Computationalist Linguistics

chapter four

Computationalist Linguistics

The idea that the human brain might just be a computer, to be actualized via the Promethean project of building a computer that would replicate the operations of a human brain, entailed a real (if, in part, sub rosa) investigation of what exactly was meant by “the human brain” and more pointedly “the human mind” in the first place. Such questions have been the substance of philosophical inquiry in every culture, not least among these the Western philosophical tradition, and engagement with these analytic discourses could not likely have produced any kind of useful working consensus with which to move forward.1 Rather than surveying these discourses, then, and in a fashion that comes to be characteristic of computationalist practice, the figures we associate with both the mechanical and intellectual creation of computers—including Turing, von Neumann (1944), Shannon, and Konrad Zuse (1993)—simply fashioned assertions about these intellectual territories that meshed with their intuitions. These intuitions in turn reveal a great deal about computationalism as an ideology—not merely its shape, but the functions it serves for us psychologically and ideologically.

Perhaps the easiest and most obvious way to show that a computer was functioning like a human brain would be to get the computer to produce its results the same way a human being (apparently) does: by producing language. This was so obvious to early computer scientists as hardly to need explicit statement, so that in Turing’s 1950 paper “Computing Machinery and Intelligence,” his now-famous Test simply stipulates that the computer can answer free-form English questions; Turing does not even suggest that this might be difficult or, as arguably it is, altogether intractable. Philosophers and linguists as widely varied as Putnam and Chomsky himself might suggest that the ability to use language is thoroughly implicated in any notion of intelligence, and that the bare ability to ask and answer free-form English questions would inherently entail part or even all of human intelligence, even if this was all the machine could do. From one perspective, Turing’s Test puts the cart before the horse; from another it masks the difficulty in defining the problem Turing wants to solve.

On grounds both philosophical and linguistic, there are good reasons to doubt that computers will ever “speak human language.” There are substantial ways in which this project overlaps with the project to make computers “display human intelligence,” and is untenable for similar reasons. Perhaps because language per se is a much more objective part of the social world than is the abstraction called “thinking,” however, the history of computational linguistics reveals a particular dynamism with regard to the data it takes as its object—exaggerated claims, that is, are frequently met with material tests that confirm or disconfirm theses. Accordingly, CL can claim more practical successes than can the program of Strong AI, but at the same time demonstrates with particular clarity where ideology meets material constraints.

Computers invite us to view languages on their terms: on the terms by which computers use formal systems that we have recently decided to call languages—that is, programming languages. But these closed systems, subject to univocal, correct, “activating” interpretations, look little like human language practices, which seems not just to allow but to thrive on ambiguity, context, and polysemy. Inevitably, a strong intuition of computationalists is that human language itself must be code-like and that ambiguity and polysemy are, in some critical sense, imperfections. Note that it is rarely if ever linguists who propose this view; there is just too much in the everyday world of language to let this view stand up under even mild critique. But language has many code-like aspects, and to a greater and lesser extent research programs have focused on these. Much can be done with such systems, although the ability to interact at even a “syntactically proper” level (seemingly the stipulated presupposition of Turing’s test) with a human interlocutor remains beyond the computer’s ability. Computers can aid human beings in their use of language in any number of ways. At the same time, computers carry their own linguistic ideologies, often stemming from the conceptual-intellectual base of computer science, and these ideologies even today shape a great deal of the future direction of computer development.

Like the Star Trek computer (especially in the original series; see Gresh and Weinberg 1999) or the Hal 9000 of 2001: A Space Odyssey, which easily pass the Turing Test and quickly analyze context-sensitive questions of knowledge via a remarkable ability to synthesize theories over disparate domains, the project of computerizing language itself has a representational avatar in popular culture. The Star Trek “Universal Translator” represents our Utopian hopes even more pointedly than does the Star Trek computer, both for what computers will one day do and what some of us hope will be revealed about the nature of language. Through the discovery of some kind of formal principles underlying any linguistic practice (not at all just human linguistic practice), the Universal Translator can instantly analyze an entire language through just a few sample sentences (sometimes as little as a single brief conversation) and instantly produce flawless equivalents across what appear to be highly divergent languages. Such an innovation would depend not just on a conceptually unlikely if not impossible technological production; it would require something that seems both empirically and conceptually inconceivable—a discovery of some kind of formal engine, precisely a computer, that is dictating all of what we call language far outside of our apparent conscious knowledge of language production. In this way the question whether a computer will ever use language like humans do is not at all a new or technological one, but rather one of the oldest constitutive questions of culture and philosophy.

Cryptography and the History of Computational Linguistics

Chomsky’s CFG papers from the 1950s served provocatively ambivalent institutional functions. By putting human languages on the same continuum as formal languages, Chomsky underwrote the intuition that these two kinds of abstract objects are of the same metaphysical kind. Chomsky himself dissented from the view that this means that machines would be able to speak human languages (for reasons that in some respects have only become clear quite recently); but despite this opinion and despite Chomsky’s explicit dissent, this work served to underwrite and reinforce the pursuit of CL. More specifically, and famously, Chomsky’s 1950s work was funded by DARPA specifically for the purposes of Machine Translation (MT), without regard for Chomsky’s own repeated insistence that such projects are not tenable.2 Given Chomsky’s tremendous cultural influence, especially over the study of language, it is remarkable that his opinion about this subject has been so roundly disregarded by practitioners.

In at least one way, though, Chomsky’s work fit into a view that had been advocated by computer scientists (but rarely if ever by linguists) prior to the 1950s. This view itself is quite similar to Chomsky’s in some respects, as it also hinges on the equation between formal “languages” and human languages. The intellectual heritage of this view stems not from the developers of formal languages (such as Frege, Russell, Husserl, and possibly even Peano and Boole), who rarely if ever endorsed the view that these rule sets were much like human language. Most famously, formal systems like Peano logic emerge from the study of mathematics and not from the study of language, precisely because mathematical systems, and not human languages, demand univocal interpretation. Formal logic systems are defined as systems whose semantics can be rigidly controlled, such that ambiguity only persists in the system if and when the logician chooses this condition. Otherwise, as in mathematics, there is only one meaningful interpretation of logical sentences.

The analogy between formal languages and human languages stems not from work in formal logic, since logicians usually saw the far gap between logic and language. But the early computer engineers—Turing, Shannon, Warren Weaver, even the more skeptical von Neumann (1958)—had virtually no educational background in language or linguistics, and their work shows no sign of engaging at all with linguistic work of their day. Instead, their ideas stem from their observations about the computer, in a pattern that continues to the present day. The computer does not just run on formal logic, via the Turing machine model on which all computers are built; famously, among the first applications of physical computers was to decode German military transmissions. Because these transmissions were encoded language and the computer served as an almost unfathomably efficient decoder, some engineers drew an analogy between decoding and speaking: in other words, they started from the assumption that human language, too, must be a code.

Two researchers in particular, Shannon and Weaver, pursued the compu-tationalist intuition that language must be code-like, and their work continues to underwrite even contemporary CL programs of (Shannon’s work in particular remains extremely influential; see especially Shannon and Weaver 1949; and also Shannon 1951 on a proto-CL problem). Weaver, a mathematician and engineer who was in part responsible for popularizing Shannon’s views about the nature of information and communication, is the key figure in pushing forward CL as an intellectual project. In a pathbreaking 1955 volume, Machine Translation of Languages (Locke and Booth 1955), Weaver and the editors completely avoid all discussion of prior analysis of language and formal systems, as if these fields had simply appeared ex nihilo with the development of computers. In the foreword to the volume, “The New Tower,” Weaver writes:

Students of languages and of the structures of languages, the logicians who design computers, the electronic engineers who build and run them—and especially the rare individuals who share all of these talents and insights—are now engaged in erecting a new Tower of Anti-Babel. This new tower is not intended to reach to Heaven. But it is hoped that it will build part of the way back to that mythical situation of simplicity and power when men could communicate freely together. (Weaver 1949, vii)

Weaver’s assessment is not strictly true—many of the students of language and the structures of language have never been convinced of the possibility of erecting a new “Tower of Anti-Babel.” Like some computationalists today, Weaver locates himself in a specifically Christian eschatological tradition, and posits computers as a redemptive technology that can put human beings back into the prelapsarian harmony from which we have fallen. Our human problem, according to this view, is that language has become corrupted due to ambiguity, polysemy, and polyvocality, and computers can bring language back to us, straighten it out, and eliminate the problems that are to blame not just for communicative difficulties but for the “simplicity and power” that would bring about significant political change.

Despite Weaver’s assessment, few linguists of note contributed to the 1955 volume (the only practicing linguist among them is Victor Yngve, an MIT Germanicist who is most famous for work in CL and natural language processing, referred to as NLP). In an “historical introduction” provided by the editors, the history of MT begins abruptly in 1946, as if questions of the formal nature of language had never been addressed before. Rather than surveying the intellectual background and history of this topic, the editors cover only the history of machines built at MIT for the express purpose of MT. The book itself begins with Weaver’s famous, (until-then) privately circulated “memorandum” of 1949, here published as “Translation,” and was circulated among many computer scientists of the time who dissented from its conclusions even then.3 At the time Weaver was president of the Rockefeller Foundation, and tried unsuccessfully to enlist major figures like Norbert Wiener, C. K. Ogden, Ivor Richards, Vannevar Bush, and some others in his project (see Hutchins 1986, 25-27). In contemporary histories we are supposed to see these figures as being short-sighted, but it seems equally plausible that they saw the inherent problems in Weaver’s proposal from the outset.

Despite the widespread professional doubt about Weaver’s approach and intuition, his memorandum received a certain amount of public notoriety. “An account appeared in Scientific American in December 1949 . . . This in turn was picked up by the British newspaper the News Chronicle in the spring of 1950, and so appeared the first of what in coming years were to be frequent misunderstandings and exaggerations” (Hutchins 1986, 30). Such exaggerations continue to the present day, when prototype or model systems confined to narrow domains are publicly lauded as revolutions in MT. Despite the real limitations of the most robust MT systems in existence (easily accessed today via Google’s translation functions), there is a widespread sense in the popular literature that computers are close to handling human language in much the same way humans do.

Weaver’s memorandum starts from a cultural observation: “a multiplicity of languages impedes cultural interchange between the peoples of the earth, and is a serious deterrent to international understanding” (15). Such a view can only be characterized as an ideology, especially since Weaver provides no support at all for it—and it is a view that runs at odds with other views of politics. The view becomes characteristic of computational-ist views of language: that human society as a whole is burdened by linguistic diversity, and that political harmony requires the removal of linguistic difference. We might presume from this perspective that linguistic uniformity leads to political harmony, or that political harmony rarely coexists with linguistic diversity—both propositions that can be easily checked historically, and which both arguably lack historical substance to support them.

Like Vannevar Bush in his famous article proposing the Memex (Bush 1945), Weaver reflects in large part on the role computers have played in the Allied victory in World War II. Not only is there concern about the tremendous destructive power loosed by the hydrogen bomb, which fully informed Bush’s desire to turn U.S. scientific and engineering prowess toward peaceful ends; there is a related political concern expressed in terms of linguistic diversity which has been exposed by the computational infrastructure used in the service of Allied powers in World War II. Both the exposure and discovery of atomic power and of computational power seem to have opened an American vision into an abyss of power, an access to what Deleuze and Guattari would call a War Machine that worries those previously neutral and “objective” scientists. Of course computers and computation played a vital role in the development of atomic power, and in the writings of Weaver and Bush (and others) we see a kind of collective guilt about the opening of a Pandora’s Box whose power can’t be contained by national boundaries or even political will—and this Pandora’s Box includes not just the atomic bomb but also the raw and dominating power of computation.

The most famous part of Weaver’s memorandum suggests that MT is a project similar to cryptanalysis, one of the other primary uses for wartime computing. Since cryptanalysis seems to involve language, it may be natural to think that the procedures used to replicate the German Enigma coding device (via Turing’s work with the aptly named “Bombe” computer at Bletchley Park) might also be applicable to the decoding of human language. Of course, in actuality, neither Enigma nor the Bombe played any role in translating, interpreting, or (metaphorically) decoding language: instead, they were able to generate statistically and mathematically sophisticated schemes for hiding the intended linguistic transmission, independent in every way of that transmission. Neither Enigma nor the Bombe could translate; instead, they performed properly algorithmic operations on strings of codes, so that human interpreters could have access to the underlying natural language.

Weaver’s intuition, along with those of his co-researchers at the time, therefore begins from what might be thought an entirely illegitimate analogy, between code and language, that resembles Chomsky’s creation of a language hierarchy, according to which codes are not at all dissimilar from the kind of formal logic systems Chomsky proves are not like human language. Thus it is not at all surprising that intellectuals of Weaver’s day were highly skeptical of his project along lines that Weaver dismisses with a certain amount of hubris. In the 1949 memorandum Weaver quotes correspondence he had with Norbert Wiener (whose own career reveals, in fact, a profound knowledge of and engagement with human language).4 Weaver quotes from a private letter written by Wiener to him in 1947:

As to the problem of mechanical translation, I frankly am afraid the boundaries of words in different languages are too vague and the emotional and international connotations are too extensive to make any quasimechanical translation scheme very hopeful .... At the present time, the mechanization of language, beyond such a stage as the design of photoelectric reading opportunities for the blind, seems very premature. (Wiener, quoted in Weaver 1955, 18)

Weaver writes back to Wiener that he is “disappointed but not surprised by” Wiener’s comments on “the translation problem,” in part for combinatorial (i.e., formal) reasons: “suppose we take a vocabulary of 2,000 words, and admit for good measure all the two-word combinations as if they were single words. The vocabulary is still only four million: and that is not so formidable a number to a modern computer, is it?” (18). Weaver writes that Wiener’s response “must in fact be accepted as exceedingly discouraging, for, if there are any real possibilities, one would expect Wiener to be just the person to develop them” (18-19). Rather than accepting Wiener’s intellectual insights as potentially correct, though—and it is notable how exactly correct Wiener has been about the enterprise of CL— Weaver turns to the work of other computer scientists (especially the other contributors to the 1955 volume), whose computational intuitions have led them to experiment with mechanical translation schemes.

Weaver’s combinatoric argument fails to address Wiener’s chief points, namely that human language is able to manage ambiguity and approximation in a way quite different from the way that computers handle symbols. The persistent belief that philosophical skeptics must be wrong about the potential for machine translation is characteristic of computational thinking from the 1950s to the present. Only Claude Shannon himself—again, a dedicated scientist and engineer with limited experience in the study of language—is accorded authority by Weaver, so that “only Shannon himself, at this stage, can be a good judge of the possibilities in this direction”; remarkably, Weaver suggests that “a book written in Chinese is simply a book written in English which was coded into the ‘Chinese Code’ ” (22). Since we “have useful methods for solving any cryptographic problem, may it not be that with proper interpretation we already have useful methods for translation?”

This perspective seems to have much in common with Chomsky’s later opinions about the universal structure of all languages, so that “the most promising approach” to Weaver is said to be “an approach that goes so deeply into the structure of languages as to come down to the level where they exhibit common traits” (23). This launches Weaver’s famous metaphor of language as a kind of city of towers:

Think, by analogy, of individuals living in a series of tall closed towers, all erected over a common foundation. When they try to communicate with each other, they shout back and forth, each from his own closed tower. It is difficult to make the sound penetrate even the nearest towers, and communication proceeds very poorly indeed. But, when an individual goes down his tower, he finds himself in a great open basement, common to all the towers. Here he establishes easy and useful communication with the persons who have also descended from their towers.

Thus may it be true that the way to translate from Chinese to Arabic, or from Russian to Portuguese, is not to attempt the direct route, shouting from tower to tower. Perhaps the way is to descend, from each language, down to the common base of human communication—the real but as yet undiscovered universal language—and then re-emerge by whatever particular route is convenient. (23, emphasis added)

This strange view, motivated by no facts about language or even a real situation that can be understood physically, nonetheless continues to inform computationalist judgments about language. Both Chomsky’s pursuit of a Universal Grammar and Fodor’s quest for a “language of thought” can be understood as pursuits of this “real but as yet undiscovered universal language,” a language that is somehow at once spoken and understood by all human beings and yet at the same time inaccessible to all contemporary human beings—again, positing an Ursprache from which mankind has fallen into linguistic and cultural diversity that are responsible for political disunity.

Even to Weaver, it is obvious that everyday language is not much like a conventional code. This prompts him to suggest two different strategies for using mechanical means to interpret language. First, because “there are surely alogical elements in language (intuitive sense of style, emotional content, etc.) . . . one must be pessimistic about the problem of literary translation” (22, emphasis in original). In fact, Weaver only proposes that a computer would be able to handle “little more than ... a one-to-one correspondence of words” (20). In “literary translation,” “style is important,” and “problems of idiom, multiple meanings, etc., are frequent” (20). This observation is offered with no argumentative support whatsoever, but for the idea that “large volumes of technical material might, for example, be usefully, even if not at all elegantly, handled this way,” even though, in fact, “technical writing is unfortunately not always straightforward and simple in style” (20). As an example, Weaver suggests that “each word, within the general context of a mathematical article, has one and only one meaning” (20).

