74 FROM MARXISM TO RADICAL CRITIQUE

logical method" attempts to recover the "local, discontinuous, disqualified,

illegitimate knowledges" of the dominated in opposition to the "unitary body
of theory which would filter, hierarchise and order them in the name of some
true knowledge and some arbitrary idea of what constitutes a science and its
objects" (Foucault, 1980: 83).
Although Foucault's method is very different, his critique is similar to
Marcuse's, not absolutist to be sure, but at least internally consistent in offer-
ing reasons for its claims. But because Foucault continually plays on the
ambiguity of power and domination, we are never sure how far he intends
to go toward a relativistic reduction of knowledge claims to social positions.
Like Marcuse's, Foucault's rhetorical strategy lends force to his condemna-
tion of the established technocracy while blurring essential issues.
Marcuse and Foucault offer a persuasive account of the condensation of
the social and the technical, and they propose a system theory of social action
that appears more applicable to contemporary societies than the traditional
Marxist class theory. But their very advances plunge them into insuperable
difficulties. Their theories of the bias of knowledge threaten rationality as a
whole; their accounts of social order seem to exclude resistance. Is there any
way to preserve the essential insights of these thinkers while avoiding their
shortcomings?
In the remainder of this chapter I will argue that Marcuse and Foucault
encounter these difficulties because they lack a theory of technological hege-
mony capable of explaining the relationship of social organizations to ideol-
ogy/science and power/knowledge. As a result, they invoke a substantivist rheto-
ric that immediately identifies science and ideology, knowledge and power.
These identifications give a particular dystopian pathos to their theories. In-
deed, if the problem is knowledge as such, or the very existence of discipline,
then criticism and resistance are equally hopeless. For all the rhetorical power
of these images of a closed world, they cannot explain the possibility of critical
knowledge of society and systematic resistance. That will require a theory of
those hegemonic mediations responsible for the problems Marcuse and Fou-
cault appear to blame on knowledge and technology per se.

The Technical Code

Double Aspect Theory

Despite many ambiguities, one thing is clear about the position of Marcuse
and Foucault: they reject the accustomed terms of the rationalism/relativ-
ism debate and affirm both that rationality is integral to a system of domina-
 The Bias of Technology 75

tion and that it nevertheless achieves cognitive success. Although this posi-
tion appears contradictory, they finesse the difficulties with the notion that
knowledge and power share a common foundation but have different desti-
nies. For Marcuse that foundation is a method of abstraction rooted in the
will to domination. His critique of that method will be discussed in the last
chapter of this book. For Foucault, it is the common fund of microtechniques
that establish both a disciplinary society and a social science and technology
adapted to it.
As we have seen, neither Marcuse nor Foucault pursues this line of argu-
ment consistently. In this section, I will therefore leave behind the details of
their positions to sketch an approach to the study of technology suggested
by their work. I will call this approach a "double aspect" theory of power/
knowledge or ideology/science because it treats hegemonic and cognitive
functions as complementary aspects of a single underlying source rather than
as separate things.
The place to begin this discussion is with the function of rationality in
modern hegemonies. An effective hegemony is one that need not be imposed
in a continuing struggle between self-conscious agents but that is reproduced
unreflectively by the standard beliefs and practices of the society it domi-
nates. Tradition and religion played that role for millennia; today, forms of
rationality supply the hegemonic beliefs and practices. This is the sense in
which knowledge has become a kind of power, not merely a tool of those in
power, without losing its character as knowledge. This change in the status of
knowledge is rooted in distinctive structures of capitalism.
In precapitalist societies workers had traditional tasks and established
codes of self-expression; they owned their own tools and formed a natural
community. The labor process was so completely enveloped in regulations
and responsibilities that precapitalist elites could rule only by escaping the
economic domain to exercise what Foucault calls "sovereign" power, the
negative power of the state.
Capitalism frees itself from these limitations to an unprecedented degree
by building workforces and markets out of atomized individuals, much like
the prisons and asylums studied by Foucault but on a far wider scale. Release
from all traditional rules and familial restraints, the capitalist has a great deal
more freedom of action than had the leaders of traditional work groups. In
chapter 2,1 called this special kind of freedom the capitalist's "operational
autonomy." It is not primarily a property of individuals but of organizations
that mobilize an array of microtechniques.
Operational autonomy is the power to make strategic choices among alter-
native rationalizations without regard for externalities, customary practice,
76 FROM MARXISM TO RADICAL CRITIQUE

workers' preferences, or the impact of decisions on their households. What-

ever other goals the capitalist pursues, all viable strategies implemented from
his peculiar position in the social system must reproduce his operational
autonomy. The "metagoal" of preserving and enlarging autonomy is gradu-
ally incorporated into the standard ways of doing things, biasing the solu-
tion to every practical problem toward certain typical responses. In indus-
trial societies, strategies of domination consist primarily in embedding these
constancies in technical procedures, standards, and artifacts in order to es-
tablish a framework in which day-to-day technical activity serves the inter-
ests of capital.
Capitalism is unique in that its hegemony is largely based on reproduc-
ing its own operational autonomy through technical decisions. This is usu-
ally sufficient because power in modern societies can be wielded through
technical control without titles of nobility or religious sanctions. The "will"
that Marx identified with Jacobin "politics" is to be found here, in the ex-
pansive dynamic of modern organizations, driven ever further toward the
accumulation of autonomous possibilities of action.
Capitalist social and technical requirements are condensed in a "techno-
logical rationality" or a "regime of truth" that brings the construction and
interpretation of technical systems into conformity with the requirements
of a system of domination. I will call this phenomenon the social code of
technology or, more briefly, the technical code of capitalism. Capitalist hege-
mony, on this account, is an effect of its code.9
In this sociological context, the term "code" has at least two different
meanings. First, it may signify a rule that simultaneously (1) classifies activi-
ties as permitted or forbidden and (2) associates them with a certain mean-
ing or purpose that explains (1). The traffic code defines permitted driving
behavior by distinguishing the safe from the unsafe. Technical manuals are
full of similar codes that determine the rule under which operations are to
be performed in service to a variety of ends such as reliability, strength,
human factors, efficiency, and so on. It is characteristic of a bureaucratized
society such as ours that we have written records of many codes regulating
behavior.
Economic codes are nowhere recorded in a manual but are implicit in
behavior and attitudes, and signify a broader range of values than the per-
mitted and the forbidden. An act of interpretation is required to extract the
meaning of this second type of code from its various manifestations. Con-
sider, for example, the prestige hierarchy of goods such as automobiles: we
"know" that Cadillacs are "better" than Fords, Mercedes "better" than
Volkswagens. In displaying one or the other car we send a message about
 The Bias of Technology 77

ourselves to others. As is clear from this example, such codes have a com-
municative function.
The technical code combines elements of both types. It is most essentially
the rule under which technical choices are made in view of preserving oper-
ational autonomy (i.e., the freedom to make similar choices in the future).
This invariant requirement of the code is not generally explicit, although like
the prestige hierarchy it can be brought to the surface without much diffi-
culty. The goal of enhancing operational autonomy is implicit in the basic
technical procedures of fields that serve the needs of business enterprises and
other similarly structured organizations. Noble's account of numerical con-
trol, discussed in chapter 2, is a clear example. As in that case, the preferred
designs are usually signified as "efficient" with the class bias Marcuse identi-
fied in his critique of Weber. And since efficiency is such a widely shared
value, that signification has a legitimating function that constitutes the com-
municative aspect of the code.
The technical code has (social) ontological significance in a society where
domination is based on control of technology. It is not merely the rule under
which means are chosen. Much more than that, it is the principle of organi-
zational identity and survival. Marc Guillaume thus defines social codes "as
the ensemble of associations between signifiers (objects, services, acts ...)
and that which they signify in society, associations created or controlled by
organizations as a basis of their existence and if possible their development"
(Guillaume, 1975: 64). However, in the case of the technical code, it is nec-
essary to go beyond this formulation. To exist, organizations must encode
their technical base, not merely associating technology with certain signifiers
but installing these signifiers in their very structure.10 How is this achieved?
This question can be answered only by carrying the theory of the condensa-
tion of the social and the technical one step further. In the process, we will
see how the dual aspects of power/knowledge are reconciled in technical
objects.
Everyone who develops modern technologies or studies their history
knows that they are built up from concatenations of more or less loosely
connected parts. The parts themselves arise out of discoveries so basic that,
although they may first have served one or another specific purpose, they
can be used for very different purposes in a wide variety of contexts. Thus,
we distinguish between the principles embodied in technologies and the form
of their concrete realization in this or that actual device.
I will reserve the term "technical element" for the specific principles, such
as the spring, the lever or the electric circuit. These are in themselves "rela-
tively" neutral, if not with respect to all social purposes, at least with respect
78 FROM MARXISM TO RADICAL CRITIQUE

to the ends of ruling and subordinate social groups. The work of discover-
ing such elements is to some extent autonomized in the research process.
Once discovered, they are like the vocabulary of a language; they can be strung
together—encoded—to form a variety of "sentences" with different mean-
ings and intentions.
Individual technologies are constructed from just such decontextualized
technical elements combined in unique configurations to make specific de-
vices. The process of invention is not purely technical: the abstract technical
elements must enter a context of social constraints. Technologies, as devel-
oped ensembles of technical elements, are thus greater than the sum of their
parts. They meet social criteria of purpose in the very selection and arrange-
ment of the elements from which they are built up.11
These social purposes are embodied in the technology and are not there-
fore mere extrinsic ends to which a neutral tool might be put. The embodi-
ment of specific purposes is achieved through the "fit" of the technology and
its social environment. The technical ideas combined in the technology are
relatively neutral, but one can trace in it the impress of a mesh of social de-
terminations that preconstruct a domain of social activity in accordance with
certain interests or values.
Bruno Latour makes a similar point. He argues that each technology draws
together a "sociogram" of alliances of social interests around a specific con-
figuration of technical elements, which he calls the "technogram." Latour
argues that "every piece of information you obtain on one system is also in-
formation on other" (Latour, 1987: 138).12 Sociogram and technogram are
essentially just two sides of the same coin; a particular technical configura-
tion reflects the influence of a particular network of actors. A precise defini-
tion of a specific technology can therefore only be found at the intersection
of the two systems.
The technical code of capitalism can now be defined as a general rule for
correlating sociogram and technogram. The assumption that, as nonowners,
workers are indifferent to the welfare of the firm is the most important social
factor that infiltrates itself into the definition of technical reason through this
code. The assembly line is an excellent example of a technology influenced
by this assumption: a strategy of technologically enforced labor discipline
forms the glue that holds together the elements from which it is composed.
This asymmetrical effect on power is characteristic of a strategically encoded
technology.
This example also illustrates the historical relativity of the process of
rationalization. The assembly line only appears as technical progress because
it extends the kind of administrative rationality on which capitalism already
 The Bias of Technology 79

