The Constitutively Social Character of Expertise
 
Steve Fuller, Dept of Communication, University of Pittsburgh,
Pittsburgh PA 15260, USA
 
1. A brief history of expertise
"Expert," a contraction of the participle "experienced," first
appeared as a noun in French at the start of the Third Republic
(about 1870). The general idea was much the same as today, namely,
someone whose specialized training enables him [sic!] to speak
authoritatively on some matter. However, the original context of
use was quite specific. The first experts were called as witnesses
in trials to detect handwriting forgeries. These people were 
experienced in discriminating scripts that appeared
indistinguishable to the ordinary observer. Thus, the etymological
root of "expertise" in "experience" was carried over as the
semantically heightened way in which the expert experienced the
relevant environment. In contemporary parlance, the tasks that
originally required the service of an expert principally involved
"pattern recognition," except that the patterns recognized by the
expert were identified in terms of an implicit explanatory
framework, one typically fraught with value connotations, as in the
case of identifying a script as a "forgery." 
    When evaluating the likelihood that a script was forged,
experts were not expected to publicly exhibit their reasoning. They
were not casuists who weighed the relative probability that various
general principles applied to the case. Rather, it was on the basis
of an expert's previous experience of having successfully
identified forgeries that his judgment was now trusted. This is not
to say that no one could contest expert judgment, but he would have
to be another expert, a colleague. If no colleague came forward to
testify against an expert's judgment, then the judgment would
stand. The climate of collegiality that harbored the mystique of
the expert led journalists of the Third Republic to distinguish
experts from the "lay" public, thereby conjuring up a clerical
image redolent of the secular religion that Auguste Comte's more
zealous followers had been promoting under the rubric of Positivism
(Williams, 1983: 129). 
    Moreover, experts were contrasted not only with the lay public
but also with intellectuals. This point is important for
understanding the source of what might be called the epistemic
power of expertise. An intellectual takes the entire world as fair
game for his judgments, but at the same time he opens himself to
scrutiny from all quarters. Indeed, the intellectual's natural
habitat is controversy, and often he seems to spend more time on
defending and attacking positions than on developing and applying
them. In contrast, the expert's judgments are restricted to his
area of training. The credibility of those judgments are measured
in terms of the freedom from contravention that his colleagues
accord him. The mystique of expertise is created by the impression
that an expert's colleagues are sufficiently scrupulous that, were
it necessary, they would be able and inclined to redress any misuse
or abuse of their expertise. The fact that they do not means that
the expert must be doing something right. 
   Collegiality enables experts to exert what both Plato and
Machiavelli would have recognized as an ideal form of power. In its
ideal form, power thrives on a counterfactual that never needs to
be realized--in less charitable terms, a persuasive bluff. If a
prince's enemies believe that the prince could squash any uprising,
the enemies will lie low, and the prince will seem invincible;
however, if the enemies challenge the prince, and the prince
defeats them only with great difficulty, then the air of princely
invincibility will disappear, and the prince will need to prepare
for redoubled efforts by his enemies in the future. Thus, ideal
power is brought down to the level of brute force (Botwinick, 1990:
133-80). Trial attorneys continue to exploit this point whenever
they try to undermine the very possibility of expertise in a field
by pitting particular experts against one another. The lawyers do
not expect a definitive judgment to emerge from the crossfire;
rather, they expect to show that no such judgment can be rendered. 
    Philosophers have traditionally shared the attorney's desire to
dissipate the power of expertise, but without in the process
undermining the possibility of knowledge. Be it embodied in machine
or human, philosophers have looked askance at the epistemological
status of expertise. For Karl Popper, the expertise conferred on
those trained in the special sciences serves to strategically
contain critical inquiry, as the evaluation of a knowledge claim
depends on the credentials of the claimant. The sociology of
knowledge perennially incurs the wrath of philosophers because it
seems to condone this tendency, which culminates in the formation
of scientific guilds, or disciplines. These, in turn, divert
inquiry away from questions of fundamental principles that go to
the heart of disciplinary identity. Disciplines proliferate
explanatory frameworks (jargons, if you will), while "science," in
the philosophically honorific sense, unifies such frameworks. Not
surprisingly, then, Popper regarded Kuhnian "normal science" as a
"danger" to the advance of knowledge, if that advance is measured
in terms of explanatory comprehensiveness, or the Newtonian virtue
of explaining the most by the least (Popper, 1970). 
    Cognitive scientists display a similar disdain when they
contrast the "knowledge-based" or "domain-specific" character of
expert systems to more general purpose problem-solving machines
that utilize principles that cut across domains. What is often
called the "orthodoxy" in cognitive science holds that an adequate
theory of cognition is not to be found simply by compounding a
number of distinct expert systems, just as an adequate theory of
knowledge requires more than simply articulating the research
conventions of all the special sciences (Haugeland, 1984). The
hope, in both cases, is that whatever principles govern the special
cases are not unique to them.