Weaver’s assertion, as most linguists and literary critics know, does not mesh well with the observed linguistic facts. Many of the most common words in English are grammatical particles and “auxiliary verbs” such as “of,” “do,” “to,” “have,” and the like, which turn out to be notoriously difficult to define even in English dictionaries, and extraordinarily difficult to translate into other languages. Because each language divides up the spheres of concepts and experience differently, there is rarely a stable, one-to-one correspondence between any of these common words and those in other languages. One of Weaver’s favorite off-hand examples, Mandarin Chinese, poses some of these problems for mechanical translation. Because Chinese does not inflect verbs for tense or number, it is not possible to distinguish infinitive forms from inflected or tensed forms; there is no straightforward morphological distinction in Chinese, that is, between the English expressions to sleep and sleep. The Chinese speaker relies on context or time expressions to determine which meaning might be intended by the speaker, but there is no process of internal “translation” according to which we might determine which of these forms is “meant.” The mechanical translator from English to Chinese would either have to automatically drop the helper verb “to,” or presume that all English forms are equivalent to the uninflected Chinese verb. Neither option is accurate, and no Chinese text, no matter how controlled its domain, “avoids” it.

In his famous debate with John Searle, Jacques Derrida challenges Searle’s extension of J.L. Austin’s speech-act distinction between performative, constative, and other forms of linguistic expression, arguing that any instance of language might be understood as performative, or cita-tional, or “iterative” (Derrida 1988; Searle 1977). Here we see an even more potent and influential version of Searle’s position, that some parts of language can be broken out as the purely logical or meaningful parts, leaving over the “stylistic” or “literary” elements. But what Wiener tried to argue in his response to Weaver is that all language is literary in this sense: language is always multiple and polysemous, even when we appear to be accessing its univocal parts.

The crux of Wiener’s and Weaver’s disagreement can be said to center just on this question of univocality of any parts of language other than directly referential nouns. This disagreement rides on the second suggestion Weaver and other computationalists make, and which continues to inform much computational work to this day: namely that human language is itself broken because it functions in a polysemous and ambiguous fashion, and that the solution to this is to rewrite our language so as to “fix” its reference in a univocal manner. One reason Weaver writes to Ogden and Richards is because of their work earlier in the century on a construct called “Basic English,” one of many schemes to create universal, standard languages (the effort also includes Esperanto) whose lack of referential ambiguity may somehow solve the political problem of human disagreement. Wiener points out in his letter to Weaver that “in certain respects basic English is the reverse of mechanical and throws upon such words as get a burden which is much greater than most words carry in conventional English” (quoted in Weaver 1949, 18). Weaver seems not to understand the tremendous importance of particles, copulas, and other parts of language that have fascinated linguists for centuries—or of the critical linguistic role played by ambiguity itself. Most English-speakers, for example, use words like “to,” “do,” “have,” “get,” “like,” “go,” and so forth with tremendous ease, but on close question often cannot provide precise definitions of them. This would lead one to believe that precise semantics are not what human beings exclusively rely on for language function, even at a basic level, but of course this observation may be unsettling for those looking at language for its deterministic features.

Weaver’s principal intuition, that translation is an operation similar to decoding, was almost immediately dismissed even by the advocates of MT as a research program. In standard histories of MT, we are told that this approach was “immediately recognized as mistaken” (Hutchins 1986, 30), since the “computers at Bletchley Park were applied to cracking the cipher, not to translating the German text into English” (30). This conclusion emerges from engineering investigation into the problem, since few linguists or philosophers would have assented to the view that languages are codes; nevertheless, even today, a prominent subset of computationalists (although rarely ones directly involved in computer language processing) continue to insist that formal languages, programming languages, and ciphers are all the same kinds of things as human languages, despite the manifest differences in form, use, and meaning of these kinds of systems. Surely the mere fact that a particular word—language—is conventionally applied to these objects is not, in and of itself, justification for lumping the objects together in a metaphysical sense; yet at some level, the intuition that language is code-like underwrites not just MT but its successor (and, strangely, more ambitious) programs of CL and NLP.

From MT to CL and NLP

Weaver’s memo continues to be held in high esteem among computational researchers for two suggestions that have remained influential to this day. Despite the fact that MT has produced extremely limited results in more than 50 years of practice—as can be seen using the (generally, state-of-the-art) mechanisms found in such easily accessed tools as Google and Babelfish— computationalists have continued to suggest that machines are not just on the verge of translating, but of actually using human language in an effective way. It is often possible to hear computationalists without firm knowledge of the subject asserting that Google’s translator is “bad” or “flawed” and that it could be easily “fixed,” when in fact, Google has devoted more resources to this problem than perhaps any other institution in history, and openly admits that it represents the absolute limit of what is possible in MT.5

With time and attention, CL and NLP projects have expanded in a range of directions, some more fruitful than others. While observers from outside the field continue to mistakenly believe that we are on the verge of profound progress in making computers “speak,” a more finely grained view of the issues suggests that the real progress in these fields is along several much more modest directions. Contemporary CL and NLP projects, proceed along several different lines, some inspired by the work originally done by Weaver and other MT advocates, and other work inspired by the use of computers by linguists. Contemporary CL and NLP includes at least the following areas of inquiry: text-to-speech synthesis (TTS): the generation of (synthesized) speech from written text; voice recognition: the use of human language for input into computers, often as a substitute for written/-keyboarded text; part-of-speech tagging: the assignment of grammatical categories and other feature information in human language texts; corpus linguistics: the creation and use of large, computer-based collections of texts, both written and spoken; natural language generation (NLG) or speech production: the use of computers to spontaneously emit human language; conversational agents: the computer facility to interact with human interlocutors, but without the ability to spontaneously generate speech on new topics; CL of human languages: the search for computational elements in human languages; statistical NLP: the analysis of human language (often large corpora) using statistical methods; information extraction: the attempt to locate the informational content of linguistic expressions and to synthesize or collect them from a variety of texts. The visible success in some of these programs (especially ones centered around speech synthesis and statistical analysis, neither of which has much to do with what is usually understood as comprehension; Manning and Schutze [1999] provides a thorough survey) leads to a popular misconception that other programs are on the verge of success. In fact, the extremely limited utility of many of these research programs helps to show why the others will likely never succeed; and the CL/NLP programs that avoid a hardcore computationalist paradigm (especially statistical NLP) help to show why the main intellectual program in these areas—namely, the desire to show that human language is itself computational, and simultaneously to produce a “speaking computer” and/or “universal translator”—are likely not ever to succeed. The work of some researchers relies exactly on the statistical methods once championed by Weaver to argue that some contemporary conceptions of language itself, especially the formal linguistics programs inspired by Chomsky, rely far too heavily on a computationalist ideology.

Two of the most successful CL programs are the related projects of TTS and voice recognition: one using computers to synthesize a human-like voice, and one using computers to substitute spoken input for written input. In neither case is any kind of engagement with semantics or even syntax suggested; contrary to the suggestions of Weaver and other early CL advocates, even the apparently formal translation of written language into spoken language, and vice versa, engages some aspects of the human language system that are at least in part computationally intractable. The simpler of the two programs is voice recognition; today several manufacturers, including Microsoft, IBM, and a specialist company, Nuance, produce extremely effective products of this sort, which are useful for people who can’t or prefer not to use keyboards as their main input devices. These programs come equipped with large vocabularies, consisting mainly of the natural language terms for common computer operations (such as “File,” “Open,” “New,” “Menu,” “Close,” etc.), along with the capability to add a virtually unlimited number of custom vocabulary items. Yet even the best of these systems cannot “recognize” human speech right out of the box with high accuracy. Instead, they require a fair amount of individualized training with the user. Despite the fact that software developers use statistical methods to pre-load the software with a wide range of possible variations in pronunciation, these systems always require user “tuning” to ensure that pitch, accent, and intonation are being read correctly. Unlike a human when she is listening to another human being, the software cannot necessarily perceive word boundaries consistently without training. In addition, these systems require a huge amount of language-specific programming, much of which has not been proven to extend to languages and pronunciation styles that are not anticipated by the programmer. There is no question of “comprehension” in these systems, despite the appearance that the computer does what the speaker says—anymore than the computer is understanding what one types on a keyboard or clicks with a mouse. The voice recognition system does not know that “File—Open” means to open a file; it simply knows where this entry is on the computer menu, and then selects that option in just the same way the user does with a mouse.

TTS systems, also widely used and highly successful, turn out to present even more complex problems as research objects. Because the voice recognition user can control her vocabulary, she can ensure that much of her input will fall into the range of the software’s capability. TTS systems, unless the vocabulary is tightly controlled, must be able to pronounce any string they encounter, and even in well-controlled language practices like English this may include a vast range of apparently nonstandard usages, pronunciations, words, and spellings. In addition, for TTS systems to be comprehensible by human beings, it turns out that much of what had been understood (by some) as paralinguistic features, like prosody, intonation, stops and starts, interruptions, nonlexical pauses, and so forth, must all be managed in some fashion, or otherwise produce speech that sounds extremely mechanical and even robotic. Most such systems still today sound highly mechanical, as in the widely used Macintosh speech synthesizer or the systems used by individuals like Stephen Hawking, who are unable to speak. Even these systems, it turns out, do not produce speech in a similar manner to human beings, but rather must use a pre-assembled collection of parts in a combinatorial fashion, often using actual recorded samples of human speech for a wide range of applications that directly generated computer speech can’t manage.

Among the best-studied and most interesting issues in TTS is the question of intonation. While in some languages formalized tones (such as Chinese, Thai, and Vietnamese) are a critical feature of language “proper,” other levels of intonation—roughly, the tunes we “sing” while we are speaking, such as a rising intonation at the end of questions—are found in all languages. Human speakers and hearers use these tonal cues constantly in both production and reception, but the largely analog nature of this intonation and its largely unconscious nature have made it extremely difficult to address, let alone manage, in computational systems. One contemporary researcher, Janet Pierrehumbert, has focused on the role of intonation in a general picture of phonology, and while she makes extensive use of computers for analysis, her work helps to show why full-scale TTS poses significant problems for computational systems:

Intonational phrasing is an example of “structure.” That is, an intonation phrase specifies that certain words are grouped together, and that one of these words is the strongest prosodically. The intonation phrase does not in and of itself specify any part of the content. Rather, it provides opportunities to make choices of content. For each intonation phrase, the speaker selects not only the words, but also a phrasal melody. The phonetic manifestations of the various elements of the content depend on their position with respect to the intona-tional phrasing. (Pierrehumbert 1993, 261)

The intonational phrase is among the most common and analog aspects of speech (analog in the sense that there is no hard-and-fast binary role for when a particular intonational melody does or does not have a specific meaning, but are also interpreted with remarkable skill and speed by human listeners). In the most obvious cases, such as brief questions and brief declarative statements, it is possible to create stereotypes that are recognizable by human beings, but computational systems fail to make the consistent alterations in such presentations that are characteristic of human speech. A human being has a range of melodies for asking questions, each of which may contribute to meaning in a more-or-less fashion; a computer has no background position from which to “choose” any of these analog options, but rather is likely to choose a hard-coded intonation phrase in these most stereotypical cases. The result is highly mechanical, even if recognizable by human beings.

Of course it is tempting, especially from the computationalist perspective, to see these intonational phenomena as paralinguistic, but Pierrehum-bert has shown conclusively (as many linguists have also known for hundreds of years) that such phenomena are critical to all spoken language (and even in our interpretation of some written language) and that they reach deeply into the nature of phonetics itself. Intonation and phonetics are in part “suprasegmental”: they are meaningful units that are larger than the apparent span of words (at least, in the way English orthography creates word boundaries, which as Wiener points out, is not consistent across all languages). Even the most apparently straightforward of phonetic facts, such as the pronunciation of individual speech sounds, turn out to be seriously conditioned by their place in suprasegmental melodies. While “there is no impasse as far as the human mind is concerned; it apparently recognizes the phonemes and the prosody together” (269), the need for discrete representations of any linguistic facts in a CL system makes it difficult to determine whether prosody or phonetics are the “base” representation on which to grow the other. Pierrehumbert calls the effects of prosody and intonation on human speech “nonlocal,” using vocabulary from physics to suggest that the impact of such phenomena can’t be assigned in a one-to-one correspondence with apparent linguistic tokens. “Nonlocal” implies a kind of mystical action-at-a-distance that may not comport well with the physical nature of the mind; a less overdetermined term might be “analog.”

The correspondence between apparently linguistic tokens and supraseg-mental melodies is something that computational researchers have hardly learned how to manage for purposes of recognition, to say nothing of natural language generation. Because “melody is not provided by the lexical choice, but rather functions as a separate simultaneous channel of information” (277), it would be necessary to categorize this channel meaningfully for computational generation; yet we do not even have a proposed routine for capturing this phenomena, despite the everyday ease with which human language users access it. Yet for speech-to-text technology to work in either direction (TTS systems and voice recognition), this phenomenon would have to be well understood and modeled not just for English but for other languages that use tone and melody in a variety of ways. For this reason, some TTS systems use what are essentially recordings of human voices speaking certain strings with “natural” (but generally unanalyzed) intonation, which are then simply played back by the system when the token is accessed.

Neither TTS nor voice recognition addresses what most in the general public assume to be the main tasks for CL and NLP, despite the clear need for these programs to handle such phenomena. Instead, putting aside the difficulties in this aspect of speech production, most CL and NLP research from the 1960s until recently has bypassed these issues, and taken the processing of written text in and of itself as a main project. (Even in Turing’s original statement of the Test, the interlocutors are supposed to be passing dialogue back and forth in written form, because Turing sees the obvious inability of machines to adequately mimic human speech as a separate question from whether computers can process language.) By focusing on written exemplars, CL and NLP have pursued a program that has much in common with the “Strong AI” programs of the 1960s and 1970s that Hubert Dreyfus (1992), John Haugeland (1985), John Searle (1984, 1992), and others have so effectively critiqued. This program has two distinct aspects, which although they are joined intellectually, are often pursued with apparent independence from each other—yet at the same time, the mere presence of the phrase “computational linguistics” in a title is often not at all enough to distinguish which program the researcher has in mind.

SHRDLU and the State of the Art in Computational Linguistics

The two faces of CL and NLP in its strong mode are either (1) to make computers use language in a fully human fashion, generally via conversational agents that can interact “in the same way as a human being” with human or other language-using interlocutors; and (2) to demonstrate that human language is itself a computational system, and therefore can be made algorithmically tractable for computation. No doubt, the most famous computer system that pursues primarily the first goal is a system built in the late 1960s and early 1970s by Terry Winograd in the Artificial Intelligence laboratory at MIT, through the same program that funded Chomsky’s work. Winograd’s system, called SHRDLU (named for the second line on a linotype printer; see Winograd nd1), took advantage of the 20 years of research following Weaver’s original memo on MT. Rather than starting with language, then, Winograd saw that a more feasible approach was to start with the world—or rather a world, one which a computer could conceivably be able to reason about—noting that this quality of “aboutness” (or what Searle and other phenomenologists call intentionality) seems to be a base feature of human language. Since a computer has such limited sensory access to the human world of experience, part of Winograd’s project was to create an ingenious “model world”—a world of objects and actions that both human beings and computers might be able to speak about.

SHRDLU’s model world persists as one of the most famous constructions in the history of computer science. Its world can be depicted in both physical space and simulated in the computer, such that the various programs contributing to SHRDLU can “perceive” objects in the model world in the same way they might be metaphorically said to perceive the elements of any running computer simulation. Rather than simply reading output devices, though, SHRDLU is constructed so as to provide answers in a limited (200-word) vocabulary of English. The model world consists of any number of a limited kind of objects: blocks of exactly three shapes (pyramids, blocks, and boxes). The blocks are colored, which provides SHRDLU with a range of real-world information about its model. It also “understands” some basic physical concepts of its world; and it can “learn” by having new terms defined in terms of words it already understands. SHRDLU also understands certain other actions, including actions it can take (moving blocks around), providing it with models of the most prominent grammatical categories (nouns, verbs, adjectives).

SHRDLU is able to produce remarkable conversations in what look like everyday written English transcriptions of conversations. Because it has physical extension, it is able to interact with the world of blocks, whether via a physical appendage or the internal simulation Winograd’s system used. Through a complex and highly modularized series of computer programs, SHRDLU does something that looks like conversing:

Person: PICK UP A BIG RED BLOCK.

Computer: OK. (does it)

Person: GRASP THE PYRAMID.

Computer: I DON’T UNDERSTAND WHICH PYRAMID YOU MEAN.

Person: FIND A BLOCK WHICH IS TALLER THAN THE ONE YOU ARE HOLDING AND PUT IT INTO THE BOX.

Computer: BY “IT”, I ASSUME YOU MEAN THE BLOCK WHICH IS TALLER THAN THE ONE I AM HOLDING.

Computer: OK. (does it)

Person: WHAT DOES THE BOX CONTAIN?

Computer: THE BLUE PYRAMID AND THE BLUE BLOCK.

Person: WHAT IS THE PYRAMID SUPPORTED BY?

Computer: THE BOX.