depends. It might not be perceived as an advance in the context of an

economy based on workers' cooperatives in which labor discipline was self-
imposed rather than imposed from above.
Of course, the parts of an invention like the assembly line have a techni-
cal coherence of their own that in no way depends on politics or class rela-
tions. Technology is not reduced in this example to production relations nor
technical knowledge to ideology. The first term in each of these pairs has its
own logic; technology must really work. But it is not merely because a device
works that it is chosen for development over many other equally coherent
configurations of technical elements. Were that the case, then by analogy one
could also explain the choice of individual sentences in speech by their gram-
matical coherence. The social character of technology lies not in the logic of
its inner workings, but in the relation of that logic to a social context.
This is even true of the social technologies Foucault studies. Techniques
of discipline and normalization do not determine a single organization of
society, but open possibilities that are disputed between dominated and
dominating social groups. Foucault did not despair because, whatever the
outcome of the struggle, modern societies will inevitably employ some vari-
ant of these techniques. On the contrary, his goal was to find a way "which
would allow these games of power to be played with a minimum of domina-
tion" (Foucault, 1988: 18).
Despite radically different purposes and institutional structures, modern
sciences, technologies, and social organizations share a similar method of
abstraction and similar microtechnical foundations. Because of this, scien-
tific and technical disciplines are able to supply the hegemony they serve with
the applications it requires. But the lower we descend toward the founda-
tions, the more ambiguous are the elements from which these applications
are constructed. This is the source of the ambivalence of technology. Thus,
a technical code is needed to bind applications to hegemonic purposes since
science and technique can be integrated to several different hegemonic orders.
That is also why new technology can threaten the hegemony of the ruling
groups until it has been strategically encoded. And that is why hegemonic
claims to monopolize rationality are subject to rational critique. The double
aspect theory can thus demystify the neutrality of knowledge and technol-
ogy without also asserting that, like ideology, they are invalidated by their
service to particular interests.
In Marcuse and Foucault, the relation between technical knowledge and
society is unclear. Because their double aspect theory of technology is im-
plicit, they lack an appropriate terminology in which to express it. At times
they appear to be saying that knowledge is merely a projection of social power;
80 FROM MARXISM TO RADICAL CRITIQUE

and yet they also want to distinguish knowledge from mere prejudices and
ideologies. These hesitations are an attempt to suggest the preestablished
harmony of knowledge and hegemony without reducing one to the other.
I argue here that the connection between knowledge and power, technique
and hegemony lies in the code that ensures that they are coordinated in
the application.

Formal Bias
The theory of the technical code offers a paradoxical challenge to the con-
ventional idea of the neutrality of technology. It is usually assumed that since
technology is founded on a generic interest, it is indifferent with respect to
particular social interests. As we saw above, Marcuse and Foucault reject this
familiar approach. Their double aspect theory of technology attributes a
certain neutrality to basic techniques, if not with respect to ends in general,
at least with respect to different encodings. Technical codes responding to a
specific social interest select one technically coherent configuration of these
basic techniques from among a variety of alternatives. This is what makes
possible modern hegemonies based on technical knowledge.
The critical theory of technology thus implies that in certain cases neu-
trality and bias are not opposites, but merely different aspects of a single
concrete object (Marcuse, 1964:156; Dreyfus and Rabinow, 1983:203). This
approach appears to be very difficult to communicate, and not uncommonly
the double aspects are collapsed back into a single one—usually it is bias—
on the assumption that the critique is an irrationalist attack on objective
knowledge. How can the coexistence of neutrality and bias be more clearly
explained to avoid such misinterpretations?
The problem is due to the fact that we usually conceive of bias as a devia-
tion from fairness, which, in common usage, refers to the application of the
same standard to all regardless of personal feelings. This background explains
why the notion of bias suggests particularity subverting universality, for ex-
ample, nepotism or prejudice slanting a hiring decision that ought to be made
on the universal ground of qualifications for the job. Neutrality, as a prop-
erty of the universal, therefore appears opposed to bias. On these terms, it is
impossible to make sense of the notion of biased technology since the ratio-
nality inherent in technical devices is incommensurable with personal par-
tiality by definition. However, there is another more subtle form of bias that
consists in applying the same standard to individuals who cannot be com-
pared or under conditions that favor some at the expense of others. This type
of bias is often difficult to identify because the application of a single stan-
 The Bias of Technology 81

dard gives the appearance of fairness. In this case neutrality is not the oppo-
site of bias but its essential precondition.13
I will borrow a distinction from Weber's theory of rationality to desig-
nate these types of bias, and call the first "substantive" bias and the second
"formal" bias. As one would expect from these terms, the second type is
peculiarly characteristic of modern societies. For example, it characterizes
conditions in which "formal" equality contradicts social "content," such as
where equality before the law is systematically frustrated by the unequal ability
to pay for legal representation, or where equal educational opportunity is
denied not by discriminatory exclusions but by administering class or eth-
nically biased tests.
In the case of biased technical arrangements, the valuative bias does not
appear as such but governs the "fit" of the formally rational subsystems and
society as a whole. The assembly line can again serve as an example. In Marx's
terms, it is guilty of "supplying capital with weapons against the revolts of
the working class," and so is clearly a biased technology (Marx, 1906:1,476.)
Yet the objective workings of this technology are as blind to social distinc-
tions as the computer that grades a culturally biased test. The bias, in such
cases, originates not in the technical elements but in their specific configu-
ration in a real world of times, places, historical inheritances—in sum, a world
of concrete contingencies. The essence of formal bias is the prejudicial choice
of the time, place, and manner of the introduction of a system composed of rela-
tively neutral elements.
These two types of bias are open to criticism on very different grounds.
Substantive bias, based on the application of unequal standards, is most often
associated with prejudice, with explicit norms that discriminate between people
of different classes, races, sexes, or nationalities. However, since unfair treat-
ment cannot be justified on the basis of mere personal preferences, such norms
are generally represented as factual judgments attributing abilities or merits,
disabilities or demerits to the more- or less-favored groups. The critique of
substantive bias proceeds by showing up its pseudofactual judgments as
"rationalizations," or, where they are highly elaborated, as "ideologies."
Formal bias implies no necessary feeling of prejudice, nor is it associated
with factual errors based on rationalized feelings. On the contrary, the facts
generally support claims of fairness aimed at justifying this type of bias so
long as embarrassing contextual considerations are ignored. Outside the
larger context, fair treatment seems to be rendered through an equal appli-
cation of the same standards to all. But in that context, it becomes clear that
the apparent fairness of the system, taken in isolation, hides systematic un-
fairness of another sort.
82 FROM MARXISM TO RADICAL CRITIQUE

Criticism of formal bias requires redefining the domain of considerations

relevant to judging the action or institution in question. It is not the par-
ticular factual claims advanced in favor of the discriminatory activity that
are challenged, but the horizon under which those facts are defined. The
enlargement of the cognitive horizon in such cases involves passing from
arbitrarily isolated elements to a larger system in which they have a functional
significance. Thus to show discrimination in the case of a technological choice
or a culturally biased test, it is necessary to demonstrate that the discrimina-
tory outcome is no accident but reproduces a relationship of domination.
The traditional neutrality thesis reifies technology by abstracting from all
contextual considerations. This approach is relatively persuasive because, as
in other instances of formal bias, the decontextualized elements from which
the biased system is built up are in fact neutral in their abstract form. The
gears and levers of the assembly line, like the bricks and mortar of the
Panopticon, possess no intrinsic valuative implications. The illusion that
technology is neutral arises when actual machines and systems are under-
stood on the model of the abstract technical elements they unite in value-
laden combinations. Critical theory shatters the illusion by recovering the
forgotten contexts and developing a historically concrete understanding of
technology.
That understanding is critical, but it contrasts sharply with the one-sided
condemnations of substantive theory. Ellul and Heidegger attribute substan-
tive bias to technology and treat it quite literally as a kind of materialized
metaphysics. Their approach confounds the essence of technology with the
hegemonic code that shapes its contemporary forms, denying the existence
of subordinated technical potentialities that could support an essentially
different type of development.
The same deficiency is often attributed to Marcuse too. This may explain
why his critics accuse him of being irrationalist and, like Habermas, beat a
hasty retreat into the conformist view of technology. But a one-sided critique
is not improved by abandoning critique altogether. What is needed is a theo-
retical account of the "other" side, that is, those progressive dimensions of
technology that would come to the fore in the course of reconstructing the
technical base. I address this question in parts II and III of this book.

Technological Figurations
The theory of the technical code and its bias establishes the social relativity
of the existing technology. But if technological domination is a contestable
hegemony rather than a dispensation of being, one would expect it to be
 The Bias of Technology 83

associated with specific forms of opposition and resistance. What would it

mean to make a counterhegemonic use of knowledge or technology? How
would such a use differ from the mere instrumentalization of neutral tools
for new purposes?
Modern ideas of resistance were originally formed in the political sphere
rather than in reflection on the dialectic of technological subjects and their
means. Political struggle lends itself to the sort of instrumentalist accounts I
have criticized as inappropriate to the study of technology. Substantivism
recognizes the failure of the instrumental view and responds with the dystopian
metaphor of society as a gigantic machine. But mechanical imagery describes
a far more stable and harmonious social order than the one in which we live.
A satisfactory model must reflect not only society's power to shape its mem-
bers but also the tensions and resistances it evokes.
In the search for such a model, some theorists have chosen to compare
society with a game rather than with a machine. Games define the players'
range of action without determining any particular move. This metaphor can
be usefully applied to technology, which sets up a framework of permitted
and forbidden "moves" in much the same way games do. The technical code
might be reconceptualized on these terms as the most general rule of the
technical game, a rule that, however, biases the play toward the dominant
contestant.14
The game metaphor is ambiguous, like the society it describes. Thus,
Michael Buroway holds that "playing a game generates consent with respect
to its rules," but he also notes that "participation in a game can undermine
the conditions of its reproduction" (Buroway, 1979:81,94). Buroway's study
of the shop-floor "game" of "making out" illustrates this ambiguity. He
wonders whether what appears to be a struggle for free time against the sys-
tem may actually be functional within it. These two positions are personi-
fied by Cornelius Castoriadis and Marcuse.