 
2. Some senses of the social
To someone whose work is primarily in the design of expert systems,
it may seem odd to juxtapose philosophical suspicions about human
and machine expertise in the way I have. Perhaps this juxtaposition
points to a residual positivist hankering that cognitive science
and epistemology (or philosophy of science) share for the "unity of
science." Without casting doubts on this diagnosis, I nevertheless
want to stress some philosophically disturbing aspects of expertise
that have yet to fully grip theorists in this area. As a first
pass, these aspects turn on the following truism: Expertise is a
constitutively social phenomenon. Indeed, as a card-carrying
"social epistemologist" (Fuller, 1988, 1993a, 1993b), I am
committed to the position that expertise can be exhaustively
analyzed as a social phenomenon. But before too many eyebrows are
raised, let me begin by listing--in descending order of
intuitiveness--four distinct senses in which expertise is
constitutively social:  
     (1) The skills associated with an expertise are the product of
     specialized training. Expertise cannot be picked up casually
     or as the byproduct on some other form of learning.
     (2) Both experts and the lay public recognize that expertise
     is relevant only on certain occasions. No expertise carries
     universal applicability.
     (3) The disposition of expertise is dependent on the collegial
     patterns of the relevant experts. Protracted internecine
     disputes over fundamentals typically erode expertise. 
     (4) The cognitive significance of an expertise is affected by
     the availability of expert training and judgment, relative to
     the need for the expertise. Too many experts or too little
     need typically devalue the expertise in question. 
So far, I have concentrated on sense (3), given its salience in the
historical development of the expert as a social role distinct from
that of the layperson, the intellectual, and, as we have just seen,
even the scientist. However, all four senses echo the twin themes
of boundedness and compartmentalization as essential to the
definition of expertise. The cognitive science literature offers
several ways of articulating these themes: Simon's heuristics,
Minsky's frames, Schank's scripts, Fodor's modules, and, most
abstractly, Pylyshyn's cognitive impenetrability. Of course, these
terms do not divide up the mind's labor in quite the same way. A
similar proviso would have to be attached to sociological markers
of expertise, such as "indexicality" and "functional
differentiation" (cf. Cicourel, 1968; Knorr-Cetina, 1981). But for
our purposes, the most striking comparison may be between, so to
speak, the cognitive and political impenetrability of expertise,
the so-called autonomy of the professions (cf. Abbott, 1988). 
    The analogy I wish to draw here is not particularly difficult
to grasp, but doing so may alert us to the conceptual baggage that
is unwittingly imported in the images we use to characterize
expertise. The profession that has most jealously guarded its
autonomy--scientists--has often struck a bargain to ensure that
professionally produced knowledge remains both cognitively and
politically impenetrable. From the charter of the Royal Society in
seventeenth century Britain to the guild right of Lehrfreiheit
(freedom of inquiry) that German university professors enjoyed
under Bismarck, the following two conditions have been met (cf.
Proctor, 1991):
     (A) The state agrees not to interfere in the internal
     governance of the profession, on the condition that the
     profession does not interfere in the governance of the
     state.
     (B) The state agrees to protect the profession from
     others who might want to interfere with its governance
     (e.g. other professional, political, or business
     interests), on the condition that the state is given the
     first opportunity to appropriate the knowledge produced
     by the profession, where appropriation includes the right
     to prevent others from subsequently appropriating the
     knowledge (e.g. for reasons of national security).
    Most theories of the mind in cognitive science are fairly
explicit in distinguishing an executive central processor or
general problem-solver from domain-specific modules that function
in relative autonomy from this unit. Sometimes (as in Simon, 1981;
Minsky, 1986) the political imagery of governance is quite strong.
However, it is not just any old image of governance, but one that
is characteristic of twentieth century thinking about the state,
namely, democratic pluralism (cf. Held, 1987: 186-220). The
pluralist portrays the state as mediating competing factions in a
large democratic society by enabling the factions to flourish
without letting any of them override each other or the national
interest. The learning process of big democracies consists of these
factions settling into interest groups and, ultimately (and
ideally), professionally governed associations whose interaction is
founded on recognition and respect for each other's work as
essential to the business of society. Gradually, then, the state's
role as mediator recedes to that of chairman of a corporate board
of directors. The only point I wish to make here is that this is
not the only, nor necessarily the most desirable, image of
democratic governance. It might repay the effort for cognitive
scientists to examine alternative forms of governance as a source
of new images for arranging the parts of the mind. 