Person: HOW MANY BLOCKS ARE NOT IN THE BOX?

Computer: FOUR OF THEM.

Person: IS AT LEAST ONE OF THEM NARROWER THAN THE ONE WHICH I TOLD YOU TO PICK UP?

Computer: YES, THE RED CUBE.

Person: IS IT SUPPORTED?

Computer: YES, BY THE TABLE.

Person: CAN THE TABLE PICK UP BLOCKS?

Computer: NO.

Person: CAN A PYRAMID BE SUPPORTED BY A BLOCK?

Computer: YES.

Person: CAN A PYRAMID SUPPORT A PYRAMID?

Computer: I DON’T KNOW.

Person: STACK UP TWO PYRAMIDS.

Computer: I CAN’T.

(SHRDLU dialogue from Winograd ndz)

It isn’t easy to capture the excitement SHRDLU created in 1968 and 1969 when Winograd first showed the system to other AI researchers. Winograd seemed to have started at just the right place, by replicating what looks like the linguistic behavior of a young human being who may not have mastery of the full range of linguistic expressions, but can demonstrate impressive knowledge of a limited set of world-word relations. This dialogue reads plausibly like an interaction between an adult and a child playing with blocks, although it lacks the inevitable digressions present in almost any child’s speech.

Like many CL & NLP projects, SHRDLU gestures in two directions at once: it hints at building more complicated systems that could handle more human language; and it also presents a model of what might be going on in the human brain. Notably, however, this model is much less subtle than Chomsky’s view of linguistic computation; instead, the system relies exclusively on expressions from formal logic. Nearly every linguistic operation SHRDLU can perform can be expressed in a logical equation, up to and including the assignment of variables to unknown terms, as in the following stretch of dialogue:

Person: DOES A STEEPLE

Computer interrupting: SORRY, I DON’T KNOW THE WORD “STEEPLE”.

Person: A “STEEPLE” IS A STACK WHICH CONTAINS TWO GREEN

CUBES AND A PYRAMID.

Computer: I UNDERSTAND.

Person: ARE THERE ANY STEEPLES NOW?

Computer: NO.

Person: BUILD ONE.

Computer: OK. (does it)

Person: CALL THE BIGGEST BLOCK “SUPERBLOCK”.

Computer: OK.

Person: HAVE YOU PICKED SUPERBLOCK UP SINCE WE BEGAN?

(SHRDLU dialogue from Winograd ndz)

Here SHRDLU demonstrates variable assignment to an unknown term (“steeple”); demonstrates the ability to extrapolate from the single noun steeple to its plural steeples without explicitly asking for a pluralization rule; and assignment of a new term when this is explicitly assigned (“Call the biggest block ‘superblock’ ”) when this requires analysis of the model world according to measurable criteria.

SHRDLU resembles the human linguistic facility so closely that many AI researchers took it to be a clear proof of concept for much more robust and usable CL systems. Underlying this view, though, is what can only be called a computationalist view of language itself. SHRDLU does not posit that it models a system that is part of human language, or a linguistic capability that human beings can also perform; rather, at least in its initial formulations, SHRDLU includes the posit that its mechanism is the fundamental operation of language, and that other linguistic operations are parasitic on, or additional to, this fundamental one. In this sense SHRDLU crosses the line between a CL research program and a model of human language capability. The programs that make up SHRDLU might be plausibly thought a simulation of the “modules” that make up the human language capability. Each module contains explicit definitions for the linguistic features it needs to “know about”; for example, SHRDLU includes explicit representations of linguistic entities like CLAUSE, PHRASE, WORD, ADJUNCT, TRANSITIVE VERB, INTRANSITIVE VERB, ADJECTIVE, and so on. Each of these categories has been hard-coded into the programming structure, so that any nonce word encountered by the system must be assigned to its relevant categories, using contextual clues.

In addition, the SHRDLU system contains explicit representations of phenomena like logical relations (“equals,” “and,” “or,” “is a member of”), spatial relations (“on top of,” “inside of”), phrasal types (question, declaration, request), and others. None of these systems are available for reflection or modification by the “speaker”; they are programmed in by the system beforehand. Phenomena that do not appear in SHRDLU’s explicit representations can’t be added to the system via conversation (as nonce words can be); instead they, too, must be programmed into the computer. Here already we encounter a significant difference between the SHRDLU system and the human one, because we simply don’t know what the human equivalent of the SHRDLU programming environment is; and it would be tendentious to say the least to suggest that analogues for the modules in the human brain could be much like SHRDLU’s modules.

Take one of the most problematic modules, namely, the one that realizes Boolean and other logical forms. It surely is the case that human children and adults can manipulate these concepts; yet it is also the case that some logical concepts can be among the difficult developmental stepping-stones for children, and that the use of logic by human beings does not seem to reflect a hard-coded “logic module” that is part of the Language Organ, simply because human beings are extremely inconsistent in their implementation of logical formulae. Much of formal logic, even in its simplest instantiations, is put off in our educational system until university. Surely if the human language organ relied heavily on hard-coded routines for critical functions like AND, OR, and NOT, it would be strange to find that they filter up to the actual human being so imperfectly. Some system of interference must be preventing the brain from accessing the logical computer operating inside of itself—which seems like an arbitrary and counterintuitive way for the biological organism to develop.

The human child, unlike the SHRDLU program, must create or realize its linguistic facility solely through conversation and introspection: positing a programmer to write routines comes dangerously close to endorsing a kind of intelligent-design view of the development of language. The child’s grasp of logic is not simply an imperfectly matured system, but a part of a complex interaction between individual and environment whose configuration is not at all clear. This can be seen by considering the ways in which the SHRDLU sample dialog is actually quite different from a parallel conversation that an adult might have with a child about the same block world SHRDLU uses. Of course any given child might well have exactly the kind of conversation shown in the SHRDLU example; at the same time, it is fair to say that the SHRDLU conversation does not quite follow the typical course of a human interaction on the same subject.

Take the definition of new terms by ostension, using the example “steeple.” Winograd’s writing is compelling and his model is impressive, so it is easy to ignore the fact that what the computer has “learned” here is an os-tensive formal definition that fails to capture exactly what would be relevant about the term “steeple” in English. While what SHRDLU understands as a steeple resembles an English-language steeple in shape and to a limited degree in function, it absolutely does not carry with it the connotations most English speakers would associate with the term (for example, that it is associated with some styles of church architecture, typically painted a neutral color, and not necessarily identical with any object of its relative shape and size). Both of the terms SHRDLU learns by definition in the dialogue, steeple and superblock, resemble English terms only by analogy. In fact, like all completely computer-based languages, or what should better be called formal systems, terms defined in SHRDLU’s language are entirely contained within the world of form; they do not make reference to the larger world and its web of holistically defined and contextually malleable language uses. They make a perfect model of what Chomsky sometimes calls I-language (for “Internal Language,” the pre-semantic formal code on which the language faculty operates), but only because despite appearances there is no “external reality” for them to deal with; there is no possibility of one of SHRDLU’s interlocutors spontaneously using steeple in a wholly different manner from what has come before, or metaphorically applying it to something that is not at all a steeple, or analogizing superblock to apply to a different kind of object altogether.

For these reasons and other similar ones, Winograd himself abandoned the SHRDLU project and, under the influence of Heideggerian thought such as that found in Dreyfus’s writings, began to think about the ways in which computers as machines interact with humans and function in the human world, and to see language as a practice that can only be understood in context. Language plays only a partial role in the social world in Wino-grad’s revised vision:

The world is encountered as something always already lived in, worked in, and acted upon. World as the background of obviousness is manifest in our everyday dealings as the familiarity that pervades our situation, and every possible utterance presupposes this. Listening for our possibilities in a world in which we already dwell allows us to speak and to elicit the cooperation of others. That which is not obvious is made manifest through language. What is unspoken is as much a part of the meaning as what is spoken. (Winograd and Flores 1987, 58).

Of course, despite the continued respect with which he is received in the computer world, Winograd’s knowing dissent from computationalist dogma, and his specific and thorough knowledge of the CL research programs and its orientation, had little impact on the actual direction of CL and NLP. They could not, because these worlds are not driven by the scientific procedures according to which they claim to operate; rather, they are ideological programs driven by a furious need to show that computers can take over language, that language itself can be formalized, and that those parts of human experience that cannot be formalized, and so are difficult to control mechanically, are also of little value or interest.

chapter five

Linguistic Computationalism

Despite the manifest difficulties found everywhere in high computationalist attempts to make computers “speak,” thinkers outside of CL proper persist in believing that such a development is likely in the near future. Few prominent CL researchers describe their projects as failures, or spend much time explaining in their published works the significant limitations of fully algorithmic efforts to produce free-form speech, since it has become quite clear that such efforts are no longer considered worthwhile or even meaningful projects to pursue. This does not mean CL and NLP are unsuccessful: on the contrary, the complexity they help to uncover in linguistic production has opened up new avenues for linguistic analysis, of both the computational and conventional varieties. They have simultaneously revealed any number of limited-scope and domain-specific linguistic tasks that are highly tractable to computational analysis.

Some computer games, for example, can list every statement that can be made within the world of the game, so that fully automated characters can speak them; yet the completely routine nature of such speech fragments is never mistaken by linguists for the syntactic and semantic creativity everywhere evident in natural languages. Such computational mechanisms can be extended, so that, for example, certain programs for natural language generation can be used to produce what look like standard news stories about limited domains: the statistics of a baseball game, for example, can be used to generate terse game summaries that read much like what a human writer might produce. Again, no serious linguistic researcher mistakes such productions for mimicking more than an important part of natural language; and critically, no serious linguistic researcher suggests that this success implies further success in free-form language production. On the contrary, such domain-specific success looks much more like the absolute limit of computational language production.

Somewhat remarkably, but in a strong testament to the power of compu-tationalist ideologies, outside the sphere of professional CL researchers, we find a much less skeptical group of researchers who do, in fact, believe that much more significant strides in computer processing of language are just around the corner. Despite the repetition of such claims throughout the history of computers, these computationalists presume that CL researchers simply have not experimented with obvious strategies for comprehension and translation, or that insufficient processing power has been brought to bear on the problem. While the presence of professional computational linguists in industry today means that companies who focus on language products make such claims less often than in the past, they can still be found with unexpected frequency outside professional CL circles. Yet the views of professional computational linguists have a surprisingly restricted sphere of influence. They are rarely referenced even by the computer scientists who are working to define future versions of the World Wide Web, especially in the recent project to get the the web to “understand” meanings called the Semantic Web. They have little influence, either, over the widespread presumption among computationalists that the severe restrictions today’s computers impose on natural language diversity are irrelevant to the worlwide distribution of cultural power.

Computationalism and Digital Textuality

During the last fifteen years, a small body of writing has emerged that is concerned with an idea called in the literature the OHCO thesis, spelled out to mean that texts are Ordered Hierarchies of Content Objects. This idea emerges in a particularly narrow band of academic scholarship situated somewhere between computer science and library science that is still notably representative of much computationalist thinking and research. It is also notably distant from mainstream philosophical practice of any sort, or from literary theory, or from textual criticism or linguistics, and suffers from this lack of contextualization. As proposed in a series of essays by DeRose, Renear, and others (Coombs, Renear, and DeRose 1987; DeRose et al. 1990; Renear 1995), and affirmed in later works by them and others, the OHCO thesis, and no less the stubborn pursuit of increasingly restricted versions of it, is supposed to have been a kind of first stab at a way to represent not merely the formal aspects of a text (in the sense that a word processor can represent any text that is typed into it), but textual semantics.1

Now largely abandoned in favor of a somewhat grudging admission that many semantic features cannot be definitively represented (a thesis called in Renear, Mylonas, and Durand [1996] “OHCO-3,” “perspectives can be decomposed into OHCOs”), OHCO’s initial acceptance within the scholarly text processing community is said to reflect normal scientific practice adjusting itself to the facts, and no doubt it does. But the particular shape taken by the initial OHCO thesis is highly revealing about a computational bias: a gut feeling or intuition that computation as a process must be at the bottom of human and sometimes cultural affairs, prior to the discovery of compelling evidence that such a thesis might be correct. In the practical world, this often translates to an assumption that the default methods already built into the computing architecture are largely satisfactory to represent the world of culture, with some minor modifications perhaps necessary “at the margins,” and often despite the explicit statement of earlier technologists that these tools will not be appropriate for those tasks.

Recent work on the OHCO thesis suggests that its most restrictive versions are untenable just because of deep conceptual issues that emerged only through prolonged contact with real-world examples. While true in its own way, this theory-internal story plays down the rationalist cultural orientation of the computer sciences themselves. The literature on the OHCO thesis is surprisingly reticent about these underpinnings. Have observations about the structure of text or of language, or philosophical speculation, led well-known thinkers to propose something like OHCO? Renear (1995) approvingly cites an extract from Wittgenstein for an epigraph, an extract that tends toward Wittgenstein’s late “healing” view of philosophical practice as a process of “clarifying,” but he does not connect this to the OHCO question in the main text. Nor does Renear reflect on the fact that this Wittgenstein (1953) is one of the writers whose work would most immediately lead one to raise questions about any OHCO-like thesis.2

Despite being presented in a topic cluster in the philosophy journal The Monist about the philosophical implications of electronic publishing, Renear locates digital textuality in a tendentious frame. Renear develops a philosophical typology via three seemingly familiar positions, which he calls Platonism, Pluralism, and Antirealism. Platonism, to Renear, is the “view that texts are hierarchical structures of objects such as chapters, titles, paragraphs, stanzas, and the like, [and] is already implicit in the early efforts to theorize about the development of text processing software, and is implied in the SGML standard itself, which is a standard for defining hierarchical models for representing texts” (Renear 1995, §5.1.2). The idea that we draw such conclusions from the practice of computer engineers is quite odd, and no less so in that it requires an analysis of the Platonic texts that is itself tendentious and that Renear does not provide. Since SGML describes hierarchical objects, we would of course expect anything it characterizes to be hierarchical. But how do we get from here to the view that text itself, in the abstract, is hierarchical, much less that such a view is Platonic?

Despite the scantiness of the discussion in Renear (1995), it remains one of the only serious examinations in the OHCO literature of the thesis that text in the abstract is fundamentally hierarchical. The OHCO literature presumes the thesis is true, and then grudgingly admits accommodations; it only rarely pauses to touch down in the philosophical and conceptual literature which it claims to have engaged, which often anticipates the same issues. No recent philosophical work on language or textuality—whether in the Anglo-American or Continental traditions—provides good underpinnings for such a formalist view of language (and even Chomsky’s formalism, arguably the most hierarchical of current linguistic theories, seems irreconcilable with anything like OHCO). Theories of literary editing, for example those of Thomas Tanselle (1992), Donald McKenzie (1999), Jerome McGann (1983, 1991), soundly reject any such theses and never support what Renear calls a Platonic orientation. So it would be difficult to say, as one might wish, that the original OHCO thesis is rooted in philosophical tradition, or editing tradition, or any tradition at all but the intuitions of particular computer scientists.

The lack of historical setting is no less problematic when we come to Re-near’s second philosophical faction, “Pluralism”: and here, unlike Platonism, there really is not even a living philosophical position one could tenably identify with the name. Renear tells us that the pluralist emerges in response to a “real practical problem”:

Briefly the difficulty is that while the SGML world assumed that text encoders would always represent the logical structure of a text as a single hierarchical structure, there in fact turned out to be many hierarchical structures that had reasonable claims to be “logical.” A verse drama for instance contains dialogue lines, metrical lines, and sentences. But these do not fit in a single hierarchy of non-overlapping objects: sentences and metrical lines obviously overlap (enjambment) and when a character finishes another character’s sentence or metrical line then dialogue lines overlap with sentences and metrical lines. (Renear 1995, §5.2.1)

Once again, Pluralism emerges not in a philosophical or conceptual context but in a computational context, generated by “text encoders” of “the SGML world.” Whether one also calls these researchers “library scientists,” “computer scientists,” “engineers,” “programmers,” etc., it is hard to grant that their basic intuitions about such a fundamental question as the nature of text should be taken seriously without deep engagement with the fields of research touching on the question, and in which the questions they raise are well-known and could hardly be said to have settled answers.

Unlike the classical position called Platonism and his relatively ad-hoc Pluralism, Renear’s third classificatory term comes straight out of contemporary philosophy: Antirealism. Renear associates this term with Wittgenstein through the work of Alois Pichler, “a researcher at the Bergen Wittgenstein Archives who is the author of some very important and philosophically illuminating papers on transcription,” who writes in an interpolated quote, apparently following Wittgenstein, that “texts are not objectively existing entities which just need to be discovered and presented, but entities which have to be constructed” (Pichler 1995, p. 774, quoted in Renear 1995, §5.3.1). Renear suggests that Antirealists like “Landow, Bolter, and Lanham do not really establish their claims,” and “neither do Huitfeldt, Pichler, and the other post-modern text encoders.”