But is making out as radical as Castoriadis claims? Or is it, as Herbert Marcuse

would argue, a mode of adaptation that reproduces "the voluntary servitude"
of workers to capital? Are these freedoms and needs, generated and partially
satisfied in the context of work, and harnessed to the production of surplus
value, a challenge to "capitalist principles"? Does making out present an anti-
cipation of something new, the potential for human self-organization, or is
it wholly contained with the reproduction of capitalist relations? (Buroway,
1979: 73)

In the remainder of this section I draw on the ideas of Michel de Certeau

and Norbert Elias to develop a theory of resistance explaining these ambi-
84 FROM MARXISM TO RADICAL CRITIQUE

guities. I will apply their insights to a noninstrumentalist account of subver-

sive practice.15
De Certeau's contribution belongs to a certain phase in the breakdown
of French structuralism. As coded objects, cultural artifacts resemble a syn-
tax regulating behavior that, like speech, follows the rules of the code. For
example, knives, forks, and spoons are not just strips of metal, but imply a
whole system of eating behavior with respect to which each actual meal is a
performance. Clothing, cars, technical devices, are all subject to a similar
analysis.
This approach has a deterministic cast, but some semiologists—for ex-
ample, Roland Barthes—offered looser formulations, applicable to culture,
in which speech practice can modify syntax (Barthes, 1969: 103-104). De
Certeau was influenced by Foucault in attempting to develop a similar theory
of cultural change. The game metaphor serves in this context to soften the
deterministic rigors of the then dominant linguistic model of society.
"Strategies," according to de Certeau, are the institutionalized controls
embodied in social and technological systems such as corporations or gov-
ernment agencies (de Certeau, 1980:1,85). The techniques of power are not
tools wielded by elites; rather, they open a space, an "interiority," from out
of which those elites act on society. The social distance implied in the meta-
phoric pair—interior/exterior—is vertical: it creates a position "above" society
from which to see and control it. To that position corresponds what I have
called the operational autonomy of a hegemonic subject. With certain modifi-
cations, this account could be generalized to any technically mediated
activity in modern societies.
Social groups that lack a base from which to act on an exteriority respond
"tactically" to the strategies to which they are subjected, that is to say with
punctual, temporary, shifting actions that fall more or less under the con-
trol of the dominant strategy but subtly alter its significance or direction.
Tactics are the inevitable response of the dominated to their domination,
unfolding on the terrain of the Other and operating in the "usage" of the
hegemonic system (de Certeau, 1980:1, 59-60).
Just as operational autonomy serves as the structural basis of domination,
so a different type of autonomy is won by the dominated, an autonomy that
works with the "play" in the system to redefine and modify its forms, rhythms,
and purposes. I call this reactive autonomy "margin of maneuver." It may
be used for a variety of purposes in technically mediated organizations, in-
cluding controlling work pace, protecting colleagues, unauthorized produc-
tive improvisations, informal rationalizations and innovations, and so on.
Action on the margin may be reincorporated into strategies, sometimes in
 The Bias of Technology 85

ways that restructure domination at a higher level, sometimes in ways that

weaken its control. Foucault's "subjugated knowledges" are elaborated in the
"space" of tactical involvement, the margin of maneuver opened by strate-
gies. De Certeau offers examples such as practices in which workers exploit
materials and tools from their workplace to make objects for their own use
(la perruque), or the investment of Christianity by local rites in colonial situa-
tions (de Certeau, 1980:1, 68ff., 106-107).
De Certeau sheds a new light on the metaphoric identification of soci-
eties with machines and games. These two metaphors are in fact angles of
vision on social activity. The mechanical metaphor describes a smoothly
working system from the point of view of those who manage it. The view
of the dominated partner is exemplified in the game, and especially in the
peculiar counterhegemonic "move" that changes the rules. The two meta-
phors together thus embrace the complementary perspectives of actors
located in different positions in the system. They exemplify the opposing
self-understandings associated with operational autonomy and margin of
maneuver.
De Certeau's distinction between strategies and tactics offers an alterna-
tive to both instrumental and substantive theories of technology. His theory
of strategies exposes the bias of the apparently neutral technical management
of modern organizations. His analysis of the role of tactics brings out the
inherent limits of dystopian rationalization. At the same time, it suggests a
new way of understanding resistance as neither individual moral opposition
nor as just another policy, indistinguishable except for the accidents of
political fortune from the dominant one. Both morality and policy are func-
tions of strategic will. Resistance, as a tactical modification to which strate-
gies are subject, belongs to another order entirely, that order which Marx
attempted to signify with his early notion of the social.
This approach has a larger context in social theory: the attempt to tran-
scend the dilemma of methodological individualism versus structuralism. A
third position, sketched variously by Pierre Bourdieu, Norbert Elias, and a
number of other social theorists, argues that individual and society, consid-
ered as separate entities, are abstractions from a more concrete unity. That
unity is a structured process of human relations. Norbert Elias calls this pro-
cess a "figuration," an ordered "pattern of bonding" among the "semi-
autonomous" individuals who make up society (Elias, 1978:6). No individual
exists outside such a framework, nor are the frameworks themselves conceiv-
able other than as systems of human interdependencies. In our society, these
relations are asymmetrical and position a few leaders to "manage" the
others.
86 FROM MARXISM TO RADICAL CRITIQUE

Elias uses imaginary games to illustrate these contests for power. In a game
in which one player, A, is much more successful than his opponent, B, A not
only has power over B but "in addition, a high degree of control over the game
as such. Though his control of the game is not absolute, he can determine its
course (the game process) and therefore also the result of the game to a very
great extent" (Elias, 1978: 81). This second-order control over the game is
very similar to what I have called "operational autonomy," the power to select
the procedures and devices (rules) that govern the behavior of those within
the system.
But Elias also allows for the case in which the dominated are able to use
their discretionary margin to strengthen their position. Then the outcome
of the play becomes increasingly unpredictable. It no longer looks like the
result of a strategy but comes to resemble ordinary social interaction. Elias
imagines a multitiered game, very much like a modern society, in which a
small number of players has the advantage over a larger number of weak
players. He describes power shifts in such a game in the following terms:

As long as power differentials are great, it will appear to people on the upper
level as if the whole game and the lower-level players in particular are there
for their benefit. As power balances shift, this state of affairs changes. Increas-
ingly it appears to all participants as though the upper-level players are there
for the benefit of the lower-level players. The former gradually become more
openly and unambiguously functionaries, spokesmen or representatives of
one or other of the lower-level groups. (Elias, 1978: 90)

Elias's model has interesting applications to the politics of technology.

Where the rules of his multitiered game are technologically embedded, they
establish a biased system within which the dominant players functionalize
the subordinate players' moves. Subordinates' initiatives tend to cancel out
as they implement the dominant players' strategy, giving the impression that
the "system" is effective in its own right rather than as a pattern of human
relations. The stronger players experience the play as the implementation of
their own strategy, which itself coincides with a specific technical rationale,
and the subordination of the weaker players then appears as an impersonal
technical necessity.
Technical mediation, however, has unforeseeable consequences. Techno-
logical strategies create a framework of activity, a field of play, but they do
not determine every move. Like all plans or rules, they are coarse grained
compared with the actual detail of concrete activity. Furthermore, the tech-
nical system is not just a plan in the heads of a few administrators; it is a real
thing with its own properties, its own logic. To the extent that this logic has
 The Bias of Technology 87

not been perfectly anticipated and mastered—and it never can be—there will
be breakdowns, imperfections in the order of the plan. The "weaker players,"
those whose lives or work are structured by the technical mediations selected
by management, are constantly solicited to operate in this range of unpre-
dictable effects.16 As a result, tactical responsiveness is not something im-
ported into the technically mediated game from the outside ("life," instincts,
etc.), but is a form of socially necessary freedom generated immanently within
the game itself (Feenberg, 1970).
Struggles over control of technical activities can now be reconceptualized
as tactical responses in the margin of maneuver of the dominated. Just be-
cause a measure of discretion is associated with the implementation of any
plan, the use the dominated make of their position in the system is inher-
ently difficult to foresee and control. It has no predetermined revolutionary
or integrative implications as such, but, like all tactical responses to strate-
gies, is essentially ambiguous. These "usages" of the capitalist technical code
are both necessary to its implementation and germs of a new society. Their
contradictory potentialities are more or less contained by management de-
pending on the extent of its operational autonomy. A strong management
can cancel the potentially subversive long-term impacts of tactical maneu-
vers. If management is forced to compromise with its subordinates over a
long period, they can transform the technical process through iterative tac-
tical responses that gradually weaken management's control and bend its
strategic line. This explains why in technical politics there are no clear "sides"
in struggles for identity that stretch all parties between their own contradic-
tory potentialities.
Workers' control simply carries this process to the limit. It is not a new
state power, but is rather a negative condition for a flowering of tactical ini-
tiative, the "organizing activity" specific to socialism. The ambivalent em-
ployment of the technical heritage depends entirely on maintaining and
enlarging the margin of maneuver required to alter the strategies encoded in
the division of labor and technology.
In sum, modern technology opens a space within which action can be
functionalized in either one of two social systems, capitalism or socialism. It
is an ambivalent or "multistable" system that can be organized around at least
two hegemonies, two poles of power between which it can "tilt" (Ihde, 1990:
144). From this standpoint the concepts of "capitalism" and "socialism" are
no longer mutually exclusive "modes of production," nor is their moral sig-
nificance captured in the manichean conflict between a prisonlike society
and the individual in revolt. They are, rather, ideal-types lying at the extremes
of a continuum of changes in the technical codes of advanced societies. As
88 FROM MARXISM TO RADICAL CRITIQUE

such, they are constantly at issue in struggles over all sorts of technical prob-
lems: at work, in education, medicine, ecology, and, as I will argue in the next
two chapters, in the development of a new technology such as the computer.
This position offers a way of understanding the continuing struggle for radi-
cal change in a world that no longer believes a new civilization can be created
by ordinary political action, or geographically localized in this or that country
or block.
 II