 
3. There may be less to expertise than meets the eye.
To explore the social character of expertise implied in senses (1)
and (2), consider what might be called a behaviorist and a
cognitivist account of how expertise develops (cf. De Mey 1982:
216):
     Behaviorist: Expertise is shaped from repeated encounters
     with relevant environments, such that increased exposure
     smooths out the rough edges in the expert's practice
     until it stabilizes at a normatively acceptable standard,
     which can then be applied, "off the shelf" as it were, in
     subsequent encounters.
     Cognitivist: Expertise consists of a core set of skills
     that are elaborated in a variety of environments, most of
     which are unforeseen. These elaborations are stored and
     themselves elaborated upon in subsequent encounters, all
     of which serves to confer on expertise a sophistication
     that is evident to the observer but difficult to
     articulate.
Thus, whereas the behaviorist sees expertise becoming more
stereotyped in practice, the cognitivist sees it becoming more
nuanced. In the ongoing dispute between cognitivists and
behaviorists, it is often asserted that the cognitivist won this
round. But before the final verdict is delivered, I wish to offer
support for the behaviorist by way of revealing the hidden social
hand of expertise.
    To hear cognitivists (not to mention phenomenologists) emote
about the "nuanced" and "craftlike" character of expertise, one
would think that an act of professional judgment was tantamount to
a magic trick, one in which the audience has attended a little too
closely to the magician's gestures and not enough to the
circumstances under which the illusion transpires. A professional
magician does not perform tricks on demand, say, by adapting his
performance to play off the specific gullibilities of his audience.
Of course, the magician adapts somewhat to his audience, but before
he even agrees to display his expertise, the stage must be set just
right, and the audience must already be in the right frame of mind
to be receptive to the "magic moment." A magician who is too
indiscriminate in his eagerness to please is bound to look bad very
quickly. An instructive case in point is the odd magician who
submits his performances to the strictures of the experimental
method (Collins & Pinch, 1982: on psychokinesis). 
    It is worth noting that, in these scientific performances, the
magician fares no worse than the expert witnesses in medicine and
psychiatry whose testimony is routinely heard--and believed--in
court (Faust, 1985). The reliability and validity of expert
judgment in experimental settings are low all around (Arkes &
Hammond, 1986). Perhaps the most celebrated historical case of the
hubris of overextended expertise was the rise and fall of the
Sophists in 5th Century B.C. Athens. After some fair success as
teachers of rhetoric and confidants of the ruling class, the
Sophists began to offer their services in competition with more
established forms of knowledge in virtually every domain. According
to a recent historical study (De Romilly, 1992), the Sophists were
perceived as opportunistic colonizers of conventional practices who
failed to cultivate the trust required for their own practices to
succeed. Consequently, the Sophists were soon ridiculed by the
people we now regard as the founders of classical philosophy and
drama--people with a greater surface respect for tradition than the
Sophists had.
    Proponents of the operant conditioning paradigm in behaviorism
would recognize what is going on here. Successful experts realize
that the secret to their success lies in noting the sorts of
situations in which clients come away most satisfied with expert
judgment (and hence reward the expert appropriately) and then
maximizing the opportunities for situations of that sort to arise
in the future. In short, the smart expert controls the
contingencies by which her behavior is reinforced (cf. Haddock &
Houts, 1992). She does not easily submit to an experimental test of
her expertise, over which she exerts little control. 
    The strategy for achieving the relevant sense of control
involves several tactics: 
     (i) pre-selecting clients before officially engaging in
     treatment, typically through an interview process, an
     important aspect of which is to gain the confidence of
     the prospective client;
     (ii) learning to refuse certain clients, no matter how
     lucrative their business might be, if it appears as
     though they will resist treatment or, in some way, be
     likely to make the expert look bad in the long run;
     (iii) persuading the prospective client that her avowed
     problem is really one that the expert has seen many times
     before; often this is done by first showing the client
     that she has not conceptualized her problem correctly;
     once the problem receives its proper formulation--which
     neatly coincides with what the expert is in the best
     position to treat--then treatment can begin;
     (iv) obscuring any discussion of the exact method of
     treatment by recasting the client's problem in the jargon
     of the expertise, which will presumably lead the client
     to infer that anything that must be described in such
     esoteric ways must be subject to a treatment that would
     be equally difficult to convey. 
The most obvious result of these four tactics is to shape the
prospective client's behavior so that her problem falls into one of
the stereotyped patterns with which the expert is familiar. A less
obvious, but equally important result is that the expert can now
exert spin control over how the client understands her situation
after the treatment. If the client's problem is solved, the expert
can confidently claim credit for the solution. No suspicions of
"spontaneous remission" are likely to be raised, especially if the
client has paid dearly for treatment. But perhaps more important,
if the client's problem remains unsolved after the treatment, then
the expert can claim, with only slightly less confidence, that
other unforeseen factors intervened, including the client's own
recalcitrance. Whatever happens, therefore, reinforces the expert's
competence and integrity.