Without reference to any background literature or recognizable positions—as to Platonism, Pluralism, Antirealism, or postmodernism— Renear then retreats from his original claim, and in fact grants the accuracy of what Huitfeldt sees as the two major components of the Antirealist view: “there are no facts about a text which are objective in the sense of not being interpretational,” and “a text may have many different kinds of structure (physical, compositional, narrative, grammatical)” (quoted in Renear 1995, §5.3.7 and §5.3.9, respectively). To Renear, in fact, both of these claims “may be true”—in fact he “think[s] that they are true”—but there “is no path of compelling argumentation from either of them, or from both of them together, to the conclusion that ‘texts are not objectively existing entities’ ”(§5.3.11). But in the end, “Antirealism may indeed be the appropriate attitude to take toward theories of textuality, but whether or not it is so I don’t think Huitfeldt or Pichler have shown that there is any motivation for Antirealism in our experiences with text encoding” (§5.3.17). This is of course an argument that requires support, but it is nowhere to be found.

It seems remarkable that this discussion constitutes the core intellectual articulation of the contemporary discussion of computational textuality. There is literally so much intellectual work in world history that bears on the underlying questions raised by the OHCO literature that it is almost impossible to cite succinctly; and in this respect the shape of Renear’s argument and no less of much of the arguments found in much other writing on the topic remains hard to fathom. Furthermore, this set of observations is not without influence or power, at first simply by displacing more reasonable and respectful models of textuality that might be usefully implemented, but also because of the ways in which it rides roughshod over so much valuable work of the past several hundred years from all over the world and in many different disciplines. In this respect the OHCO thesis represents a strong instantiation of a computationalist ideology, one according to which we should first assume worldly phenomena are computational (meaning that they work the way we imagine computers work), and then wait for exceptions (whether empirical or conceptual) to influence further iterations of our model.

Rather than the Platonism/Pluralism/Antirealism scheme that Renear offers, a more useful typology in this context can be found in the longstanding schism between Rationalism on the one hand, and something we might variously see as Empiricism, Pragmatism, or Antirealism on the other, which we will here call anti-rationalism. One place to put the stake in between these two camps is precisely language itself. According to the first camp, language is very precise and orderly, ultimately based on one-to-one correspondences with things in the real world which account for our ability to know—a means of cognition (rationality) than helps to explain our presence in the world. Putnam and other philosophers call this a “correspondence theory of truth,” according to which a version of Metaphysical Realism is erected that makes the World almost transcendentally Real.3

According to the second camp, language is often very “imprecise” and “disorderly” (if these terms have any precise meaning), based in social and cultural accommodation and history, and more about people acting together (communication) than about cognition. To the degree that language and cognition are intertwined—and this degree is quite large—it becomes difficult to see whether language is “deformative” or “constitutive” of cognitive practices themselves. On this view (and reductively), there is the appearance everywhere of one-to-one correspondence between words and things, but this says nothing about the “actual fact” of such correspondences, whatever that actual fact might be. To both Kant and the late Wittgenstein—two thinkers whose views are certainly worth at least the respect of the OHCO writers—asking questions about “what is” in the sense of “what a text is” are going to seem highly problematic.4 Whether or not there is some sort of ultimate reality of the way things are—what Kant would call the noumena— there simply is no human access to it, nor any human purpose in reasoning too much about how it might operate.

On what we are calling the anti-rationalist view, then, human language can never afford to be fully “pinned down”: human beings can never allow the dictionary definitions of words to exhaust their semantic potential, despite our need to rely on semantic regularity much of the time. Thus, we should never expect schemes to regularize, normalize, segment, or even culturally control language to be wholly effective without implementing similarly effective controls on all other social behavior. It has always been a part of the rationalist dream in particular to “regularize” language and languages to purge them of ambiguity and other “problematic” features, without recognizing that such features are everywhere and indeed constitutive of language.

There is no less a strong cultural symmetry between strong rationalism and cultural prescriptivism about language per se, as in the French Academy, characterized not by the linguistic but by the cultural good that is supposedly done by maintaining supposedly linguistic norms for language’s own sake. We can only be clear about what we mean, say the prescrip-tivists, if we stop giving in to our primitive need to play, and our social need to communicate even when we ourselves are unsure about what we want to say. We must put the planning, individualist mind where base urge had been. This is not what Freud meant by “Wo Es war, soil Ich werden.” It is a particular presumption about how people will and can function in social reality, one that few of the anti-rationalists would share.

Each of these groups has exerted a fair amount of influence over the discussion of textuality, but even more broadly, each has exerted much influence over theories of language. More forcefully, neither one of these groups would implicitly endorse an off-hand distinction between text and language, in the sense that we might expect text to have structuring properties different from those of language per se. The very early Chomsky might well have liked it if language itself turned out to be composed of hierarchically arranged objects, and much of his work was directed at just this sort of demonstration. But Chomsky himself admits that performance, which is to say language in practice, is nothing but “error”—nothing but deviations even from the appearance of ordered hierarchy. If language itself could be characterized in terms of ordered hierarchies, or even (as OHCO-3 would have it), perspectives on ordered hierarchies, we should expect Chomsky’s work to be much less controversial than it is. Chomsky’s Minimalist Program dispenses with any attempt to systematically characterize hierarchies in language, despite his continued fascination with our apparent reliance on hidden hierarchical structures. Furthermore, Chomsky’s opponents in contemporary linguistics—some of whom might be characterized broadly as anti-rationalists—base much of their work on the obvious ways in which language as such does not neatly conform to hierarchical structuring patterns. So if some version of OHCO is going to obtain, it has to be found in something special added into text that is not found in language—this despite the fact that it is not at all obvious how we distinguish between textual language and nontextual language.

From the perspective of the Text Encoding Initiative (TEI) and other language-oriented tools for computer markup, the last thing we should want to do would be to establish beforehand some kind of text/nontext distinction, since a good part of the point of TEI is to mark up linguistic data whose archival or transcription source is in turn oral data itself. Whether the source is modern oral history or indigenous and endangered languages, such spontaneous presentations offer no transparent generic “markup conventions” we could argue might be applied without radically altering “the text itself.” To the degree that computing as a project pitches itself at the whole world and not just the modern office, it is the exception rather than the rule that our “source text” is a stable document with defined generic characteristics, let alone generic boundaries that might persist across time and space. (And of course, in cases where these are clear, the anti-rationalist has no strong objections to using these features for markup purposes.) Derrida has exactly this problem in mind in Of Gram-matology (1976), where he demonstrates repeatedly that a hard-and-fast distinction between written text and spoken language is extremely difficult to maintain under philosophical pressure, and especially once we move beyond the “obvious” contemporary Western context and geography. Prior to questions of genre, form, editorial apparatus, and so on, Derrida can make us question whether we really understand what is at stake in the “clear difference” between marking words on paper and speaking them aloud.

This drives us back to that original intuition that formed OHCO in the first place: that it is possible, desirable, reasonable, or useful to try to describe the structure of texts in markup based on SGML. Remember that this is more than the straightforward truism that “computers can be used to mark documents”: after all, we have nearly 50 years of text processing to prove that point. But it is instructive to think about what Microsoft Word does in this regard: it does mark every document as a hierarchical object with XML-like metadata, but it does not try to say much about the content of the document in that structure. Every document has a title; that title exists in a hierarchical relation to the document itself. It is not an inherent feature of nature that “documents” have titles; that is an imposition of our culture and our software architecture. It is a reasonable accommodation that still causes some problems for authors some of the time, especially as a document goes through multiple revisions. In this example, the Word document is an Ordered Hierarchy of Content Objects: but the hierarchy is formal, thin, and virtually silent about the truly unlimited content it can contain.

Throughout the Microsoft Word or WordPerfect document, text is explicitly marked with PostScript and other formatting-oriented markup languages. This markup is objective because it corresponds to enumerable features of the computer: the font, font size, and so on. If one chooses one can impose further quasi-semantic structures like headers and sub-headers, but the features for creating these are quite rich and flexible, offering more default options than the ones apparent in raw XML, and still often must be overridden by users in manual fashion. The situation is even more complex for typesetting software for books. And yet even this incredibly rich system of hierarchically ordered objects only minimally corresponds to the content of the text itself. It is not easy to imagine Microsoft Word adding value to text processing by adding a suite of semantically oriented tools that will aid the human being in writing his or her text. It is specifically human creativity in text production that the word processor is enabling, and over time it has seemed that in fact most writers prefer text processors to be as formal as possible—that is, to interfere as little as possible with the creation of text content, and to make alterations to text appearance straightforward and consistent. This does not make it sound like even text producers have been waiting for the capability to generate text in terms of the semantic categories it harbors—after all, we already have terrific ways of manipulating those categories, which we call writing and thinking.

Language Ideologies of the Semantic Web

Ordinarily, a discourse like the one on the OHCO thesis might simply pass unnoticed, but in the present context it is interesting for several reasons. Not least among these is the clear way in which the OHCO writers are proceeding not (only) from the Platonic philosophical intuitions which they say drive them, but instead or also from certain clear prevailing tendencies in our own culture and world. The first is, of course, the idea that computers should be promulgated and promoted into every aspect of social life: one sees this ideology in every sphere of commercial activity, and here it is vital not to disconnect computing as a cultural project from its existence as a commercial project.5 From the side of the computing salesperson, computers are always being underutilized, could always do more, could specifically replace stubborn human workers who insist that their work can’t be computerized. From the side of the computing worker, computers should be used more, in no small part because the concentration of knowledge and power in the computer guarantees job security which is by definition otherwise less strong.

In this sense, the OHCO thesis is an offshoot of one of the strangest and yet most typical episodes of our computer culture and especially the strength of computationalist ideologies over and above the capabilities of computers. The rush of excitement that accompanied the development of the Internet—before and then including the web—had students and scholars everywhere preparing plain-text and then HTML-marked text of (especially public-domain) literary and historical texts. These texts, often poorly edited, nevertheless started to look something like a digital archive of important cultural works, and once this appearance became manifest official institutional archivists became interested in the project and started talking a lot about standards—the same kinds of standards that had been proliferating through industry at the time and that continue to proliferate today. Standards-issuing bodies, often tied to industry, like ISO and even W3C, had already been suggesting that HTML should be “replaced” or “supplemented” by a structured data variant, and XML was developed by the W3C as a “simpler” and “less restrictive” offshoot of SGML. The goal of this movement was largely industrial and commercial. Its purpose was simple: to enable accurate point-to-point data transmission for commercial purposes, for example, across incompatible legacy databases communicating over the Internet. The main beneficiaries of the implementation of structured data on the web would be commercial and financial, because commercial and financial institutions are the ones being harmed by their inability to adequately mark up data—not really text—and to use the ubiquitous Internet as a means to avoid the use of private networks.

For this purpose, XML and its variants are really profoundly effective tools, but that does not altogether explain why the web-user community and academics have been so taken with XML. Again, XML really helps to deal with data; the easiest way to think of data is in terms of databases. A database consists of fields with relatively brief amounts of data, and these fields are labeled. The relation of field contents to label is vital and it is the one that XML most easily replicates. But little about language to begin with, let alone “texts in particular,” let alone even business documents, resembles a database. Thus, even where it might well produce profit, vendors are having only limited success with software applications that parse business-related correspondence such as e-mails and provide semantic categorization that is usable by business decision makers. To some extent, the idea sounds intrusive, and the only actual implementation of such strategies is still much more keyword-based and involves monitoring employee e-mails for illegal and anticompetitive activities.

Related to this is the hard distinction between form and content that a database model of text implies. This idea is yoked to the implementation of the Semantic Web itself, suggesting somehow that the problem with the web so far and with HTML in particular has been its “blurring” of the form/content distinction.6 Tags like <b> for bold and <i> for italic are today “deprecated” because they express formatting ideas in what should be a purely semantic medium, so that we should only use tags like <em> and <strong> instead. Of course it is much more ambiguous to the author what these tags will mean in display terms than <b> and <i>: there is nothing at all to prevent implementations where <em> is realized as bold text and <strong> as italic text. The main benefit of such implementations is ideological: it is like strong grammatical prescriptivism in language teaching, ridding the world of “bad” vernaculars. When we step back from such processes, even in formation, we often find astonishing creativity among the vernacular practice. With HTML, that conjecture doesn’t need much proving out: we have the proof abundantly before us. Why has the web community in general moved toward a philosophical perspective that is bent on driving people away from their abundantly productive, if occasionally unruly, practices?

If businesses could profit by marking up the semantic parts of regular linguistic data there is little doubt they would. Amazon.com clearly uses features of semantic structuring in its selling and searching tools, but at the same time its extremely sophisticated system shows the real limits of such systems for people when they interact with text. As far as I know Amazon does not mark the contents of books with rich XML, although this clearly would be within its powers, beyond the Books in Print-based metadata it maintains for each book. Amazon does display something like the Library of Congress subject headings for books at the bottom of the page, a feature I suspect users often do not know to access. University libraries today display those same subject headings, through which mechanism we have finally come to understand that these subject headings are a cataloguer’s dream run amok: for items can be categorized in many different ways, even when this categorization is restricted to a narrow domain, that it is virtually useless as a cross-reference tool. The problems with such categorization schemes are too manifold to discuss in detail, but suffice to say it is often more difficult to search for an anthropological study of matriarchy in rural China by searching on keywords like matriarchy, China, and anthropology than it is to search through the LoC category hierarchy for the multiple headings which would fit the subject.7 Even having

found an initial result, it is usually more effective to follow keyword searches and the references found within results than to follow the semantic subject headings, because they have been created by human beings and like all human linguistic products are highly variable as to production and interpretation.

For a project like the Semantic Web to work, something more than the statistical emergence of properties from projects like del.icio.us and other community tagging projects is going to be needed.8 Despite their higher statistical frequency, these projects suffer even more greatly from the LoC problem, in that precisely what a tag is “supposed” to mean varies from person to person, and ultimately coalesces on what cognitive scientists would call category prototypes—so blogs get tagged with “politics” or “environment” as the most salient tags, precisely the words authors have already chosen as keywords.9 This is apparent even in the standard historical understanding of the success of SGML and no less, arguably, of XML:

The best use of SGML is generally made in big corporations and agencies that produce a lot of documents and can afford to introduce a single standard format for internal use. Besides IBM, early applications of the new technology include the projects developed by Association of American Publishers (AAP) and the U.S. Department of Defense. (Darnell 1997, §3.2)

The focus of XML and its associated technologies is data interchange among large organizations with standardized data objects. It is also useful for tagging and cataloging large catalogs of objects with discrete parts and labels (such as complex military vehicles, where SGML is routinely used in documentation procedures and integrated into the design and development process). These characteristics are precisely those that texts tend not to display, much as they are the ones that language and other creative human activities that are not bound by discrete limits tend to display. This is not a mystery unless baseball is a mystery (because it is analog) while computational simulations of baseball are explicable (because they are fundamentally based in binary computations).

This “truth” might sound like Antirealism or deconstruction, and it is certainly hospitable to both, but it is really not so radical, in the sense that few linguists or philosophers of whatever camp would object to it. Language is highly plastic and ambiguous, even when it appears to be solid and referential. We might be able to approximate schemes for representing certain aspects of linguistic structure, and perhaps semantic structure, as these exist in a given text, but these would always be both theoretical and partial, and they are by no means necessary. They are certainly not necessary to produce “marked-up digital text,” in the sense that I am now writing a text in Microsoft Word that is 100 percent digital and yet in no way marked by human-usable semantic tagging. Future versions of Microsoft Word will decompose files into XML for storage, but this XML will continue to conform to the Word expectations—it will not look for an OHCO-style structure in the document, in the sense that there will be any sort of overarching structure describing the document semantics, and despite the Word document’s structure literally being an OHCO.

This brings us to a more general set of questions, which have to do with the place and appropriateness of XML and similar semantic tagging schemes in humanities computing, and the role and function of digital humanities as a field. It has been quite astonishing to see the speed with which XML and its associated technologies have not merely spread throughout humanities computing, but have become a kind of conversion gospel that serves not merely as motivation but as outright goal for many projects. “Convert the texts to XML,” “everything should be XML-compliant,” “XML provides semantic searching”—to the degree that these cries have been heard within humanities circles they have been surprising. Humanities researchers don’t produce databases: they produce language and texts. Language and texts are largely intractable to algorithmic computation, even if they are tractable to simulated manipulation on the computer. Such manipulation is a primary goal of the computing industry. Whatever one thinks of Microsoft Word, it clearly serves the text-processing (and text-computational) needs of scholarly researchers perfectly well, and will continue to do so. If and when rich semantic markup of text becomes useful for the general public—an event one has reason to anticipate only with great doubt—such tools will surely be incorporated into our everyday tools for text production by Microsoft and other major manufacturers.

Some humanities projects may benefit from extensive XML markup, but for the most part what archives do is limited, precisely because it turns out that best archiving practices show that minimal markup leads to longest life and maximum portability. Microsoft Word allows users to easily create “metadata” for any document via a series of preference dialogues that most users probably do not recognize as Metadata-with-a-capital-M. Thus the recent push in digital humanities for users to become deeply involved with XML markup seems strange. Such involvement is typical of programmer-level involvement with the computer—that is, it is programmers, rather than managers and creative people, who work with XML, even in large business and government organizations. Surely the last thing we want is to say that digital humanists must be programmers. It is great to have some programmers among us, but it is vital to have nonprogrammers as well. Certain trends in computer history suggest a move toward making programmatic capabilities more available to the user, and less restrictively available to programmers alone, exactly through demotic tools like HTML, BASIC, and Perl. The way to do this with XML, I would suggest, is to incorporate it into relevant applications rather than to insist that humanities scholars, even digital humanists, must by definition be spending the majority of our time with it.