The Ambivalence of
the Computer

The last ten years has been characterized by the very rapid computerization
of society not only in the United States, Europe, and Japan but around the
world. With the spread of the technology, discourses and projects originally
associated with production reach into every aspect of social life. For example,
deskilling and automation can now be extended to such areas as education,
medicine, even food service. The computer seems to promise the final triumph
of the Heideggerian enframing.
Yet computers can also be employed to develop and apply skills and
initiative. Instead of reducing individuals to mere appendages of the machine,
computerization can provide a role for communicative skills and collective
intelligence. Computer design thus involves a choice between two different
conceptions of the relation of rational systems to human action and between
two corresponding conceptions of what it is to be human in a technological
society.
The first chapter of this part develops this theme in relation to early
theoretical debates over automation and artificial intelligence. These debates
reflect profoundly different potentials of modernity that contend for the soul
of a computerized society. A second chapter then applies the resulting
perspective to current debates over online education. Here is an issue that
clearly illustrates both the threat and the promise of computerization. On the
one hand, the production mentality follows the computer into the ivied halls
of academe. On the other hand, online education could contribute to a
radical redistribution of cultural and technical competences in a more
democratic form of modern society.
This page intentionally left blank
 4

Postindustrial Discourses

The Ambivalent Computer

Early commentaries on the computerization of society projected either opti-

mistic scenarios of social salvation or nightmares of impending dystopia. The
optimists argued that computers would eliminate routine and painful work
and democratize industrial society. The pessimists argued, on the contrary,
that computers would put millions out of work and bring universal surveil-
lance and control.1
There is a third alternative: perhaps the computer is neither good nor evil,
but both. By this I mean not merely that computers can be used for either
domination or democratization but that they can evolve into very different
technologies under the influence of different strategies of development. My
purpose here is to review a wide range of discourses representing the ambi-
valence of the computer in order to test the theory of the bias of technology
introduced in the previous chapter. By way of introduction, let's consider
the contradictory potentialities of computerization in more detail.
The computer's structure bears an ominous resemblance to mechanistic
rationalization. Computers work under the control of programs devised
outside the technology by human agents who command it from above. Un-
like other machines, the computer is an automaton that realizes a plan in-
stalled in its core rather than simply responding to external controls. This
explains the authoritarian connotation of the metaphoric "programming"
of people and social systems. What is the significance of this curious struc-
tural parallel between the computer and a hierarchical organization of soci-
ety? Is the computer predestined to strengthen the administrative grip of the
powers-that-be? Or does it contain democratic potentialities obscured in the
dominant understanding of the technology?

91
92 THE AMBIVALENCE OF THE COMPUTER

As we have learned from the Internet, computers are useful not only for
control but also for communication, and any technology that enhances human
contact has democratic potentialities. But this function of the computer was
largely invisible to the general public until this decade and is still treated with
suspicion by those whose power is based on a knowledge deficit it could
subvert. The computer's communicative capabilities could attenuate the
distinction between mental and manual labor. New forms of sociability could
emerge that would become a medium for democratic self-organization.
The ambivalence of computer technology can be summarized in two
principles that describe the social implications of technological advance. I
call the first of these the "principle of the conservation of hierarchy." Ac-
cording to this principle, the social hierarchy can generally be preserved
and reproduced as new technology is introduced. Computerization of record
keeping is a case in point, intensifying surveillance and control. A second
"principle of democratic rationalization" holds that new technology can
often be used to undermine the existing social hierarchy. Most major inno-
vations open possibilities of democratization that may or may not be real-
ized depending on the margin of maneuver of the dominated. Thus, in
many workplaces the drive to computerize has excited and sometimes ful-
filled participatory expectations.
This kind of argument was first presented by Marx over a century ago,
not specifically in relation to automation but as a theory of the ambivalence
of industrial development in general.2 While well aware of capitalist deskilling,
Marx argued that the productivity of machine technology can generally be
enhanced by inputs of knowledge and skill. More "intelligent" means of pro-
duction based on a deeper understanding of nature can be used to maximum
advantage only by more intelligent producers. Automation is merely an in-
stance of this general proposition. Marx concluded that with mechanization,
"Labour no longer appears so much to be included within the production
process; rather, the human being comes to relate more as watchman and
regulator to the production process itself.... [The worker] steps to the side
of the production process instead of being its chief actor" (Marx, 1973:70S).3
How much validity does Marx's argument retain over a century after its
initial formulation? Surprisingly, current reflections on automation repro-
duce its very structure. Thus, if the information age appears full of unprece-
dented threats and emancipatory potentials, that may be an effect of histori-
cal amnesia. We can learn from Marx that every stage in the development of
industrial technology was haunted by missed opportunities for democratic
progress. These opportunities are due to the impact of mechanization on the
potential economic contribution of cultural advance. In the next section, I
 Postindustrial Discourses 93

will explore past and current discourses of automation in the light of this
connection between economics and culture.

Automation and Ideology

In a famous book called Automation, published in 1952, John Diebold fore-

saw a new day dawning for workers. Automation, he wrote, "means... that
to a great extent the jobs in which the worker is tied to and paced by the ma-
chine will be taken over by other machines. The worker will be released for
work permitting development of his inherent human capacities" (Diebold,
1952: 162). Diebold hinted that management might have trouble adjusting
to the change: "The humility that management needs . . . is a quality that,
although always essential in administration, will be of even greater impor-
tance in the future" (Diebold, 1952: 163).
Over the years several studies cast doubt on Diebold's predictions (Buck-
ingham, 1961: 96ff.). Some thirty-five years later, Harley Shaiken reviewed
the results of a generation of automation in American industry. He concludes
Work Transformed with reflections on the following theme:
It is ironic that computers and microelectronics should be used to create a
more authoritarian workplace. They could just as easily be deployed to make
jobs more creative and increase shop floor decision-making. Rather than pace
workers, systems could be designed to provide them with more information
about the production operation in general and their own jobs in particular.
The technology could be used to bring the work under the more complete
control of the people who do it rather than the other way around. (Shaiken,
1984: 267)
The reason for the dismal failure to realize the promise of automation? Ac-
cording to Shaiken, it is "the use of technology to extend managerial power"
(Shaiken, 1984: 268).
Shaiken's comments belong to a discourse of automation that articulates
the conflict between technical potential and management resistance. We are
told now that automation requires a new type of labor process based on new
machine designs, but we are still trapped in Taylorism by various institutional
lags. In this section I will present conjointly two influential texts of this ap-
proach, a social scientific work by Larry Hirschhorn and a similarly inspired
business best-seller by Shoshanna Zuboff (Hirschhorn, 1984; Zuboff, 1988).
In Beyond Mechanization, Hirschhorn offers a historical account of how
postindustrial work has been determined by the development of modern
production technology. He argues that such undesirable social consequences
94 THE AMBIVALENCE OF THE COMPUTER

of early mechanization such as Taylorism were due to technical limitations

we can now surpass. The old mechanical technology was extremely rigid
because it employed built-in mechanical controls such as gearing and cams.
For that reason its efficient application required hierarchical management
and sharply divided labor. The new postindustrial technology employs elec-
tronic controls that can be operated separately from the machines they gov-
ern. The technical system becomes flexible both in operation and goals, with
far-reaching consequences, including an increasing role for workers with new
types of skills.
Shoshanna ZubofPs book on the computerization of the workplace, In
the Age of the Smart Machine, develops a similar argument. Zuboff claims
that computers make possible two distinct transformations of the workplace.
On the one hand, they can be used to automate production, relieving human
beings of physical effort and replacing skilled with unskilled labor. On the
other hand, they can be used to "informate," ZubofPs term for the integra-
tion of workers and machines in a reskilled labor process. Informating is not
exactly an alternative to automation in the usual sense, but a better way of
automating that realizes the human potentialities of the workforce as well as
the technical potentialities of the computer.
Hirschhorn and Zuboff attempt to pinpoint the unique properties of
computers that can support the demand for increased skill. Hirschhorn
argues that as mechanical control relaxes and the control system separates
out, "the machine has developed into a communications apparatus. The
transmission of information, not power, has become its primary purpose. ...
Only through reinterpretation and reconstruction as a communications
device can the machine play an effective role in feedback-based produc-
tion" (Hirschhorn, 1984: 37).
According to Zuboff, information technology not only produces prod-
ucts but also represents the world on which it acts. This communicative or
"reflexive" dimension of information technology gives rise to a "textualized"
work process that increasingly blurs the distinction between mental and
manual labor. A process of automation that emphasizes the replacement of
man by machine rests on the mechanical capabilities of information tech-
nology alone and "could lead to chronic suboptimization of the technology's
potential" (Zuboff, 1988: 66).4
In Hirschhorn's account, the key to the "reinterpretation and reconstruc-
tion" of automated systems lies in their very imperfections. Although they
can handle the routine problems their designers anticipate, they can never
achieve the ideal of self-regulation engineers and managers have set for them.
Unforeseeable "second-order" breakdowns arise from the vagaries of wear
 Postindustrial Discourses 95

and tear, materials quality, operator error, and changes in production sys-
tems. Work in a postindustrial society consists in dealing effectively with
these second-order breakdowns. Hence, Hirschhorn rejects "the wistful
utopianism" of engineers attempting to build the perfect system. "Machine
systems inevitably fail, given the realities of materials and human behavior.
Once we accept failure as a part of technological reality, we will gain a clearer
perspective on postindustrial work" (Hirschhorn, 1984: 86).
Workers in "symbolically mediated environments" have very different
needs from those in mechanical ones. Describing an automated pulp factory,
Hirschhorn writes, "The operators were engaged in second-order work, the
management of novelty, even as the machinery of production became more
automated and the process became more continuous" (Hirschhorn, 1984:
100). This sort of activity requires a redefinition of work as a process engag-
ing the worker's capacities as much as the machinery of production. Learn-
ing and work merge in this new technical environment.
Zuboff argues that the heritage of Taylorism is the chief obstacle to this
redefinition of work. The new approach to work does not sit well with man-
agement, the very existence of which is rooted in the expropriation of skills:
"Rationalized knowledge was the occasion for the expansion of middle man-
agement and became the basis for its legitimation" (Zuboff, 1988:232). For-
mal education and intellective skills were monopolized by management and
distinguished it from workers. Thus, it has become "second nature for man-
agers to use technology to delimit worker discretion and, in this process, to
concentrate knowledge within the managerial domain" (Zuboff, 1988: 69).
But the informating process requires the reverse, and it can succeed only
where management designs training and organizational structure to spread
intellective skill as widely as possible. Zuboff continues/'Without this stra-
tegic commitment, the hierarchy will use technology to reproduce itself.
Technological developments, in the absence of organizational innovation,
will be assimilated into the status quo" (Zuboff, 1988: 309-310). Unfortu-
nately, those with the most to lose, at least in terms of their traditional self-
understanding, are the very ones on whom change depends.
Hirschhorn discusses engineering in similar terms, as haunted by the old
mechanistic conception of work. "The very character of postindustrial work,
of second-order control tasks, of monitoring and evaluating signals and data,
increases the significance of group processes. Yet managers and engineers
continue to make work-design decisions as if group life did not exist"
(Hirschhorn, 1984:159). It is once again the heritage of Taylorism that blocks
adaptation to the new world of postindustrial technology. Taylorism is in-
compatible with a "learning approach to machine installation and develop-
96 THE AMBIVALENCE OF THE COMPUTER