    On the basis of these considerations, I conclude that the key
to understanding the distinct character of expertise may lie less
in its associated skills than in the discretionary control that the
expert has in deploying those skills. (Philosophers of science and
cognitive scientists who are uncomfortable with brute talk of
"control" may substitute this piece of genteel intellectualism:
Experts have heightened metaknowledge of the ceteris paribus
clause, or relevance conditions, for applying their expertise.)
Thus, if the above pre-selection strategy fails to mold the client
into shape, the expert can then tell the client that the problem
lies outside her expertise, which will probably cause the client to
believe that her problem remains unsolved, not because the expert
was incompetent, but because the client has yet to locate the right
expert, which may itself reflect the client's own failure to
understand the nature of her problem. The element of trust crucial
for the maintenance of expertise may be seen in the willingness
with which the client holds herself responsible for an expert's
inability to come to grips with her problem (cf. Gambetta, 1988). 
    From what I have said so far, it may seem that my thinking
about the social dimension of expertise has been strongly based on
psychiatric encounters, which have often been subject to
unflattering depictions as confidence games. However, the same
observations apply, perhaps with greater import, to experts
operating in the arena of public policy, especially those trained
in medicine, engineering, or economics. For, the biggest single
problem facing the future of democracy may be cognitive
authoritarianism, the tendency to cede an ever larger share of the
realm of participatory politics to expert rule (Fuller, 1988: 277-
88). The conversion is accomplished as government officials become
convinced that the public has ill-formed conceptions of its own
needs, needs that are best shaped and addressed by the relevant
experts. When government fails to act speedily on this conversion,
and hence does not clear the political environment to enable the
expert's stereotyped knowledge to take effect, the expert will
often appear as a moral censor, appealing to his special knowledge-
-which would supposedly be efficacious if politicians secured the
relevant background conditions--as a norm against which the state
of society is criticized. 
    But can all this talk about the strategically discretionary
character of expertise be applied to computerized expert systems?
I do not see why not. Suppose a knowledge engineer has been asked
to design an expert system that will offer advice on playing the
stock market. After some time, the knowledge engineer returns with
a product that she announces was constructed from in-depth
interviews with four of the best stock analysts on Wall Street,
making sure that respected spokespersons for all the relevant
market perspectives--fundamentalists, chartists, insiders, and
traders--were canvassed (cf. Smith, 1981). The effect of this
pedigree on the client will be similar to that of the diploma and
license that hang on the wall of the human expert's office and
invariably engage the client's peripheral vision during a
consultation. If the knowledge engineer designs the interface with
protocols that make interaction between client and expert appear
stilted, then the client will probably interpret that to mean that
the expert is concerned with getting to the heart of the client's
problem without dragging in superfluous information. Likewise, if
the expert seems to give the client advice that causes her to lose
money in the market, then the client may wonder whether a human
expert could have really done any better, or that perhaps she did
not input all the information that was relevant for the expert
system to provide better advice. Moreover, the client's inclination
to assume responsibility for the bad advice increases with the
amount of money that she had to originally spend to purchase the
system. That an AI pioneer, Joseph Weizenbaum (1976), should appeal
to moral, rather than technical, grounds for restricting the use of
expert systems reflects the propensity of clients to invest the
same level of trust in computers as in human beings. 
 
4. Global constructivism and the political economy of expertise
The senses (1)-(3) in which expertise is constitutively social are
consistent with a constructivist sociological orientation (cf.
Knorr-Cetina, 1981). Constructivists typically minimize the
attribution of intrinsic properties to cognitive agents. Instead,
they unpack such so-called intrinsic properties into relational
ones, in which the relata are two mutually interpretive agents who
jointly negotiate who will be credited with which properties. As we
have seen, an adept expert can shift the burden of responsibility
onto the client for the unpleasant consequences of following expert
advice. But for the constructivist, there is no "fact of the
matter" about whether the expert's incompetence or the client's
recalcitrance is to blame, until the transaction has actually taken
place. In discussions of cognitive science that acknowledge that
society is more than a metaphor for the mind, constructivism is
often presented as the sociological perspective. Empirically
speaking, the dominance of constructivism cannot be denied, but I
will ultimately appeal to sense (4) of expertise's sociality in
order to introduce a different, and more comprehensive,
sociological perspective.
