Some notes about transaction costs, digital libraries, peer-to-peer
computing, organized irrationality, and scientism.


In response to my arguments against automatic face recognition in
public places 
some people have objected that I am using an invalid slippery-slope
argument.  Here it is:

  (3) Even if the only people in the database today are criminals,
  the forces pushing us down a slippery slope of ever-expanding
  databases are nearly overwhelming.  Once the system is established
  and working, why don't we add alleged troublemakers who have been
  ejected from businesses in the past but have never been convicted
  of crimes?  Then we could add people with criminal records who have
  served their time, people who have been convicted of minor offenses
  such as shoplifting, [etc].  And once those people are added, it is
  then a short step to add many other categories of people as well.

One might ask why this argument is valid when I have objected to
claims that anti-spam legislation would lead to broader restrictions
on speech .  Here is
the difference.  In the case of automatic face recognition in public
places, the current debate is about whether to install a complex
infrastructure of cameras, networks, databases of facial images,
people trained to look at candidate face matches, etc.  Let's say
that we install that elaborate infrastructure for the ostensible
purpose of finding terrorists.  Then, having done all that, the
additional step of adding other faces to the database to find more
categories of people is trivial.  Taking the first step, in other
words, dramatically lowers the barriers to taking the subsequent
steps.  That's what makes it a slippery slope.  In the case of spam
legislation, by contrast, the proposed measures do not facilitate
the imagined subsequent measures in any way.  They do not make those
subsequent measures technically easier, and they do not provide courts
with legal precedents for them.  So the face recognition argument is
valid and the spam law argument is not.


Organizational boundaries and the rising tide of standards.

Here is a new way to understand transaction costs.  Coase's theory
of transaction costs, you may recall, explains a mystery: if markets
are the most efficient way to organize production, why do capitalist
economies give rise to the large command-and-control hierarchies
called corporations?  The reason, Coase argues, is that market
mechanisms are not costless.  In fact the costs of transacting
business in the market are substantial, so it is often cheaper to
organize some activities in hierarchies.  Organizational boundaries
are then determined by the balance between transaction costs and
organizing costs.  If transaction costs go down, then everything
being equal organizations break up, and if organizing costs go down,
then everything being equal organizations expand.  The problem is that
new technologies tend to affect both transaction costs and organizing
costs, so that everything is rarely equal.  As a result, the theory
is often thought to have few testable consequences, though that
hasn't stopped many theorists from deriving such consequences anyway
by making unreasonable assumptions.

The new way to understand transaction costs starts with an observation
and an analogy.  The observation is that business is fundamentally a
matter of figuring out what to standardize and what to customize.  The
two imperatives conflict, since standardization lowers costs through
economies of scale and customization makes products more appealing
to specific customers.  Because customers are diverse, all businesses
must continually search for the optimal trade-off.  Cyber gurus often
say that information technology has rendered the age of standardization
obsolete, and that new products are thoroughly customized.  But that
can't be right.  Standardization is still massively with us, and in
many areas it is even accelerating.  The cyber gurus do have a point,
though, and it is worth trying to articulate the point more precisely.
Businesses are always looking for the optimal dividing-line between
standardization and customization; that dividing-line can be drawn
in many ways, and new technologies multiply the ways that the line
can be drawn.  You can, for example, standardize the mechanisms by
which customized products are specified, manufactured, and distributed.
You can settle on three standards for three different market segments.
You can standardize some aspects of a product but not others.  You can
assemble a broad family of products from standardized modular components.
And so on.  Standardization and customization are both spreading.