ment" (Hirschhorn, 1984: 57). Engineers must get beyond the notion that
there is always a "technical fix" and come to terms with the complexity of
the social system in which their tools will be employed. Hirschhorn concludes:
There is more at stake here than competing philosophies of engineering
design. Each principle sets the stage for a different conception of work. The
principle of integration and Utopian design reinforces a Taylorist view: the
more perfect the machine, the simpler and more rational the job. Systems
theory, control engineering, Utopian thinking, and Taylorist prescriptions all
converge to limit the worker's skill. In contrast, the principle of flexibility
creates a conception of work in which the worker's capacity to learn, to adapt,
and to regulate the evolving controls becomes central to the machine system's
developmental potential. (Hirschhorn, 1984: 57-58)
Although Hirschhorn and Zuboff do not blame capitalism for the prob-
lems they discuss, their critique of the high cost of authoritarian manage-
ment generally parallels that of Marx. They show that the computer is an
ambivalent technology available for alternative developments. Automation
increases management's autonomy only at the expense of creating new prob-
lems that justify workers' demands for an enlarged margin of maneuver. That
margin may be opened to improve the quality of self-directed activity or it
may remain closed to optimize control. As Zuboff writes, "Technological
design embodies assumptions that can either invite or extinguish a human
contribution" (Zuboff, 1988: 182).

Computers, Communication,
and Artificial Intelligence

From automation to artificial intelligence (AI) appears a great leap indeed,

but both fields are divided by similar ambivalences. Of course, AI has no
precise equivalent for the ideologies of automatism and participation, but
the various currents in this field reflect a parallel conflict in the vision of
human life.
Recent debates over artificial intelligence also raise interesting philosophi-
cal questions concerning the nature of rationality. Roughly formulated, the
problem concerns the similarities and differences between human thought
and information processing. To the extent that similarities can be found,
computerized automata can replace people for many sophisticated purposes.
To the extent that differences appear, greater philosophical precision is in-
troduced into the notion of human thinking, clearly distinguished from man-
made simulacra.
 Postindustrial Discourses 97

These similarities and differences are not merely theoretical but also con-
cern computer design and programming, AI is a unique field in which con-
nections can be made between the technical preoccupations of practitioners
and the social theory of technology. And, since computer programming and
design are probably the most completely "textualized" forms of work today,
this discussion also lends support to the argument of the previous section.

The Myth of Artificial Intelligence

There are at least three different senses of the term "artificial intelligence."
In the first place, AI is a type of computer program that, despite wild over-
selling, has certain concrete results to its credit. AI has been used in medi-
cine, for example, to analyze laboratory test results. Whatever philosophers
may think of artificial intelligence, there is no reason to expect technical
progress in simulating certain intellectual functions to slow down soon.
Second, AI has inspired a new field in psychology that takes the com-
puter as a model of the mind. This approach suits the dominant rational-
istic outlook of our society. Philosophers and psychologists are pleased to
find that, having conceived of thinking as a kind of machinery, machinery
in fact turns out to be the perfect image of the process of thought. Despite
the implausibility of this premise, researchers have gained useful insight
into mental functioning.
Third, AI is the slogan of an ideological movement for reconceptualizing
man on the model of his own automata.5 At the highest level of abstraction,
this is a philosophical enterprise remote from social interests. Participants
in this movement therefore have the sense of doing "science" much like their
other colleagues in the university and tend to attribute the prestige of their
research program to its inherent virtues. But the social reception of these
speculations is an entirely different matter. Why have the most exaggerated
claims of AI become grist for popular psychology and to what practical project
can they be linked?
The theoretical advances of cognitive science affect the lives of ordinary
people only indirectly, through the plausibility they give to metaphors that
identify human beings with computers. The popularity of these metaphors
is disturbing: if computers are the very image of man, then the mechanical
world forms a closed system in which we are no more nor less than a work-
ing part. The technological obsolescence of mankind has never been closer
to achievement. Certainly the progressive political and ethical advances of
the last few hundred years cannot survive the discovery that human beings
are, after all, merely computational devices.
98 THE AMBIVALENCE OF THE COMPUTER

It is true that the French Enlightenment long ago declared that "man is a
machine." The Enlightenment made a progressive use of demystifying ma-
terialism, but the renewal of that doctrine today is not a response to religious
obscurantism. Rather, contemporary materialism appears to be the theoreti-
cal expression of the obsession with total control that David Noble identi-
fies in the managerial world.
What is the self-understanding of a machine supposed to be like? Perhaps
the answer is supplied by the theory of the "new narcissism," the intensified
pursuit of personal pleasure by individuals who have less identity than ever
before. The collapse of public life and the decline of the family seem to cut
individuality loose from its institutional moorings. No longer concretized
through real bonds and obligations, the person becomes a discontented spec-
tator on his or her own life, engaged in strategies of manipulation and control
directed toward the self and others alike (Lasch, 1979). The computerization
of the human self-image places the subject now in the position of programmed
device, now in the position of programmer. The discourse of human relations
in this new age of narcissism brings home the desolation of mechanical man.
People "push each other's buttons" today where once they might have been
sentimentally described as falling in love (Turkic, 1984: chap. 8).

Computer-Mediated Communication
There is one audience that has a unique perspective on the AI debate since it
applies theories of intelligence in its work. Among computer programmers
and designers, discussions about the nature of intelligence are not merely
theoretical but also express tensions in the self-understanding of a profes-
sion. Its members rely for the most part on unreflected projections of the
engineering culture in which they are socialized. These projections define the
"real" function of computers and the best way of using them.
The ordinary computer user is sheltered to some extent from this culture
by the higher level interfaces of application programs such as Microsoft Word,
but one still gets a hint of the engineers' world from these programs. It is a
rationalistic world that bears little or no connection to everyday experience,
in which thinking consists in linear operations on unambiguous representa-
tions of artificial, decontextualized, and well-defined objects; problems are
clear-cut and solutions definitively testable. To be sure, this is a world in
which cars, power plants, and bridges are successfully built, but it is a spe-
cialized instance of intelligence and not its paradigmatic case.
These rationalistic assumptions are embodied in the technical code of the
computer profession, the rules and procedures on the basis of which stan-
 Postindustrial Discourses 99

dard design decisions are made. It is this technical code that defines the com-
puter as a system of control, an automaton. The AI debate brings hidden
premises of this underlying code to conscious awareness.
The computer world is an especially favorable setting for the ideology of
automatism, but even there the ambiguities of the information age have an
impact. The functions of mainframe computers include communication,
usually in the form of electronic mail. Programmers and designers "live" in
an environment defined by the computer programs they use, exchange, and
discuss online. Computing is a web of communications, a social as well as a
technical network.6
In its application to communication, the computer has an astonishing
power to form the medium for a parallel world. The participants in regu-
lar online discussions find their lives doubled into a "real" and a "virtual"
segment. In their everyday world they relate to people who are geographi-
cally close, but in the virtual world, social contacts are chosen without ref-
erence to geography, exclusively on the basis of shared interests or work.
Despite a certain simplification of social interaction that results from its
decontextualization, online communication shares the ineradicable complex-
ity and ambiguity of speech in natural language (Feenberg, 1989). It does
not conform with the computer culture's standard model of intelligence any
more than would the conversation around the Coke machine in the program-
mers' office. Yet this online world is not extraneous to the computer but is
the form of its symbolic mediation in the contemporary labor process.
We are familiar on the Internet with online work and discussion groups.
These "computer conferences" or "discussion forums" are typically "asyn-
chronous," meaning that messages are stored on a server and made available
to members at their terminals whenever they call in. The earliest version of
this new medium, computer conferencing, dates from 1974 when it was in-
troduced as an improvement on simple person-to-person electronic mail.
Its first successes were in computer companies, where employees could
understand the programs and had easy access to the necessary equipment.
These applications are a perfect illustration of Hirschhorn and ZubofPs argu-
ment that postindustrial work is essentially a process of communicating and
learning organized around the "reflexivity" of computer technology.
Computer conferencing at the Digital Equipment Corporation (DEC) is
a case in point. DEC gambled very early on what is called "distributed net-
working," that is to say, the linking up of computers in integrated systems.
Instead of building huge "mainframe" computers, like IBM, each standing
in solitary splendor at the center of its own world, DEC's middle-sized "mini-
computers" were designed to be interconnected to share files and tasks. But
100 THE AMBIVALENCE OF THE COMPUTER

connecting computers means connecting those who use them, including

designers working on the company's products. DEC's industrial strategy
reacted back on the company itself, amplifying a preexisting culture based
on horizontal ties and coalitions.
In the mid-1980s DEC's 125,000 employees were scattered all over the world
and linked together by a corporate computer network. In 1986, engineers
developed the first version of the VAX Notes computer conferencing system
to improve the functioning of networked project groups. Eventually the sys-
tem grew to 15,000 conferences with tens of thousands of members, and VAX
Notes itself was polished up and distributed as a DEC software product.
The company declined to control the content of the network: conferencing
at DEC evolved entirely in function of the users' interests. In addition to work-
related conferences, many others were formed: clubs and sports groups,
employees with multiple sclerosis, executives writing international restau-
rant reviews, and so on. In short, the real social world of DEC was doubled
by a "virtual" community.
Here, then, is an unsuspected aspect of computer work. The contradic-
tion between automatism and communication built into computer practi-
tioners' daily experience offered a certain margin of maneuver that they were
able to use to modify their social insertion and activities. One of the many
VAX Notes conferences was especially symptomatic of these contradictions:
a discussion of Heidegger's philosophy. A leading design engineer and his
coworkers started the conference because they had lost faith in their ratio-
nalistic assumptions about human beings. Heidegger's phenomenology of
human action seemed to promise an escape from their naive engineering
culture toward a more realistic approach to designing interfaces and equip-
ment (Whiteside and Wixon, 1988).
I discuss some of the implications of this surprising turn in the next sec-
tion. This reaction testifies to a tension in technical professions between
widely accepted rationalistic technical codes and the everyday realities of
human thought and action. This tension was articulated in 1989, when the
philosopher Hubert Dreyfus organized an "Applied Heidegger Conference"
at the University of California, Berkeley, attended by hundreds of profession-
als not only from the computer world but from fields as diverse as nursing
and management.
These are "specific intellectuals" in Foucault's sense, intellectuals whose
resistances and revolts are rooted in their social function and its associated
knowledge base rather than in the language of politics and justice employed
by the literary intellectuals of earlier times (Foucault, 1980: 127-128). In
calling such intellectuals "specific," Foucault does not imply that their action
 Postindustrial Discourses 101

lacks universal significance, but that it grows out of a local situation in the
technical division of labor. In the terminology introduced here, specific in-
tellectuals act in the margin of maneuver associated with a technical domain
in order to transform the code establishing that domain. I would like to turn
now to a consideration of an early attempt in the AI community to articu-
late the foundations of an alternative code.