    Constructivism is a more heterogeneous doctrine than it may
first appear. Weizenbaum and others whose opposition to the
cognitive authority of either computer or human experts is mainly
moral presuppose that such experts can, indeed, exercise all manner
of authority, if they are not limited by convention. However, a
milder species of constructivism is represented by Daniel Dennett's
(e.g., 1987) instrumentalist approach to intentionality, which
makes one's cognitive status dependent on another's interpretive
stance. Dennett's constructivism is "asymmetrical," in that he does
not grant the computer the same degree of agency as most human
beings in constructing their respective identities. By contrast, a
committed sense of symmetry is the hallmark of the more radical
constructivists, the ethnographers who can be increasingly found on
the sites where knowledge engineering occurs (cf. Greenbaum & Kyng,
1991). Here knowledge engineers, human and computer experts,
clients, and other people and artifacts are portrayed as engaged in
a mutually reinforcing cooperative venture. I will now address the
limitations of this version of constructivism.
    Drawing on the work of cultural anthropologists, especially
Clifford Geertz (1983), ethnographic constructivism makes much of
the "local" character of expert knowledge, which is brought out
very clearly in the design of expert systems. The idiosyncracy of
locales is brought out in the knowledge engineer's interviews with
both client and expert, followed by the process of adapting the
expert system to the client's specific needs and abilities. These
ethnographic accounts are meant to stand in striking contrast to
the accounts typically given by the designers of AI systems in
basic research settings, who often influence the way in which the
applied researchers, the knowledge engineers, conceptualize their
activities. Specifically, workers in AI tend to attribute
properties to their programs (typically once embodied in a machine)
that the ethnographers would prefer to regard as "boundary objects"
in terms of which the identities (or cognitive capacities) of a
variety of agents are negotiated. For example, the degree of
satisfaction that the client gets from using an expert system has
implications for how much of human expertise was successfully
transferred to the computer program. While I wholeheartedly endorse
this reinterpretation of expertise as a corrective to the accounts
given by AI researchers, it nevertheless shares the fatal flaw of
its opponents. Ironically, the flaw is the tendency to universalize
from a single case. Let me explain.
    It is one thing to say that all knowledge is local. That is
rather boring. It is another to say that all locales are different
from one another. That is more exciting. Ethnographers infer the
exciting proposition from the boring one all the time. But what
licenses this inference?  Certainly, there have not been enough
ethnographies of knowledge engineering to license the exciting
proposition as an empirical generalization. In fact, the few
ethnographies available do not depict locales so radically
different from one another. My point here is not to decide the
issue by philosophical argument but to observe that the
ethnographic appeal to locality presupposes a conceptual
cartography whereby one imagines that the spatiotemporal distance
between locales represents a conceptual distance as well. In that
sense, "the local" presupposes "the global," an image of the whole.
Admittedly, sometimes this image turns out to be right; but other
times it doesn't. However, that is a matter for empirical inquiry,
and it is not clear that the ethnogrpahic brand of constructivism
encourages the appropriate sort of empirical inquiry. For, to learn
about the global properties of knowledge engineering, one needs to
discover the pattern by which expertise is distributed across a
representative sample of locales and the aggregated consequences of
such a distribution for the knowledge system as a whole. Here I
start to speak the language of political economy, and to signal the
quest for statistical correlations among variables that are
hypothesized to be salient for understanding how expertise works. 
    What sorts of people are the producers and consumers of expert
knowledge?  Provided with a serviceable answer, we can study the
distribution of expertise by focusing on a cognitively relevant
locus of scarcity: A client has only so much time and money to
spend consulting an expert, be it human or computer. Under what
circumstances does the client feel she has gotten what she has paid
for, and, when she does not feel that way, how is the blame
apportioned: Who receives the lion's share of incompetence--the
client or the expert?  A key to understanding the distribution of
expertise is to see how each side tries to convert its own sense of
frustration into a perception of the other's liabilities. It would
be fair to suppose that expert computers today receive far more
attributions of incompetence than expert humans. A constructivist
would diagnose this difference in terms of the client's lack of
time, imagination, or interest in interpreting the computer as
performing intelligently--perhaps because the client feels either
that she has better things to do at this point, and the computer is
in no position to prevent her from doing them, or that she would
have to end up interpreting the computer as doing something other
than she would have wanted (cf. Fuller, 1993b: 179-85). However, as
people become more accustomed to dealing with expert computers,
this difference in attribution is likely to disappear. But before
concluding that we are projecting a future in which experts of all
sorts are engaged in mutually satisfying relationships with their
clients, we need to consider the aggregated consequences of people
increasingly turning to computers for advice.
    As we have seen, the history of expertise teaches that the
expert is not a universalizable social role. There are no "experts"
in areas that are regarded as commonsense or part of general
education or easy to acquire without specialized training.