Another reason the OHCO materials are interesting is that they, like the general set of XML recommendations from the W3C and associated bodies, are focused on language. There is a great deal of interest in language and what language can do for computers, on how programming languages can be more responsive to human language. What seems so strange about this is that it presents as settled a series of questions that are in fact the most live ones for investigation. We don’t already know how language “works”; we are not even sure what it would mean to come to consensus about the question. In one area in which I do digital humanities work, the creation of digital archives for so-called Minority and Endangered Languages, I routinely face the assumption that XML is going to “mark up” “everything” about the relatively unknown research language. But in fact we almost always end up precisely not trying to do anything of the sort. It is hard enough just to archive “texts” made up of “sentences,” and even this sort of markup implies analytical decisions that must be written up for future users of the archive. No doubt, some researchers may wish to create highly rich, structured-data-based applications driven off of this archival language data, but it would be a grave error to try to incorporate such markup into the archive itself. That would actually obscure the data itself, and the data is what we are trying to preserve in as “raw” a form as possible.

Language: despite its claims to global utility, the web is particularly monolingual. The irony of being offered a remarkable tool at the price of sacrificing one’s language is exactly that with which many in the world are presented today. There are remarkable conceptual, practical, technical, and linguistic challenges in making the World Wide Web truly worldwide. But rather than working on this problem—in many ways part of a great refusal to recognize it—the W3C and digital text communities prefer to work on ways to theoretically expand the linguistic capabilities of the existing web, by wrapping largely English-only lexemes around existing, meaningful text. We have virtually no practical examples of such schemes working to any productive ends, and examples of them failing, not least among these the AI projects that the Semantic Web too nearly resembles.10

Today there is no more iconic exemplar of the web than Google’s main search engine. Google relies on what has become a clear and repeated success strategy in computing: by putting huge numbers of multiple, parallel, redundant, and essentially “dumb” processors together, one can churn huge amounts of raw data in relatively little time, and retrieve essentially any connected aspects of that data in response to queries. This is all done with raw-text searching, even when the data resources may have structured elements, largely because there is as yet no full mechanism for querying a host’s structured data (this is untrue, for example, in many business environments). To Google, and to philosophers, and to linguists, digital text is just text, even if it is XML. XML is just more text. It does not necessarily make it “easier” to find aspects of a text because they are “pointed out” via XML markup; it can make the searching task harder, especially when it is human beings and not machines who are the ultimate users of the search results. To the degree that Google has implemented sophisticated search algorithms that rank closeness of various search terms to each other, hidden markup may impede Google’s own ability to search the text in question, or force it to maintain separate plain-text repositories of marked-up text.

Google’s power, simplicity, and success and its utility for us relies in part on the assumption that much web data will be largely unmarked—except, ideally, for library card-like metadata that is already incorporated into web applications. This makes it difficult to see how searching might be better if text were widely marked up, especially if plain-text versions of texts are not offered in addition to richly marked versions. And unless a scheme is developed to decide which aspects of texts to richly mark in a way that does not simply repeat lexemes already occurring in the text itself, a scheme that itself seems improbable on conceptual grounds, it will be hard to see what additional lexical value the markup can add to the text in question. So instead XML is used as a series of minimal guidelines for metadata specification and formatting, which is fine enough, but in this way not a significant improvement over consistently applied HTML, and much harder for the researcher to understand. It is also used as a de facto way of standardizing web content. If the goal was simply to standardize web formatting of important documents we could have talked about that as an open political question, but it has not been considered.

I see nothing in particular in the W3C proposals suggesting that raw XML should become an authoring tool, and I do not see anything in Oxygen or other XML tools to suggest that they are meant to be used as primary document processors. (When XML or database content management is appropriate, for example in blogs, the XML is kept in the background and used to create a user-focused interface for content production, as in all modern blogging tools.) I see no reason to expect that the coming generations of digital humanists will prefer to write raw XML when writing connected prose; I know few who do. It is hard to imagine how connected prose would benefit from extensive XML markup; it is easy to see how a minimal sort of structuring markup might be useful for standardized institutional storage, but while this is a typical use of XML it is not necessarily in philosophical harmony with the rest of the web. Perhaps as importantly, corpora of language—both textual and spoken—are now a routine part of linguistic practice, and these corpora are not usually richly marked with semantic terms, for reasons already discussed. When such semantic tagging is used, it is usually done for further application processing and for research purposes— which has not yet produced powerful means of accessing otherwise-hidden aspects of the semantics in question.11

Digital text, for the most part, is just text, and for the most part text is just language. Language, a fundamental human practice, does not have a good history of conforming to overarching categorization schemes. It seems not merely reasonable but likely that the best-known humanistic worldviews would resist applications of schemes like the Semantic Web to ordinary textual data, and in particular to focus attention on what humanists have always put first: our responsibility to each other as human beings, and our profound interest in and respect for human creations, including both computers and language. While such a view is compatible with the perspective that embedded markup is itself potentially harmful (as no less a thinker than Ted Nelson has suggested, 1997; also see Smith 2001), it at least leads to what practice itself has shown the OHCO theorists: where semantic markup is concerned, less is more. Where the humanities are concerned, the linguistic imperatives offered by computers and the web seem less promisingly those of markup-based machine communication than the communicative opportunities computers may open for human beings who need them—and no less the avenues of linguistic communication whose existence computers help put into jeopardy.12

Monolingualism of the World Wide Web

Another tempting but inaccurate analogy between programming languages and natural languages can be found along the axis of linguistic diversity. It is no accident, and no less remarkable, that insofar as something akin to natural language “runs” computers, that language would have to be identified with contemporary standard written English. Contemporary standard written English is remarkably effective at projecting itself as typical of “the way language works”; it is also remarkably atypical, both from an historical perspective and in a synchronic perspective, that is, as one out of the world’s 6,000 or so languages. Of the approximately 120 writing systems in active use, only a handful can be said to have developed English-style alphabetic orthography “internally”—that is, much prior to the imposition of direct colonial administration by European countries (Coulmas 1989, 1996; Skutnabb-Kangas 2000; also see Derrida 1976, 1979, 1996a; Go-lumbia 1999; Ong 1977, 1982). Some societies have found themselves compelled to re-write their languages so as to either comport with Western standards or, simply, to become Romanized or quasi-English (Skutnabb-Kangas 2000). Few have been standardized to anything like the Greco-Roman degree (in other words, most non-Western writing systems include many “left over” characters, alternate forms, competing systems, and so on); each of them presents unique difficulties for computers, such that native speakers of any of the world’s non-European major languages experience a routine degree of difficulty in using these languages in the computational environment.13

Few English-only speakers realize the implications of the fact that almost all programming languages consist entirely of English words and phrases, and that most operating systems are structured around command-line interfaces that take English writing, and specifically imperative statements, as their input (Lawler 1999). The extraordinary success of software engineers in India, Russia, Japan, and Hong Kong (among other places) maps onto the metropolises created and maintained by world empires, and correlates no less with the spread of English-style education and enforced English language and orthography (Skutnabb-Kangas 2000). It seems no accident that computers rely on the availability of standardized text and that the availability of persons who are fluent in computer engineering emerge from cultures where English-style standardization is produced and often enforced. This has not resulted in the wide availability of native-language resources, even in the widely distributed alternate orthographies, most especially at the software development level. One might say that, despite the ability of computers to produce documents in Hindi or Japanese, computers and networks themselves speak and operate in only a fragmentary, math- and logic-oriented chunk of English.

This is visible most readily on the Internet. In at least three, connected ways, the web looks like an instrument of multilingualism, but on closer examination seems largely to be organized around Westernized, Englishbased categories and language concepts. First, the HTML for most web documents, the markup which surrounds the document’s content (along with JavaScript and several other noncompiled script languages), is fundamentally in English, so that it is necessary to understand the English meaning of certain words and abbreviations in order to read a document’s source (and in a critical sense, to interpret the document). Second, web servers and web software are themselves usually confined solely to English, necessitating (among other things) English-based CGI (Common Gateway Interface) programs and directories, meaning that the URLs of most web documents are largely in English (in fact, it is only recently that it has been proposed that browsers allow any non-Roman characters in URLs, despite which few web servers actually take advantage of this capability). Related to this is the fact that the entire operating system world is run by English products and with English directory structures, such that most web-based categorization tools (like Yahoo!) are organized around profoundly Western-style categories and category systems—often, the exactly equivalent English titles, viewable as the HTML link for any categories on any non-English Yahoo! website, including most of the pages that have Roman character representations and those that do not.

From the perspective of world linguistic history, programming and scripting languages represent not a diversity of approaches so much as a remarkable extension of an already highly standardized phenomenon: English. This might seem a strong claim, were it not exactly in line with one of the most remarkable features of contemporary world culture, namely, the global replacement of local languages with English and, to an important but lesser degree, other standardized languages (Illich 1980; Nettle and Romaine 2000). We have been taught to think of the computer revolution as a fundamental extension of human thinking power, but in a significant way mass computerization may be more accurately thought of as a vehicle for the accelerated spread of a dominant standard written language. (Computerbased writing may only be less successful, that is to say, than are mass media such as television and pop music as media for spreading prestige English as a spoken tongue.)

At the same time, one of the things computers have also done is to have helped expose and call into question the spread of English and other standardized languages, just as this spread has taken on new power exactly through the spread of computers and the bureaucratic systems of control they bring with them (Illich 1980; Phillipson 1992; Skutnabb-Kangas 2000; Spivak 1999; Warschauer 2002, 2003). The problem with this spread, which must always be presented as instrumental, and is therefore always profoundly ideological, is that it arises in suspicious proximity to the other phenomena of cultural domination toward which recent critical work has made us especially sensitive. I am thinking here of the strong tendency in the West (although not at all unique to us) to dismiss alternative forms of subjectivity, sexuality, racial identity, gender, kinship, family structure, and so on, in favor of a relatively singular model or small set of models. Of course some of this is due to the cultural oppression that goes along with every empire; what in particular seems to be at issue the world over is a loss of cultural diversity and a coordinate loss of linguistic diversity, and the coordination between these two is complex.

By “cultural diversity” here I mean something broader than what is sometimes understood by the phrase—not merely styles of dress or manners but the principles and guides by which the whole of social structure is elaborated, and within the matrix of which identity is crafted. Today we radically underestimate the power of these forces, despite their clear implication in the production of modern society, itself far more uncharacteristic than typical in the tens of thousands of years of human history (Mander 1992). This has become so much the case that we have encased in terms like globality and modernity the apparently inevitable spread of written English and other European languages (Spivak 1999). We have convinced ourselves that, because the people on the ground seem to want to learn these languages, then they must only constitute tools to attain economic status, rather than a deep part of one cultural position that has historically spread as much through explicit imposition as through mutual agreement.

Mass standardization and geographically wide linguistic and cultural uniformity are not necessities. In world linguistic history, there have been any number of periods of deep, longstanding linguistic interchange between high numbers of linguistically diverse groups, where high levels of structural diversity correlate with low cultural diversity over small geographic areas, but higher cultural diversity over larger areas (Dixon 1997; Nettle 1999; Nettle and Romaine 2000; Nichols 1992; Sapir 1921). In other words, there are areas where high linguistic diversity over a wide geographic area correlates with a high level of linguistic and cultural adaptation to local geographic conditions, and a great deal of linguistic variation and change over relatively short periods of time. There even seems to be anecdotal evidence promoting the usefulness of the kind of ecological management exercised by humans in such geographic areas, coupled with any number of attitudes toward technological progress that are quite different from ours (Abram 1996; Nettle and Romaine 2000).

There is simply no doubt that computers and languages are closely tied together, for reasons that are as much ideological as they are intellectual or philosophical. It is hard to understand how human beings could take advantage of computer power without linguistic interfaces of some sort; many anti-computer prejudices hang on in the memories of those who had to interact with computers solely through languages composed almost entirely of mathematical equations (so-called assembly languages, machine languages, and so on) and (to human readers) meaningless symbolic abstractions (as in the physical substrate of punch cards and computer tape). At the same time, these linguistic interfaces exact a significant price in performance and, to a lesser extent, reliability, such that the great bulk of computer programming in the contemporary world—programming for devices that are embedded in larger machines like automobiles, airplanes, manufacturing equipment, and medical devices—is still performed in quasi-mathematical languages by a narrow band of experts.14 In an interesting sense, these computers and languages are more l anguage-neutral than are the great bulk of “proper” computers with which the general public is familiar, but their pervasiveness gives the lie to the view that our computing infrastructure is the only one possible; rather, the “English-only” characteristic of modern computers is clearly a layer that mainly serves social, and therefore ideological, functions.

There is no doubt that the widespread access to computers, and so distributed access to information that is updated frequently, is part of what has made linguists and social critics aware of the situation of so-called Endangered Languages. Computer users for whom the large majority languages are not their mother tongue are acutely aware that computers demand language sacrifice as a price of entry; this is not merely true, as might be expected, of speakers of minority and endangered languages, but even of non-Indo-European majority languages that do have computerized orthographies (like Japanese, Mandarin Chinese, and Korean); it can be no accident that English is rapidly becoming a shared lingua franca across the Indian subcontinent just as computers are spreading, nor is it accidental that Indian writing systems, despite their widespread availability, are much less widely used on computers than one might expect. Like Hollywood films and U.S. television, the computer exerts a powerful attractive pull toward English that is perceived by many non-English-speakers as “modern opportunity.” To succeed in the world, to be something, to be somebody, one must abandon the old-hat, traditionalist models of thinking associated with “home languages,” and move into the modern, technological language of the metropolis, which just means English. In this way computers serve one of the most disturbing and the least-considered powers of globalization: the economic power to tell people that their way of doing things is worth less than culture empowered with modern technology, that their ways of life are stagnant and uninteresting, and that to “make something of themselves” they must “get on board” with modernity, represented by the computer and the significant education computing requires, often (due to children being the ones who feel this pull most strongly) to the real disparagement of whatever they choose to leave behind.

While it may be a coincidence that something like half of the world’s 6,000 languages are said to be “dying out” just as the computer spreads across the world, and this is exactly the frame in which the two events are presented in the literature even by sympathetic writers (see Crystal 2004), it is conceivable that these trends are in fact two sides of a single coin. Networked computing has helped to inform people of the eradication of minority languages, but often in a frame that suggests these languages are already “lost,” and in any case so “backwards” and nontechnological that they are of little use to the modern age. Wikipedia is dedicated to presenting its content in every possible language, but like Unicode it is limited in practice to those languages that have already been “reduced” to written orthographies, and the significant lack of emphasis on making the web operate via sound helps to enforce the focus of these projects almost exclusively on majority practices. The lack of modernization and standardization found in some nonmajority practices often produces a kind of exhaustion in computer scientists; if Southern Indian languages cannot be fixed down to a small number of scripts, why bother with them at all? If the letter forms change so often when writing that one can scarcely determine a fixed set of forms to encode, computationalist ideologies suggest, maybe it would be better to just leave them off the computer altogether.

There are few more obviously disturbing applications of such thinking than in the One Laptop Per Child (OLPC) project spearheaded by Nicholas Negroponte. This project, which promises to give many of the world’s disadvantaged children inexpensive laptops which they can use freely for any kind of education, has much to recommend it and yet on cultural grounds seems highly disturbing. There could be almost no more efficient means of eradicating the remaining non-Western cultures of the world than to give children seductive, easy-to-use tools that simply do not speak their languages. The fact that this problem is so rarely discussed in the literature on Negroponte’s project shows the degree to which computers already have encoded into them a profound linguistic ideology, and the fact that few researchers if any are working on ways to solve this problem before “giving computers to everyone” shows exactly the values that propel much of the computer revolution. Like the Semantic Web, the fact that such resources are profoundly majoritarian is considered entirely secondary to the power they give to people who have so little and to the power such projects create. Of course disadvantaged people deserve such access, and of course the access to computer power will help them economically. The question is whether we should be focusing much more of our intellectual energy on making the computer infrastructure into an environment that is not majoritarian, rather than spending so much of our capacity on computerizing English, translating other languages into English, getting computers to speak, or, via projects like the Semantic Web, getting them to “understand” language for us.

part three

CULTURAL

COMPUTATIONALISM

chapter six

Computation, Globalization, and Cultural Striation

In the late 1960s and early 1970s, Marxist economists outlined a theory that was received with a certain amount of surprise, one that has been largely pushed aside today. The thesis was that despite the appearance of competition, most contemporary global economic power was held by a few, massive, concentrated centers—in short, monopolies. In critiques of Joseph Schumpeter (1942), orthodox pure “free market” capitalist economy, and also of more moderate, statist Keynesian economics, the writers Harry Braverman (1974) and Paul Baran and Paul Sweezy (Baran and Sweezy 1966; Sweezy 1972) suggested that capitalism exerts a continuous pressure, even in apparently democratic societies, toward monopolistic and oligarchical organization, and that the concentration of profit among a small group of institutions makes the systems that operate under the name capitalism nevertheless obey both macro-level laws of competition and at the same time laws of power politics, again demonstrating the resemblance of the modern corporation to the apparently pre-modern institution of the principality. Without a socialist revolution, which like many orthodox Marxists these theorists at times argued was inevitable, Braverman, Baran, and Sweezy specifically worried that a strong antidemocratic impulse in the most powerful corporations was starting to gain traction in the United States and worldwide, even as our culture told us that democracy was emerging everywhere.