Toward a New Paradigm

The AI field is divided into two camps, a majority "cognitivist" camp and a
"neoconnectionist" minority. Cognitivists attempt to simulate the essential
operations of human thought with very powerful computers of conventional
design. These "serial computers" move quickly from one operation to the
next, manipulating symbols in sequence according to syntactic rules con-
tained in their programs. So-called expert systems work in this way, sorting,
classifying, and calculating with symbolic materials supplied by the users.
Such computers finally beat the world chess champion in 1997.
Neoconnectionism's best argument against this approach is the fact that
the human mind does not think in linear sequences, but "processes" data in
complex parallel operations. "Parallel processing" appears to be essential to
such activities as vision, which would explain why it is easier for a serial com-
puter to beat Kasparov than to imitate the eye of a fly.
The neoconnectionists hope that their "neural networks" can overcome
these limits although so far there has been far less progress than promised a
decade ago. This is a new computing technique that applies parallel process-
ing to tasks for which ordinary programs seem in principle unsuited. The
operation of these networks more nearly resembles an apprenticeship through
trial and error than a programmed processing of symbols. The neural net-
work interacts with the environment in such a way as to reorganize its own
internal state in a coherent manner that can be used for some purpose such
as recognizing or imitating patterns. Because it is based on statistical regu-
larities, it can work with approximations and improve its own performance.
A few products have been developed using these techniques, but the field is
still in its infancy.

The Paradox of Self-Organization

According to Jean-Pierre Dupuy, this division in the AI research commu-
nity was prefigured in the debates of the cybernetics movement that preceded
102 THE AMBIVALENCE OF THE COMPUTER

it. Mainstream cybernetics attempted to show that self-organizing systems

such as living things could be explained on the basis of the same principles
of feedback, homeostasis, and control that apply to machines. Meanwhile, a
smaller group attempted to distinguish between self-organizing systems and
mechanical ones, but at first only the study of mechanical systems prospered
(Dupuy, 1985).7
The theory of self-organization was taken up again by a group of original
thinkers with better approaches in recent years. This emerging field, which
Heinz Von Foerster calls a "second cybernetics," is represented principally
by Von Foerster himself, and Henri Atlan, Humberto Maturana, and Fran-
cisco Varela (Dupuy, 1982:227). Dupuy argues that this "second cybernetics"
is in the process of resolving fundamental problems in the heritage of the
first.
Early cybernetics bequeathed biology and neurology a set of concepts
derived from mechanical models, such as the notions of genetic "codes" and
mental "programs." Whatever the fruitfulness of such concepts in particu-
lar applications, insofar as biological and mental life are self-organizing sys-
tems they are fundamentally different from machinery, even such sophisti-
cated machinery as computers.
As machines, computers are turned on the one side toward action in the
world and on the other side toward a human user. Like a hammer, which
possesses a head for striking and a handle for holding, the computer's very
structure implies an operator who intervenes in the mechanical environment
but is not a part of it. This structure—control from above—appears self-
evident: the programmer operates the computer and not vice versa. When
the order is reversed, when the operator is also the object of action, as, for
example, when the hammerer strikes his own thumb, the operation falls
outside the domain of technical action proper and is counted as a mistake.
Russell and Whitehead explored the logical structure of such irreversible
hierarchies. They wanted to eliminate reflexive paradoxes, such as the famous
"liar's paradox," which occur when certain types of propositions refer to
themselves. The logical equivalent of hammering on one's own thumb is
exemplified by the statement "This sentence is false." Russell and Whitehead
introduced the "theory of types" to expunge such paradoxes from language.
This theory requires a clean separation between levels of discourse. In the
accepted terminology, the higher-level "metalanguage" refers to the lower-
level "object language" but it cannot refer to itself. Russell and Whitehead
permit Sentence A (the metastatement) to claim that Sentence B (the object
statement) is false only if A and B are different, thereby avoiding the liar's
paradox.
 Postindustrial Discourses 103

But, however inconvenient reflexivity is in logic, it is essential to the world

as we know it. Living things are "programmed" by genetic materials that are
themselves the objects on which the genetic program operates (Atlan, 1979:
21-23). And, although some mental operations are describable in terms of the
metaphor of external programming, the brain as a system largely "creates" it-
self by operating on its own states; human thought more nearly resembles a
neoconnectionist neural network than an ordinary computer. We are, in short,
self-programming beings, an apparent contradiction in terms (Varela, 1984).
Social applications of the concepts of the "first cybernetics" resonate with
the ideology of total control. The separation of the (controlling) metalevel
of the programmer or operator and the (controlled) object level reflects the
split between conception and execution in modern technical systems. In
contrast, the idea of a self-programming or self-organizing system has a para-
doxical structure and emancipatory implications: in a democracy all indi-
viduals are both objects of administration and administrators of each other.

Ontological Designing
We appear to have wandered far afield from artificial intelligence, but in fact
the divisions in AI can be reformulated in terms of the concept of self-
organization. Two Chilean neurophysiologists, Humberto Maturana and
Francisco Varela, joined the debate with an innovative conception of the brain
as a self-organizing system (Maturana and Varela, 1987). Their theories in-
fluenced a small group of computer scientists and designers who challenged
the dominant rationalistic technical code. This influence was primarily me-
diated through the account of Maturana's theories in Understanding Computers
and Cognition by Fernando Flores and Terry Winograd (Winograd and Flores,
1987).
Maturana rejects the prevailing model of mental functioning according
to which the mind is essentially an observer of the world. On that account
the mind forms mental representations of what it observes, and these repre-
sentations then serve as mediations between sensory inputs and outputs of
action. Mental "programs" are said to organize the construction of such rep-
resentations and the response to them. Maturana's theory of the nervous
system breaks with this representationalist paradigm of cognition and con-
ceives the mind not as an observer but as an actor immediately engaged with
reality.
Operationally considered, cognition is not the construction of represen-
tations in the brain but the patterning of behavior. Such patterning aims at
the preservation of the structure of the organism in and through interaction
104 THE AMBIVALENCE OF THE COMPUTER

with the environment. To this end, the organism must achieve what Maturana
calls "structural coupling" with the world around it (i.e., effective responses
to the perturbations it experiences). To explain the brain as a self-organizing
system is to show how it continually reproduces itself under these dynamic
conditions (Maturana and Varela, 1987: 75-80).
According to Varela, the mind is not basically a manipulator of symbolic
representations like a computer. He observes that "only a predefined world
can be represented," but, as he points out, we do not live in such a world
(Varela, 1988:92). The world is not given to us as a collection of well-defined
objects and problems but as an infinitely rich context of action. We do not
discover the unambiguous truth of that context in knowledge but "enact" a
viable "world" on the basis of our experience and culture. This is what human
intelligence is all about and it is quite different from representing a world
the outlines of which are clear prior to action.
Varela and Maturana show that the representationalist paradigm of knowl-
edge presupposed by expert systems works only against a background of
practical involvements it cannot explain. They agree, of course, that we are
able to construct representations of aspects of the real world, but a category
mistake is involved in treating those representations and the expert systems
based on them as general models of the world and intelligence. Accordingly,
Winograd and Flores argue that

the current discourse about computers is based on a misinterpretation of the

nature of human cognition and language. Computers designed on the basis
of this misconception provide only impoverished possibilities for modeling
and enlarging the scope of human understanding. They are restricted to rep-
resenting knowledge as the acquisition and manipulation of facts, and com-
munication as the transferring of information. As a result, we are now wit-
nessing a major breakdown in the design of computer technology. (Winograd
and Flores, 1987: 78)8

Winograd and Flores conclude that what is needed is "new ground for
rationality—one that is as rigorous as the rationalistic tradition in its aspira-
tions but that does not share the presuppositions behind it" (Winograd and
Flores, 1987: 8). For the authors this alternative tradition is represented by
Martin Heidegger. It is interesting to note that these technologists have no
use for Heidegger's later substantive theory of technology; they are concerned
only with the early theory of action developed in Being and Time, which they
apply to the human relation to computers. There Heidegger argues that being
and subjectivity are inextricably intertwined in "being-in-the-world." We are
"thrown" into the world, obliged to establish our own meanings and objects,
 Postindustrial Discourses 105

always already in the midst of action. The objective representation of "things,"