That's the observation.  The analogy starts from Bill Mitchell's
argument about the geographic consequences of pervasive networking.
All of us, Mitchell says, are bound geographically to various things:
family, work, recreation, community, climate, culture, and so on.
Different people are bound to different things, and we experience
different bonds of different strengths to different things in
different places.  To reconcile these conflicting bonds we may travel
regularly, or we may move at certain points in our lives when the
balance changes among the strengths of the various bonds.  Computer
networks and the general trend toward digitization, however, loosen
some geographical bonds.  We can use cheap communication technologies
to stay in more continual contact with our families or customers,
so we have less need to be geographically close to them.  We can
obtain access to digital libraries at a distance, so we have less
need (relatively speaking) to live near a physical library.  Cheap
air travel, itself greatly facilitated by computing, also loosens
geographic bonds, though perhaps that will change now.  In any case,
as the bonds change their strengths, it follows that people and things
will rearrange themselves around the landscape.  People will reckon
the strengths of the bonds that attract them to one place or another,
and they will relocate accordingly.  The point is not that people
will become totally unbound, but that the bonds that computer networks
do not loosen will become relatively more important in determining
settlement and travel patterns.

The analogy I want to draw is between geographic and organizational
bonds.  Just as people are geographically bound to certain people,
places, and things, likewise different activities are bound together
within the same organization.  Of course, the activities could
still be conducted in separate organizations, but putting them in
the same command structure helps to coordinate them.  The greater
the need for coordination, the greater the force the binds them
into the same organization.  No force operates in isolation, though,
and every activity will be subject to numerous conflicting forces.
The boundaries of organizations will be determined by the resultant
of these forces, just as geographic settlement patterns are determined
by the forces that Mitchell describes.  As computer networks loosen
some bonds, it stands to reason that activities will group themselves
into different organizational boundaries, with some coordination
arrangements that were formerly conducted through command-and-control
happening instead through the market, and vice versa.

Let us put the observation and the analogy together into one story.
Organizations are constantly searching for the right dividing-line
between standardization and customization.  Every organization is
conducting this search simultaneously, and different organizations are
likely to come upon similar answers.  That is one reason why standards
emerge across multiple organizations and even multiple industries.
Conforming to industry standards is often valuable in itself, for
example because of the ready supply of people who can work with the
standard, and so new standards are likely to emerge from this process
and take on a life of their own.  These include product standards,
but just as importantly they include process standards, accounting
standards, evaluation and testing standards, training standards,
and so on.  Some of these standards will have official ISO standards
numbers and others will be written into law, but many will be
unofficial, even unrecognized, as employees carry solutions with them
that have worked well in other jobs.

The result of all this is a rising tide of standards, most of which
will not be visible to customers.  The great paradox of standards,
familiar to anyone who understands the Internet, is that standards and
customization are not in conflict.  To the contrary, a standard often
supplies a platform or building-block from which customized products
can be made.  So long as companies have identified accurately where
the dividing-line between standardization and customization should
fall, and so long as this dividing-line stands still long enough
for new standards to actually take hold, the rising tide of standards
will facilitate greater outward diversity even as it reduces diversity
behind the scenes.

Think of the rising tide of standards as the accretion of successive
layers.  In the old days, every organization was its own stovepipe,
cooking its own custom solution to every problem.  As standards become
established, the bottommost reaches of the stovepipe are removed
and the rest is moved onto the standards.  As the sedimentation of
standards grows deeper, the stovepipes get shorter and shorter.  This
is important because the binding force of different activities within
an organization is determined in part by the number of nonstandard
components they have.  Industry standards effectively displace the
effort of coordination, moving it from the firm level to the industry
level, or to a monopoly that supplies the standard.  As coordination
becomes easier, binding forces are reduced.  It does not follow that
organizations fragment into separate pieces that are not bound to
one another at all.  It does follow, however, that they are likely
to regroup.  Some organizations will specialize in supplying a single
standard, thus becoming a focused monopoly that provides one thin
slice through industry as a whole (see Lowell Bryan et al, Race for
the World, Harvard Business School Press, 1999).  Other organizations
will specialize in managing the boundary between standards and
customization.  Because customization is hard to manage on a large
scale, these latter organizations will be major consumers of standards
produced by others.

This explanation of transaction costs is still very abstract.
It depends heavily on an unexamined notion of the optimal boundary
between standardization and specialization -- a boundary that takes
very different forms in different industries.  It also, like most
theories of organizations, presupposes that the forces it describes
are the only forces in operation.  That will rarely be the case.
But at least it has some heuristic value.  By posing the question
of the exact nature of the trade-off between standardization and
specialization, it sets in motion in any particular case an inquiry
that in my experience is quite productive.