Among the most powerful tools for what I will call oligarchical capitalism is the use of large-scale pricing power to manipulate human behavior and the actions of the working class so as to deprive them of real political choice and power, in the name of apparently laudable goals like efficiency, personalization, individual desire and need. While the discourses of these goals are also advertised as part of the computer revolution, when looked at from a perspective inside oligarchical capitalism they can be seen, like the work of Customer Relationship Management (CRM) and Enterprise Resource Planning (ERP) and other software regimes, to be variables that can be directly manipulated at the service of corporate interests. It is no accident that the changing nature of the machine is one of the tropes we see repeatedly in the works of the monopoly capital theorists, just as today the “transition to a service economy” is one of the unexamined truisms of contemporary culture. Despite the obvious differences between what we characteristically understand as physical labor and what we understand as service or informational labor, I am again arguing that this difference may in fact be far less salient to politics at both a personal and community level than the degree to which members of a given institution in fact participate in the management of the organization or are instead subject to strong, hierarchical rule.

Cultural Striation

Among the most widely used of computer technologies with impact not just on business per se but perhaps an even more clear and also covert impact on public life and culture are technologies whose explicit purpose is to striate the contents of public cultural existence, which in virtue of their creation as private, proprietary, and sometimes patented technologies puts them almost entirely beyond the oversight (and often even the knowledge) of the public. Financial, legal, and health care information often exceeds the mere recordkeeping functions of which the public is aware; software applications in these industries enable so-called data mining and other forms of analysis that allow the creation of what can only be understood as new knowledge, which is completely inaccessible to the general public. Yet this information is used exactly and explicitly for the control of public behavior, for the manipulation of populations and individuals as well as for the small-scale provision or denial of services. It is well known that so-called managed care insurance companies or HMOs engage in the denial of medical service to individuals precisely so as to maximize profits, even in cases of life-threatening illness. What is l ess-well-understood is the degree to which computer models have been used to dictate such policies, so that even the inevitable litigation that follows a pattern of service denial is precalculated in the system, and certain classes of individuals and conditions are specifically targeted for denial, not for policy or medical reasons but because they have been statistically identified as the best candidates for profitable service prevention.

Such models are often proprietary and as such inaccessible for examination by the public and academic researchers; the emergence of this circumstance from the historical protection of business secrets is understandable, but this history seems not to have anticipated the degree to which such policies could infiltrate almost every aspect of human life and, in part because of the closed nature of many computer systems, be almost wholly inaccessible for outside examination. Such a problem is only now being recognized in the provision of electronic election services by Diebold, itself a company with a long history in the development of closed, proprietary computational systems. Today, businesses in every industry make use of such systems of predictive analysis, data mining, and customer behavior emulation, all of which raise profound questions about democratic control of government, and the relationship of oligarchical capitalism to the regulatory functions of the State.

Among the few spheres in which such tools are visible for public view, if not control, or oversight, is the use of tools of highly precise statistical striation used throughout the advertising and sales industries under vague terms like “target marketing.” In our age of the rhetoric of multiculturalism, postcolonialism, and identity politics, the overcoded racialist and gendered politics of such systems are strikingly overt, as is the ubiquity of such systems in contemporary circulations of capital.1 In the trade this practice is called “geodemography,” a term coined by Jonathan Robbin, “a sociologist who [during 1963-83] developed a computer-powered marketing technique” (Burnham 1983, 90), largely to predict the behavior of voters in political campaigns. The best-known current provider of such services is a U.S. company called Claritas, founded by Robbin himself, and today owned by the same Nielsen corporation that conducts surveys on media usage. Claritas and its forerunners have used computers to develop a suite of marketing applications which striate the consuming population into statistical aggregates that allow pinpointed marketing, financing, advertising, and so forth. In the popular media these efforts are sometimes referenced in terms of TV shows that “reach the crucial 19-to-34 male consumer,” and so forth, but this characterization deliberately undersells the culturalist and racialist presumptions of contemporary computer-based marketing systems.

Typical among the product Claritas produces is one called Claritas PRIZM, which “divides the U.S. consumer into 14 different groups and 66 different segments” (Claritas 2007). The 14 different groups include Group U1, “Urban Uptown”:

The five segments in Urban Uptown are home to the nation’s wealthiest urban consumers. Members of this social group tend to be affluent to middle class, college educated and ethnically diverse, with above-average concentrations of Asian and Hispanic Americans. Although this group is diverse in terms of housing styles and family sizes, residents share an upscale urban perspective that’s reflected in their marketplace choices. Urban Uptown consumers tend to frequent the arts, shop at exclusive retailers, drive luxury imports, travel abroad and spend heavily on computer and wireless technology. (Ibid.)

“Urban Uptown” can be viewed in contrast to Group U3, “Urban Cores”:

Urban Cores segments are characterized by relatively modest incomes, educations and rental apartments, but affordable housing is part of the allure for the group’s young singles and aging retirees. One of the least affluent social groups, U3 has a high concentration of Hispanics and African-Americans, and surveys indicate a fondness for both ethnic and mainstream media and products. Among the group’s preferences: TV news and daytime programming, Spanish and black radio, telephone services and pagers, cheap fast food and high-end department stores. (Ibid.)

Each Group is further striated into what Claritas calls “segments.” For example, “Urban Uptown” includes five segments (each of which includes a representative graphic icon), among them “04, Young Digerati” and “07, Money and Brains”:

04. Young Digerati—Young Digerati are the nation’s tech-savvy singles and couples living in fashionable neighborhoods on the urban fringe. Affluent, highly educated and ethnically mixed, Young Digerati communities are typically filled with trendy apartments and condos, fitness clubs and clothing boutiques, casual restaurants and all types of bars—from juice to coffee to microbrew.

07. Money and Brains—The residents of Money & Brains seem to have it all: high incomes, advanced degrees and sophisticated tastes to match their credentials. Many of these citydwellers—predominantly white with a high concentration of Asian Americans—are married couples with few children who live in fashionable homes on small, manicured lots. (Ibid.)

Similarly, “Urban Cores” includes five segments, among them “59. Urban Elders” and “65. Big City Blues”:

59. Urban Elders—For Urban Elders—a segment located in the downtown neighborhoods of such metros as New York, Chicago, Las Vegas and Miami— life is often an economic struggle. These communities have high concentrations of Hispanics and African-Americans, and tend to be downscale, with singles living in older apartment rentals.

65. Big City Blues—With a population that’s 50 percent Latino, Big City Blues has the highest concentration of Hispanic Americans in the nation. But it’s also the multi-ethnic address for downscale Asian and African-American households occupying older inner-city apartments. Concentrated in a handful of major metros, these young singles and single-parent families face enormous challenges: low incomes, uncertain jobs and modest educations. More than 40 percent haven’t finished high school. (Ibid.)

While marketing methods like the ones employed by Claritas existed before the widespread use of computers, the close tie between computerization and this kind of heavily striated and statistical method could not be clearer: they are what Lisa Nakamura calls “cybertypes,” which work to preserve taxonomies of racial difference (Nakamura 2002, 29; also see Nakamura 2007)—perhaps they are even elaborated with a particularly striated, hierarchical quality in cyberspace. Mass computerization is what enables such methods to be easily distributed and what allows the underlying data to be collected and analyzed.

The methods and categories employed by Claritas PRIZM exemplify the computationalist view of culture. Despite neoliberal claims to equalaccess democracy and to the cultural power of multiculturalism and antiracist discourse, the PRIZM categories could not be more explicitly racialist. Perhaps more powerfully than the identitarian discourse often attacked by neoliberal mouthpieces, PRIZM sees individuals only and exactly in how they fit into racial, ethnic, and economic groups; Claritas customers purchase the PRIZM products exactly to target their messages (which may be commercial, charitable, or even political) to PRIZM groups and segments, and typically cannot see further down into the database to the individual level. Instead, PRIZM segments are specifically tied to various kinds of commercial and political behavior, and its customers choose to adjust their behavior according to the statistical predictions implied or offered by this and other Claritas software packages.

Explicitly racialized practices like so-called mortgage redlining have been at least partially eliminated by litigation and regulation, but such governmental practices presume corporations operate via exclusively racist operations: that a bank decides to charge higher mortgage loan rates to black people, for example. Governmental oversight has not found a way to manage—if legislators are even aware of them—the complex and highly striated groupings created by packages like PRIZM, despite the fact that they are in some ways at least as culturally disturbing as straightforward racist policies. Because a segment like “Big City Blues” includes not just “the highest concentration of Hispanic Americans in the nation” but also “downscale Asian and African-American households occupying older inner-city apartments,” it is not even clear how EEOC-type legislation could be crafted so as to regulate its operation, unless it simply and overtly targets exactly the practices in which Claritas and other similar companies engage.

Yet the commercial services, financial conditions, and even political circumstances of any individual’s life in the United States (and, increasingly, worldwide) are in many ways determined by services like the one Claritas advertises in PRIZM. Membership in a given Claritas segment is largely (though not exclusively) determined through postal zip code; if one belongs to a given segment in aggregate, there is of course no way to adjust one’s membership in the group, nor could there be. Few Americans are even aware of the PRIZM segments or their power over so much of their lives, much less the other systems that are not even advertised to the public—in fact it is a testimony to the business-internal existence of such powerful computational tools as PRIZM that even a portion of it is exposed to the public via advertising, considering how overt its implementation of bias toward various races, genders, nationalities, and ages. PRIZM is just a tool to striate the U.S. population at a level of sophistication beyond that which most human beings can conceptualize without computational assistance, and in this sense both its existence and its effects can be and have been essentially hidden in plain sight; like the complex operations of airline ticket pricing software, it has grown so effective that its functions can be exposed through their effects with little fear of public reprisal.

As a type of product heavily used throughout many global institutions, PRIZM represents the kind of oligarchical power computers help to instance, concentrating power and pushing it upwards while reducing large masses of smooth data into tractably hierarchical, striated layers. These reductive striations become in some important way more real than the underlying data they represent, even if, as in the PRIZM case, what is represented is precisely people’s everyday social behavior. Like the financial models propagated by banks and companies like Fair, Isaac, these representations can be said to have more commercial and even political salience than do individuals or explicit political groups, since individuals and even most groups have little access to them. To be sure, political parties in particular take great advantage of such tools, and there can be little doubt that they have been effectively used so as to predict and statistically manipulate political behavior, especially voting behavior, in ways that would seem quite foreign to the average citizen. Since both parties theoretically have equal access to at least some of these tools, one can argue that they are somewhat equally distributed—but at the same time the lack of awareness of the general public about such tools and their use in politics and public institutions suggests that they continue, today, to be used primarily for the enrichment of the power elite (Mills 1956).

Computationalist World History

Despite the obscurity of the processes used by companies like Claritas and software packages like ERP and CRM, to some extent their operation, since it is characteristic of so much computation, is readily visible to and even widely used by members of the public. In fact, within a wide variety of computer games, one can see a process exactly isomorphic to such software applications, in which quantified resources are maximized so that the player can win the game. The most direct example of such games are the so-called Real-Time Strategy (RTS) games, in which players build kingdoms by supervising the growth of basic resources such as food, stone, wood, and so on. By using “villagers” or “drones” to maximize the collection of such resources, the player attempts to outwit the similar efforts of his or her opponents. Such games reveal exactly the oligarchical-monopolist, Statist, even fascist politics at issue throughout the computerized world, its reductionist conception of social process and politics, the intertwined nature of economics and politics within the computer world, and its particular conception of how culture relates to the rest of the social world.

Playing an RTS game such as Warcraft, Starcraft, Civilization, Alpha Centauri, Age of Empires, Empire Earth, or any of a number of others, one experiences the precise systemic quality that is exploited for inefficiency in ERP software and its particular implementations (see Chapter 7). One simply sets the “difficulty level”—or degree of economic efficiency—to the degree demanded by the rest of the business system. This is of course a profit level determined by the various constituents of a company, most especially its financial determinants including its accountants and accounting software. Of course the values of some elements of the ERP formalization are bounded within limits set by the real world; presumably a car company is better off if it tracks the precise number of cars it is actually producing, windshields available, and so on—presumably. But it is clear that the values of these variables are set not by absolute, objective reality but by human beings whose ability to meet the demands of quantification are limited; companies today seem to be failing in no small part due to the ability of human beings to deceive themselves about the objectivity of the values ascribed to these abstractions. Though ERP runs on the logic of the game, winning at running a business is not the same as winning a game. Maximizing human social satisfaction is a goal we hardly understand, and it is not something we talk about often; in RTS terms, a “satisfied system” becomes static and uninteresting very quickly. This would seem to be somewhat the opposite of how we would like society to be arranged; we would seem to want our society to be as satisfied as possible, with as little conflict as possible between members (on a brief analysis).

Much of the time, such games are played against computer or so-called Artificial Intelligence opponents, avatars of the player represented by the computer, often representing possible real-world strategies. To a neutral observer, the play of an AI player and a human player are virtually indistinguishable (that is, if one watches the screen of a computer game, rather than the actions of the human player). What we call AI has among its greatest claims to successes within the closed domain of computer games, inside of which AI opponents might even be said at times to pass a kind of Turing Test (or perhaps more appropriately, human beings are able to emulate the behavior of computers to a strong but not perfect degree of approximation). Of course this is mainly true depending on the difficulty level at which the game is set. Because the world-system of an RTS game is fully quantified, succeeding in it is ultimately purely a function of numbers. By setting the difficulty level high enough, one can guarantee that the computer will win: it is simply possible to apply more computing power to a simulation than a human being can possibly muster.

One does not, at any rate, win an RTS game by maximizing efficiency; one wins the game, just as much, through the armies and destructive weaponry that seem unavoidably attached to almost every RTS model, which often represent explicit iterations of historical and science-fiction conquest narratives. Much of the game architecture was developed with either an “adventure” model (in which the hero gathers resources in order to fight successively more powerful opponents), or a so-called “civilization” model, in which either abstract (often space-based) civilizations, or actual historical civilizations from world history, fight with each other using successively more powerful weapons. In Civilization or Age of Empires, the user takes the role of an historical “conqueror” such as Napoleon, Cortes, or even, in one version, Montezuma; the increasingly elaborate and technologically sophisticated wars one produces lack precisely the unique contingency of history itself. Users play these games over and over, “addictively,” playing one side and now the other, pitting two Aztec kingdoms against each other, displacing the unconscious of history itself onto the unconsciously driven will to power of winning the game through acquisition and violent, oppositional destruction. Somehow the rationality of software representation leads not to the world of choice but to the inevitability of realpolitik. Instead of a paradise of informatic exchange, the world of computer representation is a bitter and political one of greedy acquisitiveness, callous defeat, and ever-more-glorious realizations of the will-to-power. It seems inevitable that the result of such pursuits is most commonly boredom. Like a sybaritic emperor, the unconscious desires served by such an approach are never satisfied, and can never find enough of the resources and excitement they need. The game thus “covertly” satisfies the unconscious, but offers only traumatic repetition as satisfaction, more consumption as the solution to the problem of being human.

Among the most paradigmatic of RTS games is Microsoft’s Age of Empires, which has been issued so far in three major iterations. In Age of Empires, as in similar games, a player begins by choosing which race/-civilization/nation to play. While some games allow a certain amount of customization to such details, in Age of Empires these choices are programmatically constrained. One chooses a nation or race based in part on the graphical appeal of the animations used to represent the group, but also practically because each group gives the player certain characteristics (benefits and drawbacks) that are programmatically tied to the initial choice. While such choices are the precise domain of discussion in nearly all fields of humanities and education today, within the computerized world they are presented with a certainty that few thinkers of any stripe would today endorse. Again, population thinking within the striated computer world trumps individual variation.

In the full version of Age of Empires II (including its Conquerors expansion), for example, the player chooses from one of 18 “game civilizations,” including such textbook standard groups as Britons, Huns, Vikings, Japanese, Koreans, Chinese, Aztecs, and Mayans. Of course for the sake of efficient gameplay such civilizations are restricted to ones that established geographically large scale empires, and also present most nation-states as internally consistent; while there are both Mongols and Chinese in the game, there is no question of representing the Mongolian minority that exists in the non-Mongolian part of China, or of politically problematic minorities such as Tibetans and Uyghurs, or of the other non-Han Chinese minorities (e.g., Li, Yi, Miao). There is no real ability to depict intergroup interactions, hybridization, or blending; the existence of racialized and national groupings is presented as a kind of natural fact that cannot be questioned or changed in the course of gameplay, as if this were also true of the global history the game will display: as if from the beginning of history until the present day there simply were groups with names like “Briton,” “Hun,” and “Byzantine” whose coming-into-being was outside of the history that, in historical fact, produced them.