Consequently, knowledge engineers are in the curious position of
potentially destroying expertise as they diligently codify it and
make it available to more people in user-friendly packages. While
this consequence is bound to elude any on-site description of the
knowledge engineer's work, it is nevertheless felt by professional
associations that believe that knowledge engineers are indirectly
deskilling their members. For, even as the human expert retains
discretionary control over when, where, and how she uses her
expertise, she may be losing discretionary control at the meta-
level, namely, over who--or what--else counts as an expert in her
field. Librarians have so far been most vocal in their concerns
(cf. Pfaffenberger, 1990), but attempts by many doctors and lawyers
to limit the scope of the interviews they give to knowledge
engineers reflect similar worries. 
    At first glance, it may seem that the proliferation of expert
systems is the ideal vehicle for democratizing expertise, as it
would seem to put expert knowledge within the reach of more people.
Just because the knowledge engineer can extract elements of
expertise from her interviews with experts, it does not follow that
the expertise remains intact once it is programmed into a computer
for a client. After all, if expertise is indeed constitutively
social, then altering the context in which expert knowledge is
deployed should alter the character of the knowledge itself. Such
change may be witnessed in the course of designing the interface
that enables the client to interact with the expert system. Here
the tendency has been to "go ergonomic" by designing interfaces
that require the client to change his ordinary patterns of thought
and behavior as little as possible (Downes, 1987). Less charitably
put, the ergonomic approach reinforces the client's cognitive
biases and thereby minimizes the learning experience that he might
derive from engaging with the expertise as a form of knowledge. A
potentially "dialectical" exchange is thus rendered merely
"instrumental" (cf. Adorno & Horkheimer, 1972). The result is a
spurious sense of autonomy, whereby the client's powers appear to
be extended only because the environment in which he acts has been
changed to his advantage (Fuller, 1986). 
    Thus, while the expert humans may lose some of their power as
their clients increasingly rely on computerized systems, the
clients themselves may not, in turn, become epistemically
empowered. Experts are deskilled without clients being reskilled.
Where has the original power of expertise gone?  That power would
seem to have dissipated somewhere in the knowledge engineering
process, specifically when expertise was converted into a tool that
exerted few of its own demands on the user (cf. Fields, 1987).
    The utopian vision of democratized expertise is foiled by the
simple fact that expertise, and perhaps knowledge more generally,
is what economists call a positional good (Hirsch, 1977). A
positional good is one whose value is directly tied to others not
having it. Economists have generally refused to count knowledge as
a positional good, preferring instead the classical philosophical
position that knowledge is an "ethereal good," one whose value is
not determined by the laws of supply and demand (Fuller, 1992).
However, professional associations realize all too well the
positional character of expertise. 
    The existence of positional goods is the dark secret of the
welfare state. According to welfare economics, capitalism can avert
a Marxist revolution because lingering inequalities of wealth will
be resolved once a level of productivity is reached that enables
everyone to be supported at a minimally acceptable standard. At
most, stabilizing this situation will require a modest
redistribution of income through progressive taxation. Overlooked
in this scenario is that, as more goods are made more generally
available, the perceived value of the goods may decline as they no
longer serve to discriminate people in socially relevant ways
(Bourdieu, 1984). Knowledge-intensive goods display such
positionality effects. Higher education is perhaps the most obvious
case in point: As it becomes easier for people to complete college,
more postgraduate degrees are needed to acquire the same
credentials. Should it become impossible either to stop the
production of degree-holders or to set up additional barriers in
the credentialing process, higher education will then no longer be
seen as imparting an especially valued form of knowledge. Instead,
it will take the place of bare literacy and the high school diploma
as the minimum threshold for entry into the job market. Can
anything be done reverse such positionality effects, or is the
value of knowledge-intensive goods doomed to continual deflation?
    The political scientist Yaron Ezrahi (1990) has argued that the
scientific enterprise has come to consume so many resources and to 
produce so many questionable consequences that we may be reaching
the end of the period (which began with the Enlightenment) when
knowledge is presumed to be a public good. Ezrahi envisages that
scientific forms of knowledge will gradually acquire the social
character of artforms: their support will be privatized and their
products customized to client tastes, which are presumed not to be
universalizable. The expansion of intellectual property law to
cover more instances of "basic research" suggests that Ezrahi's
prognosis is already taking shape (Fuller, 1991). Given their own
interest in customizing expertise for user demand, knowledge
engineers clearly contribute to this overall trend toward 
privatization. Indeed, human experts may soon find the need to seek
legal protection for their expertise, if only to earn royalties
from the expert systems that are based on from it. In that way,
knowledge engineers would not benefit too much from the "Japan
Effect" of learning how to manufacture expertise more efficiently
than the original experts themselves (cf. Weil & Snapper, 1990).
