzsche is a notable advocate of this view, in a relevant passage which I cite below. McLuhan makes a similar point by playing off the common etymological roots of ‘metaphor’ and ‘translation,’ linking both concepts to mediation more generally: “All media are active metaphors in their power to translate experience into new forms. The spoken word was the first technology by which man was able to let go of his environment in order to grasp it in a new way. Words are a kind of information retrieval that can range over the total environment and experience at high speed. Words are complex systems of metaphors and symbols that translate experience into our uttered or outered senses. They are a technology of explicitness. By means of translation of immediate sense experience into vocal symbols the entire world can be evoked and retrieved at any instant” (McLuhan, 1994, p. 57).
same technological extensions of our minds which fascinated him have also proven amongst the most productive sources of metaphors for characterizing our ‘internal’ cognitive and experiential selves.
It is axiomatic to my analysis that the literal and the figurative are not two neatly separable kinds of language use. The dream of a perfectly literal language is about as realistic as the recurrent dreams of a ‘universal character’ – most notably expounded by Leibniz (Cohen, 1954) and John Wilkins, then renewed in a modern guise with logical positivism. In my view these remain two domains worth distinguishing analytically, but in practice the boundary between them is exceedingly porous and mobile. Any actually existing text is a hybrid of the two. I am more sympathetic with Nietzsche, who saw literal truth as produced from many layers of metaphor:
What is a word? The image of a nerve stimulus in sounds... The ‘thing in itself’ (for that is what pure truth, without consequences, would be) is quite incomprehensible to the creators of language and not at all worth aiming for. One designates only the relations of things to man, and to express them one calls on the boldest metaphors. A nerve stimulus, first transposed into an image—first metaphor.
The image, in turn, imitated by a sound—second metaphor. And each time there is a complete overleaping of one sphere, right into the middle of an entirely new and different one (Nietzsche, 1873).
As the famous phrase from this essay has it, ‘truth’ is thus refigured as “a mobile army of metaphors, metonyms, and anthropomorphisms,” an assemblage of tropes which have become routinized to the point that we have forgotten their metaphorical character. Prefiguring later developments in the philosophy of language and sociology of knowledge, this makes truth into a “sum of human relations” (ibid). But while the front lines are always shifting between the figurative and the literal, it is unproductive to elide the distinction altogether.
In my view this is the wrong way to read both Nietzsche and later, like-minded scholars in science and technology studies.5 As Nietzsche goes on to point out, further layering on the
metaphors himself, while the construction of concepts was perhaps once a merely linguistic affair, now science has taken pride of place in this endeavour. They continually labour on this ‘bulwark’ of concepts despite the fact that they can never access the ‘things in themselves’ to serve as its
foundation; the scientist seeks shelter beneath the bulwark of their constructed truth, “and he requires shelter, for there are frightful powers which continuously break in upon him, powers which oppose scientific truth with completely different kinds of ‘truths’ which bear on their shields the most varied sorts of emblems” (ibid.). This is a quite polemical figuration which I cannot fully endorse, but has significant parallels in STS, particularly the earlier work of Latour.6 In my view it does suggest the best way to understand this underdetermination and undecidability. Philosophically speaking, there is no privileged access to reality. We can construct multiple systems of truth
consistent with the available data, and we cannot decide a priori which sorts of claims are to be interpreted literally and which figuratively. This plays out most evidently of course in truths contested between science and religion. Accepting underdetermination does not imply relativism, however. Practically and sociologically speaking, what we have to do is determine and decide in the absence of any absolute standard of truth or universal decision procedure. What the ubiquitous
5 Haraway for instance makes the point in an interview that her fascination with the ‘cyborg’ concept is tied with the porosity of the literal/figurative divide: cyborgs are “places where the ambiguity between the literal and the figurative is always working. You are never sure whether to take something literally or figuratively. It is always both/and. It is this undecidability between the literal and the figurative that interests me about technoscience. It seems like a good place to inhabit. Moreover, the cyborg involves a physicality that is undeniable and deeply historically specific” (Haraway, 2004, p. 323). I further explore the history of cyborgs, cybernetics, and scholarly engagements with both in the following chapter – here, though, I am more concerned with this ambiguous, undecidable boundary in itself.
6 By framing this as a struggle exclusively between science and ‘completely different kinds of truths,’ Nietzsche’s poetic phrasing does miss the point that many of the most significant struggles occur between different propositions claiming the mantle of the real scientific truth – only after the resolution of such a controversy, as Latour observes, can one set of propositions be deemed non-scientific and not in accord with Nature (Latour, 1988).
polemical metaphors for such situations do rightly illuminate is that the social processes by which this occurs can often be quite antagonistic.
It is not the case, then, that the discourse of the scientist is wholly literal and unlike the
‘merely figurative’ discourse of the poet or the theologian. All discourses blend together the literal and the figurative. Yet were it not for the existence of processes by which some propositions are collectively deemed figurative, others literal, and still others as tropes conveying or ‘simplifying’
literal truths, then in its labours and social functions science would indeed be indistinguishable from poetry. This is really just another way of viewing the public processes of interpretive closure
whereby ‘potentially limitless’ debates over scientific knowledge are settled in practice, however provisionally and incompletely (Collins, 1983). A proposition becoming literal as opposed to figurative, as Nietzsche framed it, is indeed partly just a matter of convention and habit—as when we cease to ‘hear’ the metaphor in referring to the end of a bed as its ‘foot,’ or the analogy between a grouse’s foot and a family ‘tree’ diagram in the old term ped de gris is elided by the modern ‘literal’
concept of ‘pedigree’ (G. Lakoff & Johnson, 2003). In the context of science, however, it is
something more than habit and convention which carries us from metaphor to literal description. In some contexts the boundary between the two kinds of proposition is relatively well-established, but in others, particularly biology and psychology, substantial controversy endures over the right metaphors and just how far they can be taken as literal models. The mechanisms by which these controversies are settled, as I have suggested, are social, but my understanding of social is a broader one more in line with both actor-network theory and the views of human actors I study.
something more than habit and convention which carries us from metaphor to literal description. In some contexts the boundary between the two kinds of proposition is relatively well-established, but in others, particularly biology and psychology, substantial controversy endures over the right metaphors and just how far they can be taken as literal models. The mechanisms by which these controversies are settled, as I have suggested, are social, but my understanding of social is a broader one more in line with both actor-network theory and the views of human actors I study.