The main reason for players to choose an ethnicity or nation in the course of the game is to realize certain benefits and drawbacks that affect gameplay as a whole. For example:

• • Britons

• • Foot archer civilization

• • Shepherds work 25% faster

• • Town Center costs 50% less wood

• • Foot Archers gain + 1 range in Castle Age and an additional + 1 in the Imperial Age (for + 2 total)

• • Team Bonus—Archery units produced 20% faster

• • Koreans

• • Tower and naval civilization

• • Villagers + 2 Line of Sight

• • Stone miners work 20% faster

• • Tower upgrades free (Bombard Tower requires Chemistry)

• • Towers range + 1 in Castle Age, +1 in Imperial Age (for + 2 total)

• • Siege units have + 1 range

• • Aztecs

• • Infantry and monk civilization

• • Start with Eagle Warrior, not Scout Cavalry

• • Villagers carry + 5

• • All military units created 15% faster

• • Monks + 5 HP for each Monastery technology

• • Team Bonus—Relics + 33% gold

• • Mayans

• • Archer civilization

• • Starts with one additional villager and Eagle Warrior (not Scout Cavalry) but less 50 food

• • Resources last 20% longer

• • Archery Range units cost -10% Feudal Age, -20% Castle Age, -30% Imperial Age

• • Team Bonus: Walls cost -50%

Much like the segmentation employed by Claritas PRIZM, these categorizations reduce the complexities of social life to measurable quantities; even further, they employ capability-based labels that define civilizations in terms of particular technologies and characteristics, so that, for example, the Britons can be characterized as a “foot archer civilization.” These labels are tendencies, regardless of the degree to which they may or may not apply to historical entities—nations that never developed sea power, for example, will nevertheless develop sea vessels and weapons in at least some editions of Age of Empires and similar games.

The whole idea of applying modifiers and bonuses to each civilization hangs on the notion that there is one central history of cultural development that is at least in part common to each nationality and race. In the Age of Empires model, at least, this is strictly true: each nation starts in the “Dark Ages,” when it has access only to materials made of wood and something like smelted iron, so that it can use swords; then advances to the “Castle Age,” when it can use gold and stone to create masonry buildings and learns technologies like gunpowder; and then to the “Imperial Age,” when it gains putatively cultural attributes like religion (Monasteries) and higher learning (Universities). The appearance of such buildings, like the animations for citizens, varies just a bit from civilization to civilization, but the underlying progressive history remains identical for each civilization, giving the impression that this progressivist idea of history is universal and was experienced in history by each civilization. As such, Age of Empires and similar RTS games do not merely argue but demonstrate that human history is unified, progressive, and linear, and that in particular the development of technology stands as the ultimate end of civilization. Culture per se is seen to be no more or less than decoration—a kind of costume, much as culture is seen in the neoliberal versions of museumized cultures the world over that have failed to fully integrate into the noncultural, central (and typically Western) view of technologically developed humanity.

In the most typical game played in Age of Empires, there are exactly two ways to win. Since every culture is by definition militaristic, committed to developing a similar kind of army and navy, one can choose to defeat each other nation and thereby achieve military victory (alliances are possible, but are downplayed significantly in this game). One can also win by virtue of overwhelming economic development: by setting the difficulty level of the game low enough (or by playing against human opponents of low-enough skill) that one can generate more economic resources than any other player, thereby creating and sustaining military power beyond the capability of any other player.

In addition to military conflict, the main gameplay of Age of Empires and other similar RTS games is twofold: the accumulation of resources and the development of city-states. In Age of Empires II resources come in four types: food, wood, stone, and gold. Each civilization may create dozens of basic citizens, which other than being gendered are otherwise indistinguishable, and similar but for clothing details across each civilization. These citizens (unlike in some more complex games like Civilization) never protest about the repetitive nature of their work; they simply devote themselves without rest to the collection of resources or to the erection and repair of buildings. The player’s actions are limited to directing the citizens to a resourcegathering or building task; unless interrupted by military skirmish, the citizens simply carry out the task perpetually until the game ends. While gathering of wood, gold, and stone is limited to a single function, food can be accumulated through hunting (of sheep and deer conveniently located at random points on the world map), fishing, and both marine and terrestrial farming. Again, these activities are completely identical no matter the civilization one is playing, and food products are not really identified at a detailed enough level to identify what food is being produced nor how it is being consumed.

Instead, all the basic resources produced by a city are used by the player in the service of creating ever-more elaborate buildings and ever-more sophisticated soldiers; the bulk of the program’s “technology tree” (the sequence of available upgrades for various devices, implements, and characters in the game) is devoted to military technology. While some attention is paid to the racially identified attributes for a given civilization—so that a civilization specializing in siege weapons will have more elaborate siege devices and more technology upgrades available—the general rule of thumb is that every basic unit and every basic technology is available to every civilization, and that each civilization passes through exactly the same phases of development as it advances in the game. Development ends with the Imperial Age, a triumphal goal toward which every civilization aims.

While its developers would surely claim that such features are a consequence of gameplay necessities and the use of the tools available at hand, there can be little doubt that a game like Age of Empires II instances a com-putationalist perspective on world history. According to this view, history is a competition for resources among vying, bounded, objectified groups; human beings are all driving toward the same accomplishments, even if our racial and national backgrounds tend to pull us with some force in certain directions; everyone would like to rise to advanced levels of technological accomplishment, which can be labeled with a term like “Imperial”; every society, if given the opportunity, would choose to become an empire. In this view there is no hybridization in Homi Bhabha’s sense; there is no talking back to power; other civilizations are simply eliminated as their settlements are replaced with those of the winning side.

More than this: the pursuit of historical change is the pursuit of striated power, realized as both an individual “leveling-up” and a societal achievement of historical epoch, themselves licensed by the accumulation of adequate resources, which are always channeled back into an even more intensive will-to-power. There is no question of corruption or self-interest; the State simply is the user, in a position of mastery that is not a difference of degree but of kind. The user is a godlike presence, controlling almost everything in the game but wholly opaque with regard the game narrative, such as it is. Despite being the most important single presence in the game, the godlike-princelike user does not engage in any of the activities of his or her characters, but instead revels in the total activity of all his or her minions. A true Hobbesian Prince, the user of Age of Empires allows his subjects no interiority whatsoever, and has no sympathy for their blood sacrifices or their endless toil; the only sympathy is for the affairs of state, the accumulation of wealth and of property, and the growth of his or her power.

Age of Empires II was one of the first mainstream games to include not just some apparently non-Western empires (Korea, China, Japan) but also indigenous cultures that, whatever their internal imperial ambitions, come down to us today as the subjects rather than the architects of colonization. In a 2006 expansion to Age of Empires III, called The War Chiefs, players can choose from three indigenous civilizations: Aztec, Iroquois, or Lakota. These figures are depicted with careful attention to detail and admiration, as if to say that if only their war technology had been a little better, perhaps we would today be living in an expanded Iroquois Confederacy instead of the United States—with a kind of condescension appropriate to history’s winners. The three indigenous groups are advertised as follows:

The Sioux

With the discovery of gold in their sacred Black Hills, the Sioux had sporadic conflicts with the new immigrants, leading to the Indian Wars and eventually the Battle of the Little Big Horn. The Sioux Nation consists of three geographically distinct divisions of peoples who speak a Siouan language: the Lakota, Dakota, and Nakota. In the language of the Sioux, the names Lakota, Dakota, and Nakota mean “friends.”

The Sioux were famed for breeding and training horses. Military advantages of the in-game Sioux civilization are primarily cavalry based. The Sioux strike hard and fast.

The Iroquois

The Haudenosaunee, or Iroquois, formed a League of Five Nations long before Europeans arrived in North America (a sixth Nation joined later). The Iroquois Confederacy had a constitution with rules for the selection of war chiefs, guidelines for council meetings, and even suggestions for proper oratory at funerals. Benjamin Franklin expressed great respect for the Iroquois Confederacy. During the American Revolution some tribes sided with the British, while others sided with the colonists, effectively dismantling the once-powerful Confederacy.

Equipped with artillery and siege weaponry, the in-game Iroquois civilization can mount a powerful but slow-moving assault.

The Aztecs

At their height, the Aztec represented the most powerful civilization in Mesoamerica. The Aztec constructed great cities, most notably Tenochtitlan on the site of modern-day Mexico City. Described by some early Spanish as grander than most European cities, Tenochtitlan’s advanced infrastructure included temples, markets, and canals. In Aztec society, membership in the calpulli established each individual’s religious and secular schooling, as well as warfare training. The men of a calpulli served together in battle and on numerous public works projects. The in-game Aztec civilization is based around a strong infantry consisting of several units, including elite infantry.

Such descriptions continue to reinforce the computational view of history and progress. Civilizations are described exclusively in terms of economics (especially resource accumulation), military technique, and contribution to modern Statecraft. Cultural characteristics are reduced almost exclusively to military attributes, which in turn reduce to the attribute characteristics realized inside of the game; the user is implicitly located above all such characteristics, a technological Prince of whom all the world’s peoples are unaware and yet to whom they are all subject.

There is more than a passing similarity between race and civilization as it is realized in RTS games like Age of Empires and the Claritas segmentations: they are both faces of the computationalist view of culture, in which race and the State are not at all obliterated but instead essential attributes of the person at a group level which no individual can transcend. The only way to move out of a Claritas segment is to change where one lives, presumably by also changing economic situation as well; but there is little (if any) way to change the internal characteristics of the striated segments. There is no way to change what it means to be a Sioux or Briton or Korean in Age of Empires, other than by winning or losing the game; there is no means to alter the fundamental capitalist basis of economy or the fundamental nature of the State or the individual’s relationship to it—despite the fact that in, for example, both Lakota and Iroquois societies in historical reality, it seems arguable that something like a relatively more nomadic and thereby less rigidly emplaced State were in strong evidence.

In Age of Empires the entire world is reconceptualized as fully capitalist and Statist from the outset, as if capital accumulation and Western-style technology are inevitable goals toward which all cultures have always been striving. “Under” each civilization is something like bare human life without culture, as if culture and humanity can be separated, and there can be little doubt that this bare humanity corresponds in some important way to the existence of the godlike Player who manipulates the game behind the scenes. Only to him or her are attributes changeable, learnable, flexible; the world he or she views is one of complete striation, a story already told and waiting for some kind of inevitable end. Those who don’t get on board with the story simply are not worth representing at all, and even if we pay tacit obeisance to those cultures that were, in fact, the subjects of empire and colonization, it is only in so far as to explain why, in the crudest economic and racialist terms, they do not sit where “we” do today, in the abstract and godlike position of the subject (rather than object) of History.

Writing of an RTS-based world history game that has substantially more cultural flexibility than Age of Empires, the Civilization series developed by Sid Meier, Alex Galloway suggests that “the more one begins to think that Civilization is about a certain ideologically interpretation of history (neoconservative, reactionary, or what have you), or even that it creates a computer-generated ‘history effect,’ the more one realizes that it is about the absence of history of altogether, or rather, the transcoding of history into specific mathematical models” (Galloway 2006, 102-3). “ ‘History’ in Civilization is precisely the opposite of history,” Galloway writes, “not because the game fetishizes the imperial perspective, but because the diachronic details of lived life are replaced by the synchronic homogeneity of code pure and simple. It is a new sort of fetish altogether. (To be entirely clear: mine is an argument about informatic control, not about ideology: a politically progressive ‘People’s Civilization’ game, a la Howard Zinn, would beg the same critique)” (ibid., 103). Despite the fact that Galloway sees some of the inherent ideological commitments in the RTS genre, this view gives in to exactly the neoliberal/progressivist/neoconservative view that technology can stand outside of history and that computational processes are themselves synchronic, abstract, and deserving of new analytic methods and frames to be understood—exactly what I am arguing here is untrue. Surely a “People’s Civilization” computer game might in fact deserve critique along the lines of informatic control; but it is to place computers and ourselves outside of history not to see that there is an historical reason why the successful and popular RTS games are not of the Zinn variety. Computation and striated analyses; essentialist understandings of race, gender, and nation; and politics that emphasize mastery and control do not merely walk hand-in-hand: they are aspects of the same conceptual force in our history.

Also writing of the Civilization genre, McKenzie Wark speaks of a “double development, which at one at the same time deepens and proliferates lines of the possible and the actual,” which

can be called America. It is what is both desired and feared under the rubric of that name, which no longer marks a particular topic but stands for the very capacity to mark and manage space itself, as topology. When playing Civilization III, it doesn’t matter if the civilization you choose to play is Babylon or China, Russia or Zululand, France or India. Whoever wins is America, in that the logic of the game itself is America. America unbound. (Wark 2007, 75)

Wark comes closer than Galloway to seeing the inherent logics and politics of RTS games and the history they embody, but the proper name America is arguably both too specific and too general to capture what is at issue in the culture of computation. It is too specific because America is not pure computation and not the exclusive progenitor of computational views; at best computationalism can lay claim to a segment of American imperial thinking and history (something like the Federalist perspective associated with Alexander Hamilton); in this sense computationalism, like the conceptual authoritarianism of which it is but a contemporary avatar, has a deeper connection to and association with the development of everything we call the West (and, as RTS games suggest, everything that we call Empire in our world) than the name America would suggest. It is too general because, of course, computationalism is today distributed across varied geographies, nations, and groups, and finds not only advocates but dissenters in all of those societies. Races, nations, and genders are both striated and dismissed as decoration in RTS games in the name of a putative technologically driven ur-humanity to which not just America, but other political groups lay claim. In some sense, at least, the impact of computationalism on the world is much less like the historical development of America, and much more like the worldwide reemergence of principality like domains of imperial control today associated with medieval and pre-modern political forms (Hardt and Negri 2000, 2004). A more apt name might be “neoliberalism.”

Empires and Computation

Today the world is unevenly covered many times over by entities we routinely refer to as multinational corporations, just as much as and in some ways more completely than it is covered by State-based governments. The emergence of the multinational corporation is itself both a recent development and a return to formations that had been effectively brought under democratic control only with the most intense and directed efforts of both voters and workers. Despite the extreme claims of computational advocates to a general democratization of all forms of information and political control—which on some readings must be identified with something like the means of production for an information age—it is much less noted that the resurgence of neoliberal economics and politics has created a vast array of entities with extra-State political power and control, no longer subject to either the internal (labor union) or external (governmental) oversight that had been the hallmark of liberal political economy.

While at one level it may be harmless enough to imagine that the tremendous power placed at the hands of many individuals offers a democratization of culture, at another level such claims smack of something between naivete and outright, condescending colonial thought—just, in fact, the intellectual regime we tell ourselves we have surpassed in this nominally postcolonial age. Among the most influential accounts of such cultural transformation is the New York Times columnist Thomas Friedman’s (2005) The World Is Flat: A Brief History of the Twenty-First Century. Exactly as its presumptuous subtitle implies, Friedman’s book outlines a series of cultural-economic changes whose shape can only be dimly glimpsed today, a messianic vision in which the “essential impact of all the technological changes coming together today” is that “people can plug, play, compete, connect, and collaborate with more equal power than ever before” (x). While the overt topic of Friedman’s book is social change, its substance is almost exclusively technological, and it evinces everywhere a profound technological determinism, simply arguing that inevitable technological changes lead inextricably to social changes of a radical and egalitarian sort. Friedman suggests that what he calls “Globalization 1.0 and 2.0 were driven primarily by European and American individuals and businesses... . Because it is flattening and shrinking the world, Globalization 3.0 is going to be more and more driven not only by individuals but also by a much more diverse—non-Western, non-White—group of individuals. Individuals from every corner of the flat world are being empowered” (11).

The trope of empowered individuals, as if they are unconnected from the political and economic and cultural institutions in which they are embedded, is one of the most familiar of our recent times, and we see again and again that this trope does not indicate just how those who already have relatively less power will have any more political influence than they do now, when those with the most power already also have their power increased. Looking simply at computer technology, it seems arguable that this kind of information-technology-based power boost follows a kind of logarithmic curve: while the relatively disempowered are certainly enfranchised by getting on the Internet and getting access to a relatively basic computer, the powerful are themselves radically reinforced via high-speed Internet access; multiple desktop, laptop, and miniature computing devices; computational systems in their automobiles and home entertainment systems, etc. Nobody would, in fact, doubt that the very rich and powerful are among those who take the most advantage of new technologies. How does a 50 percent boost to someone in the bottom rung of society increase his or her power or influence when the top 10 percent is simultaneously experiencing a three- or fourfold magnification of their own power and influence? How much influence and power are there in our world?