This form of legal protection would, in turn, require a new
category of intellectual property beyond the usual three of patent,
copyright, and trademark. 
    But even if expertise were to become entirely market-driven,
the skills surrounding the expertise would still attract human
practitioners for the same reasons as art continues to do. The
skills would be detached from the fame, fortune, or power that had
been previously tied to them. Most of the perverse consequences of
positional goods rest on such coupling (Crouch, 1983). For example,
higher education is populated by a few people who are interested in
the education process itself and many more who view it as a
credentialing process, the surest route to a job. Decoupling those
two groups would presumably help restore the integrity of higher
education. 
    Knowledge engineers have a crucial role to play in the future
disposition of expertise and knowledge more generally. Customized
expert systems will hasten the demise of expertise and turn
Ezrahi's image into a reality. However, the proliferation of such
systems may also limit the client's potential for cognitive growth.
 
A page from the history of manufacturing may prove instructive in
resolving this dilemma. 
    Once the demand for manufactured products grew to a critical
level, customization yielded to mass production (Beniger, 1986:
291-343). This transition was accompanied by the design of quality
control standards for the mass produced goods. In the process of
defining the minimum level of acceptability for a particular good,
manufacturers effectively forced potential customers to adapt their
behavior to the set dimensions of the good. Typically, these
adaptations were dictated by the manufacturer's desire to cut
costs, but knowledge engineers could collectively set guidelines
for the design of expert systems, the successful use of which
required clients to expand their cognitive repertoire. The sort of
behavioral changes I envision here may be quite subtle. For
example, an online library search system may discourage
disciplinary provincialism by requiring the client to initiate
searches by using protocols that are tied less to the jargon of
specific disciplines and more to the exact topic or problem that
client wishes to tackle. The system's database would, in turn, draw
on the literatures of several disciplines so as not simply to
confirm the course of inquiry that the client would be naturally
inclined to follow (Cronin & Davenport, 1988: 316-27). 
 
5. But is expertise really knowledge?
It remains an open question whether the epistemic power of science
is tied more to its sheer practice or to its status as a positional
good. Those keen on retaining the ethereal quality of knowledge may
no doubt want to make a strong distinction between "genuine
knowledge" and "mere expertise" (cf. Ford & Agnew, 1992). They will
object to my apparent conflation of these two concepts. Whereas
expertise may ultimately reduce to matters of status and trust, the
objector may argue, the test of knowledge is precisely that it does
not lose its force as its availability increases. In response, let
me grant the objector's distinction as the basis for an empirical
hypothesis. If there are indeed types of "information" or "skills"
whose power does not diminish with their increased availability,
then I will gladly call them "genuine knowledge." However, my guess
is that there are limits to the optimal distribution of these
cognitive products. Consider the following homely observations:
     (a) Everyone in a town may know the location of a
     particular store and the time it opens for business.
     However, if many of these people decide to act on that
     knowledge at roughly the same time to purchase the same
     goods, then a larger percentage of them will probably
     return home empty-handed than if fewer of them knew about
     the store in the first place. Here knowledge lacked
     efficacy because the knowledgeable got in each other's
     way. (A more realistic version of this situation is one
     in which everyone decides to take the same expert's
     advice on which stock to purchase.)
     (b) It is often assumed that information freely exchanged
     among a large network of peers breeds the sort of
     critical inquiry that is necessary for genuine epistemic
     progress: the larger and freer the network, the more
     critical the inquiry. Unfortunately, this assumption
     presumes, contrary to fact, that inquirers have an
     inexhaustible ability and inclination to attend to each
     other's work. Yet, by the time the network of inquiry
     attains the dimensions of "Big Science," inquirers become
     more concerned with finding allies than opponents, and
     hence are likely to simply ignore work that cannot be
     immediately used for one's own purposes (cf. Fuller,
     1994). Here knowledge lacked efficacy because more of it
     was available than could be assimilated. 
Knowledge was undermined in (a) because too many people possessed
the same information, whereas in (b) it was because each person
possessed too much information. In neither case did these skewed
distributions actually convert a truth into a falsehood, but from
a pragmatic standpoint, they might as well have. In other words,
attention to the socially distributed character of knowledge may
help explain the intuitions that have traditionally led
philosophers to posit a conception of knowledge that "transcends"
constitutively social character expertise.   REFERENCES  
 
Abbott, A. (1988). The System of Professions. Chicago: Chicago.
Adorno, T. & Horkheimer, M. (1972). The Dialectic of Enlightenment.
     New York: Continuum.