in the specific sense of stable, independent objects, is a secondary process
and not our basic relation to reality.
Maturana appears to be making similar claims. The representational
model of cognition, in which things and their properties are presumed to
precede activity, is a theoretical construct built by observers who are outside
the situation of active involvement they describe. In fact, cognition occurs
against a background of practical assumptions, called "preunderstandings"
by hermeneutics, that construct the domain of experience as an action do-
main. Knowledge articulates distinctions already made at the practical level;
but these distinctions cannot be explained after the fact by reference to the
very objectivities that they establish in establishing a world.
So far the theory seems quite abstruse, but Winograd and Flores oper-
ationalize it through focusing on Heidegger's concept of "breakdown."
Heidegger holds that action is not a simple response to the objective quali-
ties of things. Rather, what we take for things in a contemplative mode actu-
ally arises from the "breakdown" of practical behaviors in which we engage
with reality at a more fundamental level.9 In breakdown, the "ready-to-hand"
objects of action become "present-at-hand," that is, they are viewed from a
distance as things and not experienced immediately as a dimension of an
action system. This theory has a certain similarity to Maturana's concept of
structural coupling, which involves a kind of "readiness-to-hand." "What
really is is not defined by an objective omniscient observer, nor is it defined
by an individual... but rather by a space of potential for human concern
and action" (Winograd and Flores, 1987: 37).
These concepts suggest a very different paradigm of computer design from
the rationalistic tradition, with its emphasis on thought, planning, and deci-
sion. Rather than constructing an exhaustive rational map of the program
for the user, "the designer of a computer tool must work in the domain gen-
erated by the space of potential breakdown in that [structural] coupling"
(Winograd and Flores, 1987: 72).10 This is reminiscent of Hirschhorn's dis-
cussion of industrial design. Because of the inevitability of breakdown, "the
allocation of responsibility between the controls or computer and the op-
erator must be dynamic, based on the operator's learning needs as well as
the performance requirements of the system" (Hirschhorn, 1984: 97).
Heidegger's phenomenology of action can contribute to these technical
discussions because it looks at the world from the standpoint of the involved
subject rather than from that of the external observer. That subject has ap-
peared in our discussion before as the individual engaged in tactical maneu-
vers in an environment shaped by an alien rationality. In that context, the
106 THE AMBIVALENCE OF THE COMPUTER

theory of breakdown refers not to a purely contingent feature of the human

relation to tools, but more specifically to the limits of control from above.
These themes now come together in the idea of an alternative rationality, a
rationality of implementation rather than of planning and control, based on
self-referential processes of communicating and learning in the course of
using and modifying tools.
Winograd and Flores argue that computers are not automata, artificial
intelligences, but "machines for acting in language" (Winograd and Flores,
1987:178). AI needs to lower its sights considerably if this is true. From this
standpoint, "The relevant questions are not those comparing computers to
people, but those opening up a potential for computers that play a mean-
ingful role in human life and work" (Winograd and Flores, 1987:12). It makes
more sense to compare expert systems to word processing than to treat them
as mental prostheses. Word processors are not intelligent but enable us to
act effectively in a particular domain, the preparation of text. Expert systems
that supply aids for accomplishing definite tasks have a similar relation to
professional activities.
These aids make possible a new form of human-machine interaction that
gives the illusion of partnership. But however "intelligent" it may appear to
be, the computer is not a mind but "a structured dynamic communication
medium that is qualitatively different from earlier media such as print and tele-
phones" (Winograd and Flores, 1987: 176). It is the programmer and those
who use it who are engaged in communication, not the computer system.
This view leads to a revalorization of the communicative functions of
computers. A new field of "collaborative technologies" has emerged to adapt
computer programs to the needs of work groups. Instead of appearing as tools
for individuals, programs are designed as "groupware" for use by a whole
team (Johansen, 1988). The social and technical dimensions of computer-
ized activity are integrated here in a way that recalls Hirschhorn's commu-
nication theory of automated machinery and Zuboff s discussion of the
textualization of work.
The stakes in this debate over artificial intelligence are not merely techni-
cal. If we understand computers rationalistically, as automata, we prepare a
revised self-understanding along the same lines. People become information
processors and decision makers, rather than participants in shared commu-
nicative activity. "Computer systems can easily reinforce this interpretation,
and working with them can reinforce patterns of acting that are consistent
with it" (Winograd and Flores, 1987: 178).
Considered as a communication medium, the computer is an environ-
ment for an increasing share of daily life. In this conception, computers are
 Postindustrial Discourses 107

not "images of man" but domains in which we act and which shape us in
return. As Hirschhorn suggests, postindustrial "technology can potentiate
latent cultural trends. Control system failures may help to bring out in the
culture a developmental concept of the self, a concept that leads people to
seek out learning opportunities throughout their lives" (Hirschhorn, 1984:
4). One of the chief obstacles in this path is the hidden cultural agenda of
industrial design. "In their search for fail-safe systems engineers demonstrate
the hubris of most design professions. The designers of a machine, a build-
ing, or a policy are attempting to imprint their minds on other people's lives"
(Hirschhorn, 1984: 86). This attempt is not merely a theoretical error but
reflects the practical requirements of the capitalist technical code, with its
overriding emphasis on operational autonomy.
The design of computers is thus humanly significant as well as instrumen-
tally important, for "in designing tools we are designing ways of being"
(Winograd and Flores, 1987: xi). Winograd and Flores call this "ontological
designing." They write, "In ontological designing, we are doing more than
asking what can be built. We are engaged in a philosophical discourse about
the self—about what we can do and what we can be" (Winograd and Flores,
1987: 179). That discourse, I would add, is also political.
This discussion of artificial intelligence leads to the same conclusion as
the earlier discussion of automation. The place computers are intended to
hold in social life is intimately connected with their design. Systems designed
for hierarchical control are congruent with rationalistic assumptions that treat
the computer as an automaton intended to command or replace workers in
decision-making roles. Democratically designed systems must instead re-
spond to the communicative dimension of the computer. As a medium it
facilitates the self-organization of human communities, including those tech-
nical communities the control of which founds modern hegemonies.

The Myth of Automatism

Although technologies are first and foremost tools for solving practical prob-
lems, they are not fully understandable in functional terms. This is especially
true in cases where their function is itself in dispute. As we have seen with
computers, these disputes are not merely technical but go to the cultural sig-
nificance of the technology. The critical theory of technology is therefore a
cultural theory.
Jean Baudrillard suggests a semiological approach to understanding cul-
tural investments in technology. He argues that technical objects have an
108 THE AMBIVALENCE OF THE COMPUTER

equivalent of "denotation" through their function, and "connotation"

through their relation to the fantasies and sociopsychological needs of those
they serve (Baudrillard, 1968:15-16). Ambiguities in the definition of a tech-
nology such as the computer are resolved through interactions between de-
signers and users in which the still fluid boundary between connotations and
denotations is fixed.11
For his understanding of the functional aspect of technologies Baudrillard
relies on the French philosopher of technology Gilbert Simondon. Accord-
ing to Simondon, technical objects generally begin as loose concatenations
of separate mechanical structures, each devoted to a single function. As the
object becomes more technically elegant, single structures incorporate
multiple functions, and powerful synergisms emerge from the interactions
between structures. This type of development, which Simondon calls "con-
cretization," defines an immanent criterion of progress (Simondon, 1958:
chap. 1).
Technological "connotations," on the other hand, lack a basis in the struc-
ture of technical development and may invest machines with inappropriate
functions (Baudrillard, 1968:14). Where this occurs, the technical object does
not advance toward a higher stage, but instead becomes complex and cum-
bersome. The evolution of the automobile in the 1950s offers a case in point.
As cars became symbols of prosperity and sexual prowess, they grew in size
and weight; their gadget-encrusted bodies were burdened with fishtails and
heavy chrome bumpers. Needless to say, these dinosaurs were less efficient
as means of transportation precisely to the extent that they were better at
serving their symbolic functions.
These theories are interesting, but they are too deterministic: they assume
the existence of purely technical criteria of progress, but technical develop-
ment generally opens onto several different paths of concretization. The alter-
natives are signified first in connotations that gradually determine shifts in
the very definition of the technology. Thus, some of the goals that were clum-
sily pursued by the automobile industry in the 1950s, such as improved com-
fort, have been attained today through concretizing innovations in design.
The floating living room is gone, but more appropriate solutions to the prob-
lem have been found through improving suspensions and seat design.
The case of the computer suggests, however, that not all paths are equal:
the managerial ideology of total control, like the rationalistic ideology of
artificial intelligence, responds to fantasies that distort technical development
for political purposes. These ideologies are expressed in the discourse of
"automatism." Here we can see a connotative dimension of a new technol-
ogy in the very process of conversion into a denotation. Ancient dreams of
 Postindustrial Discourses 109

power, embodied until modern times in emblematic objects designed to

exemplify the human capacity for godlike creativity, are confounded with
the actual workings of society and the mind as automatism captures the
modern imagination. The fantasy of the totally automated factory, cleansed
of human effort and the necessity of employing obstreperous workers to
supply it, replaces the innocent dreams of earlier times. As Noble puts it,
"Thus did the capitalist mentality appropriate the primitive enchantment
with automation and turn it to practical and pecuniary ends, where it now
fueled fantasies not of automatic birds and musicians but of automatic fac-
tories" (Noble, 1984: 58). These same dehumanizing connotations of the
computer appear in the notion of a mechanical mind.
What is the peculiar fascination of automatic functioning? Baudrillard
addresses this question in an interesting discussion of its symbolic signifi-
cance. He begins by dismissing the idea that the pursuit of automaticity is
technically motivated. Automatism does not respond to the rational drive
toward increasing efficiency and technical concreteness. In fact, it compli-
cates objects needlessly, making them more effective as symbols of pure
technicity at the price of rendering them ever more elaborate, fragile, and
rigid. Machines actually progress not through automatism but through in-
creased flexibility and responsiveness to more subtle external instruction.
Automatism is thus not rational but contains "the imaginary truth of the
object," our fantasy of mechanical perfection (Baudrillard, 1968: 156).
Carried to the limit of its possibilities, automatism is exemplified in the
useless gadget, a marvel of purposeless complexity. Baudrillard calls this the
"functional delirium" of technique, a kind of baroque predilection for
complexity that encrusts otherwise useful devices. The gadget is a technical
object that "no longer obeys any other necessity than that of functioning"
(Baudrillard, 1968: 159-160). It may possess some ostensible purpose, for
example, to lock or unlock car doors from a distance, but in reality it exists
simply to display its own workings.
Automatism is a fantastic way of experiencing "technicity," that is to say,
what is essentially technical in machines. Through automatism, technicity
appears as a symbol of pure operativity, signifying not some specific techni-
cal function but an imaginary investment of the world as a whole by tech-
nique: "Automatism is the object acquiring the connotation of an absolute
in its particular function" (Baudrillard, 1968: 153-154). The clever device
that automatically cores apples or magnetically suspends a pen in an upright
position serves not so much a practical need as a sort of functional super-
stition that is comforted by the thought of "nature as a whole reinvented
according to the technical reality principle" (Baudrillard, 1968: 164).
110 THE AMBIVALENCE OF THE COMPUTER

Baudrillard rejects the widespread belief that our problems are due to the
rapid advance of technology while social science and moral reflection stag-
nate. With the fantastic demand for automatism, technology and morality
are both caught up in the same contradiction. "In our technical civilization
. . . techniques and objects suffer the same servitudes as men" (Baudrillard,
1968: 175). Consumer society exemplifies technical failure as well as moral
regression through the corruption of design and conception by symbolic
demands that block the concretization of technology. Baudrillard concludes:
"Between men and the world, technique can be an effective mediation: that
is the hardest path. The easiest path is that of a system of objects which in-
terposes itself as an imaginary solution to every sort of contradiction, which
short-circuits, so to speak, the technical order and the order of individual
needs, exhausting the energies of the two systems" (Baudrillard, 1968:183).
Baudrillard's analysis converges with that of Norbert Wiener, one of the
founders of cybernetics and a skeptical observer of early automation. Wiener
warned that any machine capable of making decisions would either have no
capacity to learn, in which case we would not be wise to place much reliance
on it, or it would have such a capacity—it would be a true automaton—in which
case there is no guarantee its decisions would be acceptable to us. "For the man
who is not aware of this, to throw the problem of his responsibility on the
machine, whether it can learn or not, is to cast his responsibility to the winds,
and to find it coming back seated on the whirlwind" (Wiener, 1950: 212).