Not only does Friedman’s analysis simplify what he calls Globalization 1.0 (the colonial project from 1492 until about 1800) and Globalization 2.0 (approximately 1800 to 2000) in terms of the cooperation of ruling elites the world over with powerful interests, but his conception of the radical change underwritten in our current age relies specifically on quasi-magical transformations engendered by computer technology whose mere presence—in some cases, despite its obvious uses and affordances—through some mysterious mechanism makes our world much more equal than it is. Friedman’s argument hinges on what he calls “ten forces that flattened the world” (51), all of which are innovations in Information Technology, and few of which transcend majoritarianism, concentration, and segmentation, and Friedman’s rhetoric suggests there is no need to explain how his ’’forces” lead to the radical democratic organization suggested by the apparent meaning of the world being “flat” (and despite the strange and perhaps unintended suggestion that our world is subject to exactly the same rule of empire as was that of Columbus, whatever the operative geometric world pictures may have been in 1492).

Thus the first three of Friedman’s “forces” is the introduction of the personal PC with Microsoft Windows; the second is the global connectivity that is approximately coterminous with the rise of the world wide web, but for Friedman also entails the 1995 commercialization of the Internet and Netscape going public; and the third is the introduction of what he rightly calls “work flow software,” again almost exclusively a corporate innovation. Sounding virtually like a software advertisement, Friedman writes:

Boeing wanted it so that its airplane factories in America could constantly resupply different airline customers with parts, through its computer ordering systems, no matter what country those orders came from, and so its designers could work on planes using airplane engineers from Russia to India to Japan. Doctors wanted it so that an X-ray taken in Bangor could be read in a hospital in Bangalore, without the doctor in Maine ever having to wonder what computers that Indian hospital had. And Mom and Dad wanted it so that their e-banking software, e-brokerage software, office e-mail, and spreadsheet software all would work off their home laptop and be able to interface with their office desktop or BlackBerry handheld device. And once everyone’s applications started to connect to everyone else’s applications, work could not only

flow like never before, but it could be chopped up, disaggregated, and sent to the four corners of the world as never before. (83)

This is supposed to be an account of the world changing away from corporate power, but in practice it isn’t even necessary for Friedman to demonstrate that corporations are getting less powerful; our current rhetoric, perhaps following U.S. popular and judicial usage, suggests that powerful corporations just are powerful individuals. While Friedman’s apparently middle-class “Mom and Dad” certainly benefit from the availability and convenience of computerized tools, the main benefits Friedman describes empower centralized corporations to distribute their own work widely, and in some sense even more fully subordinating their workers to management.

Surprisingly, though, despite the ways in which Friedman’s first three benefits replicate colonial logic and largely benefit large corporations, the rest of his ten “forces” are even more explicitly focused on corporations. In fact, they are largely devoted to exactly the kinds of business-process tools discussed in Chapter 7: ERP, CRM, and other supply-chain management software. Friedman champions open source software, especially Linux, but primarily because IBM realized that it could profit from the distributed community that had created it (which is itself largely made up of corporate employees working without compensation in their free time); arguably, we still talk about Linux precisely because of its utility to profitmaking corporations. Yes, Linux was created through distributed means, but how the outsourcing of work to essentially donated labor contributes to economic or cultural flatness remains a mystery. Then Friedman champions the “withering global competition” (149) enabled by WTO-backed treaties and IT-fueled communication via what he calls “offshoring”; the supply-chain management techniques used by Wal-Mart: “as consumers, we love supply chains, because they deliver us all sorts of goods—from tennis shoes to laptop computers—at lower and lower prices and tailored more and more precisely to just what we want” (155), despite the fact that “as workers, we are sometimes ambivalent or hostile to these supply chains, because they expose us to higher and higher pressures to compete, and force our companies to cut costs, and also, at times, cut wages and benefits” (ibid.).

Here it is plain to see the rhetoric of IT-based transformation giving way to the most traditional, neoliberal, high capitalist claims: what has been flattened via IT is not at all individual access to culture, economics, or political power, but rather the “playing field” for capitalist actors. Thus the flat world of Friedman’s title gives way to the familiar ideal of a completely free market, constantly changed by Schumpeterian “destructive transformation” which workers must simply endure; there is no question of worker rights, labor organization, or in fact any of the social values that, for at least much of the first half of the twentieth century, characterized the reformist capitalism in the United States and Europe. What is flattened in Friedman’s program is not human interaction at all: it is the marketplace, and with it, labor. Just as Braverman suggests in Labor and Monopoly Capital (1974), technology may make an individual worker more powerful, but it simultaneously gives capital and the corporate manager even more power, both for the generation of capital and over the workforce itself: “it is of course this ‘master,’ standing behind the machine, who dominates, pumps dry, the living labor power; it is not the productive strength of machinery that weakens the human race, but the manner in which it is employed in capitalist social relations” (Braverman 1974, 158).

In a series of incisive and often scathing reviews of the claims for technology and their relation to political power, a number of U.S. technology historians (see especially Adas 2006, and Headrick 1981, 1988; also see Wright 2005) have begun to trace out how the notion of technological progressivism has been especially critical for U.S. identity and how it has promoted this progressivism as a necessary correlative of democratic politics over the globe. Along the way, other approaches to understanding and measuring the worth and value of cultures have been summarily dismissed, usually via the power of attraction rather than explicit critique. Promoting simply the manifest capabilities of modern technology, the avatars of U.S. imperialism implicitly disparage less technologically progressivist ideologies, with the powers of fascination and economic leverage to support them. Thus those groups in East and South Asian societies that also become fascinated with technology often implicitly (and often explicitly) begin to endorse exactly the same pro-gressivist ideology, putting something that looks like “America” to much of the world at the top of a technological heap.

Yet in our current climate one wonders more and more if it is America, or the United States, or any specific nationality that has placed itself at the zenith of technology, or whether, viewed more subtly, a different kind of power has emerged to some degree of its own making. Rapid change in communication technologies and transportation, in particular, have made the geographic reality of nation-states far less meaningful than they once were. Changes in U.S. and worldwide corporate law, along with these technological changes, have allowed the so-called multinational corporation to be much more than it ever was. Even in the days of the State colonial companies, geographic reach was often accompanied by communicative lag, thus allowing time for Bhabha-style colonial ambivalence to develop in colonial outposts, at least to a limited extent.

Today’s corporations no longer need to wait for information to percolate in from the remotest outposts; information is not merely solicited but actively collected from every point on the Earth. Where once much of the Earth was largely “smooth” terrain to which lines of segmentation might be intermittently applied, today all of the Earth’s land mass, and a great part of its waters, are constantly surveilled by electronic monitoring, all of which inherently places locations on a single, global grid. The grid is so unquestioned and so available now that even mass applications are made available for individuals (see especially Google Earth), though without the analytic and surveillance tools to which corporations and governments have access. There would seem to be no position at all from which to question whether it is desirable or even ethical to persistently map every square inch of global terrain and make it available for electronic processing; since the benefits of such a scheme are so apparently obvious, only cranks or luddites might stand in opposition to them.

Alternate conceptions of culture and politics, which have at times receded so far into the background as to be nearly invisible, might cause one to wonder about such assumptions. Does technology really drive human development? Are we unable to choose what technology will do or should do, and instead must follow technological leads wherever they take us— with “what can be done, shall be done” as the whole of the law? Clearly that is the law of today’s culture, from biotechnology to computer technology and even to industrial development, and the efforts of citizenry not directly involved in any given industry to monitor, control, or contain such efforts is met with extreme levels of political and media resistance; the opposition to the nearly unanimous scientific consensus about global warming suggests that our technological progress may make us more, not less, susceptible to deliberate ideological control.

Government, deliberately denuded in part through a series of ideological control measures applied over decades by corporate interests, today exists as much to license and pave the political way for corporate action as it does to constrain that action. And since it has been deliberately depowered, there is no longer a central authoritarian or even oligarchical ruler to worry about this loss of power. Instead, that power has been transferred to the corporations themselves, which appear as a series of globally overlapping quasi-totalitarian entities engaged in a “cold” but nevertheless endless war over resources. The entities fighting this war are nearly indistinguishable; as multinational corporations merge and acquire, divest and split, any sense of a “core business” often vanishes, so that all that is truly left of actual business practices is what is commonly today and perhaps rightly called a brand, and which might as well be understood as a national flag or heraldic shield. This sign to which employees are expected to give every aspect of their lives but for their actual blood is one that brooks no disloyalty whatsoever, and under which the employee loses nearly all of the rights governmental theorists like the founders of the American state thought critical to the well-being of any citizenry.

In 2003, one of the former editors of the Harvard Business Review published in that journal an essay called “IT Doesn’t Matter” (Carr 2003; also see Carr 2004). The article caused a small explosion in the business community, in no small part because Carr was putting forward some obvious truisms that, from an economic-historical perspective, simply must be true: as new tools become more and more widely distributed, their effects become less distinctive and less capable of creating corporate advantage, rather than more so. “You gain an edge over rivals,” Carr writes, “only by having something or doing something that they can’t have or do. By now, the core functions of IT—data storage, data processing, and data transport— have become available and affordable to all. Information technology’s power and presence have begun to transform it from a potentially strategic resource into what economists call a commodity input, a cost of doing business that must be paid by all but provides distinction to none” (Carr 2004, 7). Putting the matter rightly into historical perspective, Carr observes that “as was true with railroads and electricity and highways, it is only by becoming a shared and standardized infrastructure that IT will be able to deliver its greatest economic and social benefits” (ibid., 11).

While Carr’s observations seem eminently grounded in history and logical analysis, they were roundly dismissed by business advocates, particularly those in the IT fields. Less noticed, though, is what Carr has observed about the role IT is coming to play in the contemporary corporation and in contemporary social life, which is namely that it is becoming ubiquitous for institutions of a certain size—not merely ubiquitous but unavoidable, a “cost of doing business.” But where the other technologies Carr mentions—railroads, electricity, highways—were either built or owned by governments or recognized fairly quickly as the potential loci for oligarchical control, today such an argument is rarely if ever applied to IT, despite the fact that IT is arguably just as much if not much more invasive, much more personal, much more striating, and much more subject to authoritarian misuse than these earlier technologies. Carr’s article and subsequent book have been roundly denounced by corporate computationalists, among them Bill Gates, arguably not merely because Carr notes the immediate falsehoods surrounding computers in particular but also because he begins to outline precisely the historical and ideological contexts that computers demand:

Because it marks a break with the past, the arrival of any new infrastructural technology opens the future to speculation. It creates an intellectual clearing in which the imagination is free to play, unconstrained by old rules and experiences. Futurists spin elaborate scenarios of approaching paradises (or, less commonly, infernos), and every new vision of the future serves as a foundation for wider conjecture. . . . Soon, the entire public becomes caught up in the excitement, sharing an intoxicating communal dream of renewal.

Although information technology is a far less revolutionary technology than electricity, it has nevertheless spurred a particularly extreme version of this phenomenon, which culminated in the millennial fervor of the 1990s, when visions of digital utopias became commonplace. With an almost religious zeal, the metaphysicians of the Internet promised to free us from the burdens and constraints of our physical selves and release us into a new and purified world of cyberspace. (Ibid., 138-9)

Despite our clarity about this historical pattern, it is striking how difficult it has been to resist as it emerges again—despite the fact that we know in some important way that such promises will always accompany significant technologies and that they must be false, we seem surprisingly recalcitrant to incorporating that historical understanding into our contemporary moment. No doubt this is in part due to a kind of social hope, a recognition that we face deep and significant problems in our world that demand resolution.

While it is clear that a certain strand of utopian enthusiasm inherent in every technological development—perhaps, as Wark might argue, one closely tied to particularly American ideologies of novelty and renewal (see, e.g., Adas 1989, 2006; Marx 1964; Noble 1977)—there is also a profoundly specific character of the IT revolution that we are especially reluctant to face head-on, despite the fact that our most trenchant social critics have tried continually to bring it to the fore. Computation is not a neutral technology; it is a means for expanding top-down, hierarchical power, for concentrating rather than distributing. The main argument for computers as a means of distributing power is that they increase the capabilities of individual users; but in order for this argument to reach the end its advocates claim, this expansion of individual power would have to somehow cease when it reached the levels of power elites and oligarchical institutions. Yet this is clearly not what has happened; instead it is specifically power elites and oligarchies who have access to the most powerful computers and the newest tools; to ignore this situation simply because those of us relatively low on social hierarchies also receive benefits is to miss the forest for the trees.

Among the only discourses of critical thought about computers today has to do with surveillance; we are repeatedly warned, and rightly so, that computational tools “increase surveillance” in society and increase the sense in which average citizens typically feel (and, in fact, are) watched. This worry is correct in so far as it goes, but it still neutralizes computation by viewing it as an instrument that carries with it certain mechanically produced effects: it remains a kind of technological determinism. The problem we still avoid in these discussions is not technological but social and political: there are powers at work in society that want to watch the mass of the citizenry, and that seek to control social movements via such surveillance. If we had no worries about this political power we would worry much less about the means available to it; if we knew that we could in fact supervise our own government via democratic control, it seems clear we would worry far less about the clandestine surveillance and political manipulation secretive organizations like the CIA have undertaken even before the wide spread of computers. Because computation does empower individuals of all stripes, including those of us who are already extremely powerful, we cannot hope that this sheer expansion of power will somehow liberate us from deep cultural-political problems; because computation sits so easily with traditional formations of imperialist control and authoritarianism, a more immediately plausible assumption would be that the powerful are made even more powerful via computational means than are the relatively powerless, even as everyone’s cultural power expands.

Perhaps every bit as disturbing as these considerations is the question of just what power is being inflated as computation grows more and more widespread. Among the oddest but most telling of the cultural changes accompanying the computer revolution is the one that emerges out of the late 1960s social movements, in which a significant segment of youthful intelligentsia embraced the computer as a revolutionary technology that might transform the world along some of the same lines suggested by the counterculture in general (see Turner 2006 for an especially pointed history). In retrospect we can see that this has to be one of the most successful instances of co-optation in the history of social movements; for despite their appearance of transformative power, it is the ability of the computer to expand the feeling and fact of mastery that is most compelling about it. Much like their extended experiments with the profoundly capitalist medium of rock music and profoundly self-gratifying mind-altering substances—visible especially as the supposedly cognitively liberating psychedelic substances gave way to destructive and strongly isolating substances like alcohol, cocaine, and heroin—the counterculture was unable to escape the capitalist and hierarchical strand of dominant American culture it thought itself to be resisting. In the end, this revolution became about exactly the individualistic power it said it was resisting, in no small part via the embracing of a technology that wears its individualist expansionism on its sleeve.

We live in a society that has scarcely begun to learn how to situate itself historically—how to understand the world in which we have arrived and the processes that brought us to where we are now. Looking at the advent of television, railroads, electricity, light, and no less fundamental technologies like the wheel, printing, fire, gunpowder, we can see that they always entail as many problems as they solve, and that their ability to change some (but by no means all) fundamental features of human social life often effects ruinous changes on parts of the human world that in retrospect we may wish not to have lost. It is remarkable that in the act of selling ourselves a story about the revolutionary and progressive nature of technological change via the computer, we ignore the real effects the computer has on the world, and the way in which technology does not and cannot lead to the cultural and political change many of us hope can be created in our world. Just as we have begun to recognize that we may have been wrong for hundreds of years about the “backwardness” of the world’s thousands of cultures and languages, especially with regard to their cultural and political sophistication, we have developed a technology that threatens, via a power that is predominantly cultural, to finish the project of eradication that we now say we reject under the name of colonialism. Just as we have begun to learn who the Inca, Aztec, and Maya really were and are, we are writing into life a story about them that says their cultures failed for good reason—because they failed to develop the technologies that characterize cultural evolution and that we use to justify our own mastery over the social and natural worlds (see Bowers 2000, 76-107, on the close ties between computationalism and social Darwinism). Even as we know this control cannot be complete and may in fact be engendering its own (and our own) destruction, the “old story” has emerged with an unexpected power: exactly the power of computation.

We want to imagine computers as deterritorializing our world, as establishing rhizomatic, “flat,” nonhierarchical connections between people at every level—but doing so requires that we not examine the extent to which such connections existed prior to the advent of computers, even if that means ignoring the development of technologies like the telephone that clearly did allow exactly such rhizomatic networks to develop. What computers add to the telephonic connection, often enough overwriting exactly telephonic technology, is striation and control: the reestablishment of hierarchy in spaces that had so far not been subject to detailed, striated, precise control. Of course in some individual cases these forces may be resisted, but we must not allow these instances of resistance to obscure the tendencies created by our computation, in our world: and here, contrary to received opinion, it is the nature of our computers to territorialize, to striate, and to make available for State control and processing that which had previously escaped notice or direct control. In this sense the computer works to bring under particular regimes of cultural knowledge those aspects of culture that may have previously existed outside of official oversight; perhaps more disturbingly, these striation and territorializing effects have the consequence of establishing profoundly hierarchized, oligarchical institutions in which individuals are seen not as equal participants but as cogs in larger political machines, as objects available for the manipulation of those with real power. In this sense the rhetoric of revolution that surrounds computers today can be seen as a striking displacement for the real effects of consolidation and concentration that computers in fact enable, and that suggest the transformation of institutions that function via democratic participation, and especially that value the unexpected contributions of minorities, into profoundly majoritarian, unified, authoritarian structures, and that offer the renewal of colonial oversight just as we tell ourselves that the era of colonialism has passed.