Arkes, H. & Hammond, K., Eds. (1986). Judgment and Decision Making.
     Cambridge: Cambridge.
Beniger, J. (1986). The Control Revolution. Cambridge: Harvard.
Botwinick, A. (1990). Skepticism and Political Participation. 
     Philadelphia: Temple.
Bourdieu, P. (1984). Distinction. Cambridge: Harvard.
Cicourel, A. (1968). Cognitive Sociology. Harmondsworth: Penguin.
Collins, H. & Pinch, T. (1982). Frames of Meaning. London: 
     Routledge & Kegan Paul.
Cronin, B. & Davenport, E. (1988). Post-professionalism: 
     transforming the information heartland. London: Taylor
     Graham.
Crouch, C. (1983). Market failure. In A. Ellis & K. Kumar (Eds.),
     Dilemmas of Liberal Democracies. London: Tavistock.
De Mey, M. (1982). The Cognitive Paradigm. Dordrecht: Reidel.
De Romilly, J. (1992). The Great Sophists in Periclean Athens. 
     Oxford: Clarendon.
Dennett, D. (1987). The Intentional Stance. Cambridge: MIT.
Downes, S. (1987). A philosophical ethnography of human-computer
     interaction research. Social Epistemology, 1, 27-36.
Ezrahi, Y. (1990). The Descent of Icarus. Cambridge: Harvard.
Faust, D. (1985). The Limits of Scientific Reasoning. Minneapolis:
     Minnesota.
Fields, C. (1987). The computer as tool. Social Epistemology, 1,
     5-26.
Ford, K. & Agnew, N. (1992). Expertise: socially situated, 
     personally constructed, and "reality" relevant. Paper
     presented at the AAAI Spring Symposium on the Cognitive
     Aspects of Knowledge Acquisition. Stanford.
Fuller, S. (1986). User-friendliness: friend or foe? Logos, 7, 93-
     98.
Fuller, S. (1988). Social Epistemology. Bloomington: Indiana.
Fuller, S. (1991). Studying the proprietary grounds of knowledge.
     Journal of Social Behavior and Personality, 6 (6), 105-
     28.
Fuller, S. (1992). Knowledge as product and property. In N. Stehr
     & R. Ericson (Eds.), The Culture and Power of Knowledge
     (pp. 157-90). Berlin: Walter de Gruyter.
Fuller, S. (1993a). Philosophy of Science and Its Discontents. 2nd
     ed.  New York: Guilford. 
Fuller, S. (1993b). Philosophy, Rhetoric, and the End of Knowledge:
     The Coming of Science & Technology Studies. Madison:
     University of Wisconsin Press.
Fuller, S. (1994). The social psychology of scientific knowledge:
     another strong programme. In W. Shadish & S. Fuller
     (Eds.), Social Psychology of Science. New York: Guilford:
     162-80.
Gambetta, D., Ed. (1988). Trust. Oxford: Blackwell.
Geertz, C. (1983). Local Knowledge. New York: Basic.
Greenbaum, J. & Kyng, M., Eds. (1991). Design at Work. Hillsdale
     NJ: Lawrence Erlbaum.
Haddock, K. & Houts, A. (1992). Answers to philosophical and 
     sociological uses of psychologism in science studies. In
     R. Giere (Ed.), Cognitive Models of Science (pp. 367-99).
     Minneapolis: Minnesota.
Haugeland, J. (1984). Artificial Intelligence: The Very Idea. 
     Cambridge: MIT Press.
Held, D. (1987). Models of Democracy. Oxford: Polity.
Hirsch, F. (1977). Social Limits to Growth. London: Routledge & 
     Kegan Paul.
Knorr-Cetina, K. (1981). The Manufacture of Knowledge. Oxford: 
     Pergamon.
Minsky, M. (1986). The Society of Mind. New York: Simon & Schuster.
Pfaffenberger, B. (1990). Democratizing Information. Boston: G. K.
     Hall.
Popper, K. (1970). Normal science and its dangers. In I. Lakatos
     & A. Musgrave (Eds.), Criticism and the Growth of
     Knowledge (pp. 51-58). Cambridge: Cambridge.
Proctor, R. (1991). Value-Neutral Science? Cambridge: Harvard.
Simon, H. (1981). The Sciences of the Artificial. Cambridge: MIT.
Smith, C. (1981). The Mind of the Market. Totowa NJ: Rowman &    
     Littlefield.
Weil, V. & Snapper, J., Eds. (1990). Owning Scientific and 
     Technical Information. New Brunswick NJ: Rutgers.
Weizenbaum, J. (1976). Computer Power and Human Reason. San 
     Francisco: W. H. Freeman.
Williams, R. (1983). Keywords. Oxford: Oxford.