Technology and Finitude

The rationalization of modern societies has been carried out by subjects—

capitalists or government bureaucrats—whose defining characteristic is their
operational autonomy. This fact is articulated through images drawn from
the sphere of technique because modern power relations resemble the op-
eration of a machine. Thus, Lukacs argues that modernity brought with it
a specific type of formal rationalism that, in the seventeenth century, iden-
tified the workings of the universe with the mechanical creations of the
human hand and brain. The scientific logic of classification and cal-
culation is the metaphoric equivalent of the techno-logic of machinery. The
input of data, the raw material, is worked over by the axiomatic of the
system, yielding an output of truths, goods, or wealth. The identity of
syllogistic-mathematical procedures with mechanism inspired science and
technology, and gave a distinctive practical aspect to modern Reason that
culminates in the computer.
 Postindustrial Discourses 111

Descartes was the first to articulate the self-understanding of the subjects

of this new form of rationality. De Certeau calls the shaping of an "interior-
ity" from which to act on a correlated "exterior" the "Cartesian gesture" (de
Certeau, 1980: 85). This gesture lays the foundation of leadership in mod-
ern societies. One manages organizations from without, rather than as a full
member of the community. There is thus an inner link between the "posses-
sive individualism" of emerging capitalist society and the Cartesian cogito,
which is also a figure of alienation from immediate involvements.12 The struc-
tural parallel between these subject positions is the basis for the social gen-
eralization of modern Reason, its transformation from an intellectual method
into the cultural basis of modern society. A technical paradigm of thought
and action triumphed in this transformation.
A finite subject is constituted by its own actions on the world. In using
technique it is shaped by technique and becomes something quite different
from what it intends. Meanwhile, as it incorporates its objects into technical
systems, it changes them from immediate "natural" objects into mediated
social objects. Thus, prior to the actual unfolding of any particular technical
action, the subject and object have already been restructured. Modernity is
the general collapse of religious and folk tradition in the face of this process.
As a vision of the world, it is characterized by the fundamental misrecognition
of finitude associated with the naive self-understanding of technical subjects.
A little god, the modem subject sees itself as autonomous, as indepen-
dent of the system on which it operates through technical means. The mod-
ern subject places itself beyond the web of consequences of its own actions.
From this beyond it elaborates projects based on formal, mechanistic think-
ing and a representation of reality as essentially an object of technical control.
But as Adorno and Horkheimer argue, what ultimately conquers humanity in
these projects is not a particular elite but a new form of life based on total
technologization.
One of the paradoxes of the twentieth century is that, just as the entire
world was enrolled into Western technological rationalism, the foundations
of science and philosophy changed radically, undermining the assumption
that the subject can remain external to the systems it designs and operates.
Modern physics, philosophy, and biology, not to mention avant garde art
and literature, increasingly challenge what Heidegger calls the "onto-theo-
logical" constitution of the subject as "beyond" objectivity (Heidegger,
1977a). These challenges demystify the procedures by which the illusion of
technical interioriry is produced and suggest the need for a new understand-
ing of rationality. Although not always progressive, these theoretical inno-
vations open potential political challenges; reason appears as inherently
112 THE AMBIVALENCE OF THE COMPUTER

ambivalent and can either support a technocratic order or subvert it depend-

ing on how it is deployed socially.
Yet the new ideas have had little lasting social impact. Heidegger and
Lukacs lent their thinking to totalitarian schemes that exaggerated to the
breaking point the very things they were attempting to overcome. The rela-
tively justified reaction in favor of liberalism that began after World War II
continues with ups and downs to this day. At its best, it saves what can be
saved within the existing system, but liberalism appears now as a practical
necessity, in the absence of workable alternatives, rather than as a solution
to the fundamental problems of modernity.
As a culture, modern technological rationality is not dependent on science
and philosophy but on hierarchical forms of social organization and tech-
nologies such as the computer. As hierarchy obtains a technical function,
social subjects are placed in a technical inferiority from which they control
the systems on which they act. To attack the belief that hierarchy is destiny
is an essential philosophical task requiring new concepts of social and tech-
nical action based on a radical acceptance of human finitude. Ethics has al-
ways involved the recognition that our actions on the world are ultimately
actions on ourselves, on our way of being in the world and on our very nature.
This insight must now be extended to technology as well.
At the beginning of the democratic era, Saint-Just expressed this contra-
dictory structure through the figure of speech called "paradoxisme" in the
rhetorical theory of his day. The National Assembly "ingeniously enchained
the people with their own freedom"; "The people is a submissive monarch
and a free subject" (Saint-Just, 1963: 27, 39). These are paradoxes of reflec-
tion in which the subject is also the object. They express the emergence of
political rationality in the peculiarly modern form of self-consciousness.
This reflexive logic is diametrically opposed to the one-way movement
down a Russellian hierarchy from metalanguage to object language. The
politics of self-organization has the form of a "strange loop," which Douglas
Hofstadter describes as a "phenomenon [that] occurs whenever, by moving
upwards (or downwards) through the levels of some hierarchical system, we
unexpectedly find ourselves right back where we started" (Hofstadter, 1979:
10). The most elegant illustration of Hofstadter's concept is Escher's "Draw-
ing Hands," an etching of two hands, each of which holds a pencil and draws
the other. Strange loops play many roles in our lives, appearing in various
reversals of agency with which we are all familiar. For example, the child on
whom we impose rules may insist we follow them. Or consider the case of
President Clinton's rising ratings during the impeachment, which pollsters
concluded was due to an attempt by those polled to use polling to stop the
 Postindustrial Discourses 113

process, regardless of their opinion of Clinton. Controlling people exposes

the controllers to a counteraction on the part of their human objects and
the strange loop is closed.
Political democracy institutionalizes such reversals in the electoral sub-
ordination of the rulers to the ruled. Democratic notions of management and
political organization can be rethought on these terms as rational systems
without assuming an external source of control (Jantsch and Waddington,
1976). A self-referential logic of action is needed to grasp a democratic pro-
cess that would have as its goal not escape from the community to a com-
manding position above it but internal self-development in common with
others.
 5

The Factory or the City

Which Model for Online Education?

Technology and Modernity

Much recent discussion of the Internet emphasizes its promise of epoch-

making changes in our lives. In no domain are these anticipated changes more
radical than in education. We are told that the substantive content of instruc-
tion can now be delivered better by computers than by teachers. Are we on
the verge of a fundamental transformation of all our assumptions about
education as we enter a postindustrial information age, or are we instead
witnessing significant but more modest changes in education as we know it?
As a participant in the early development of online education, I hope to be
able to bring a touch of realism to the debate.
The debate is not limited to education, which is simply one among sev-
eral fronts in the struggle to define the society of the future. The meaning of
modernity is at stake in this struggle. One possible outcome is a society re-
flecting in all its institutions the logic of modern production, obsessed by
efficiency achieved through mechanization and management. The Internet
could serve this technocratic project in hitherto protected domains such as
education. But one can also envisage a very different outcome modeled not
on the factory but on another modern institution, the city.
The city is the place of cosmopolitan interactions and enhanced com-
munication. Its god is not efficiency but freedom. It is not dedicated to the
rigid reproduction of the same, the "one best way," but to the flexible test-
ing of possibilities and the development of the new—not hierarchical con-
trol but unplanned horizontal contacts; not simplification and standard-
ization but variety and the growth of the capacities required to live in a

114
 The Factory or the City 115

more complex world.1 The Internet extends this urban logic in a radically
new way.
The question implied in the debate over educational technology is there-
fore: Which model, the factory or the city, will shape the future of educa-
tion? Online education can serve either strategy in different technical con-
figurations. Automated education is certainly possible although at the price
of a redefinition of education itself. The generalization on the Internet of a
more traditional concept of education centered on human interaction would
facilitate participation by underserved groups and might raise the cultural
level of the population at large.
This latter prospect recalls a significant precedent. It is clear that the gradual
disappearance of child labor and the consequent establishment of universal
education has transformed modern societies and shapes the kind of people who
inhabit them. To the extent that we are capable of understanding the complex
technologized world around us and acting independently within it, this is owing
to the extended time for learning modern societies allow.
However, there is a strong link between education and the division of
labor, with the latter determining the former over long periods. Where
deskilled production governs educational expectations, cultural levels remain
relatively low. Marx saw no escape from this situation so long as capitalism
survived to impose its division of labor. But capitalism is alive and well long
after the demand for skill has risen to encompass a significant fraction of the
labor force. The consequence has been tremendous educational dynamism.
Adult education, for example, now embraces more than half the students in
American college programs, a reflection of the shortage of competencies in
the labor pool.
Yet one wonders how far this trend can go under capitalism. In the first
place, the growing demand for educated labor in the advanced capitalist world
is accompanied by the export of manufacturing to poor countries. While
skilled and unionized manufacturing workers suffer steep declines in income
and job security in the advanced countries, old-fashioned patterns of indus-
trialization appear everywhere else. The net effect may well be a global in-
crease in deskilled work despite the contrary appearance in places such as
Silicon Valley. Second, business leaders appear to be increasingly alarmed
by the high cost of education, which is now the largest budget item in prac-
tically every advanced capitalist nation. In the United States, the promise of
the Internet has inspired an ideological offensive in favor of automating and
deskilling education. These problems suggest the continuing relevance of
critical theory to educational policy.