Digital libraries and the nature of texts.

Let me suggest an analogy.  In his analysis of social consciousness
under capitalism, Marx complained that people are led into a certain
mistake: treating commodities as if they were things unto themselves,
when in fact they are embedded in webs of relationships among people.
When commodities appear in the marketplace, they are clean and
packaged, standardized and branded; they are portrayed in dream-world
advertisements that ignore the complexities of real life.  As a
result, we tend to forget that commodities are made by particular
people in particular circumstances, and that they are shaped by
various institutional pressures.  Marx's politics were wrong, but
his descriptive analysis is often useful.  The critique of commodities
directs us to throw off the illusions of advertising and use the
tools of social science to look at the place of commodities in chains
of human relationship.  Having done so, we can see them no longer in
isolation but as pieces in a larger puzzle.

Let me bring this perspective to a particular category of things,
namely texts -- novels, news articles, scientific papers, and so
on.  I call this an "analogy" because not all texts are commodities;
scientific papers, for example, may be purchased by libraries, but
the social relations around them -- peer review, for example -- are
largely organized on non-commodity terms.  When texts are printed
and sold on paper, they certainly seem thing-like.  You can buy them,
file them, shelves them, mark them with a highlighter pen, or read
them on the subway.  They seem fully detached from the circumstances
in which they were produced.  Michael Curry has pointed out that books
in particular come with a certain subliminal promise, that you can
take them anywhere.  This is not really true, however, because every
book presupposes that its reader lives in a world that is structurally
related in way to the world of the author.  Scientific papers,
for example, presuppose not simply the mastery of a certain jargon
but participation in a certain ongoing dialogue among the field's
members, as well as an appreciation of the practicalities of doing
and reporting that particular type of science.  The scientific paper's
ideal reader, therefore, is nearly certain to be another member of the
scientific community.  The same is true in one way or another, to one
degree or another, of many if not all categories of texts.

This analysis may seem abstract, but it has concrete consequences in
people's lives.  New graduate students are making a huge transition
from one social status to another, and they are now the sorts of
people to whom scientific papers (or whatever sorts of scholarly
papers they are learning to write) are addressed.  Because of this,
graduate students need to think of the papers they read as turns in a
conversation and as moves in a complicated social world that has its
own politics and economics.  Before entering graduate school, students
are likely to read scientific papers (if at all) as the emanations
of an authority -- the blank wall that too many institutions present
to people at a distance -- and they are unlikely to think of those
papers as having been written by real people that they might expect
to meet.  Having entered graduate school, however, part of their job
is preparing themselves to build professional networks, and that means
meeting the authors of the papers they have read and cited in their
work.  This transition, from treating scientific papers as dead things
to treating them as embedded in a set of social relationships of you
yourself are also part, is very much what Marx was talking about, and
it can go wrong unless the student is provided with a decent theory of
it.  That is why I wrote "Networking on the Network".

With the advent of digital libraries, it will become more natural to
understand texts in terms of their social embedding.  This is partly
because digital libraries simplify the same uses of texts that were
always possible on paper, but it is also because the radical changes
in technology will help us to see even the old, predigital world in
a new light.  Think, for example, of the author's contact information
that is present in some kinds of texts.  Scientific papers have
long carried the author's paper mailing address, and more recently
they have begun carrying e-mail addresses and home page URL's as well.
This contact information makes it possible to "reach through" the
text to the author, whether directly by sending an e-mail query or
indirectly by making it easier to peruse the author's other papers
and research projects.  In fact, you can think of research libraries
as directories that scientists and scholars use to identify potential
professional friends.

The institutions of scientific research make it relatively natural to
drop a line to the author of a text, but other institutions may work
differently.  Every institution organizes its own set of relations
among people, and these are reflected (among other ways) in the
presence or absence of contact information and the custom of using
it.  The authors of mass-market fiction, for example, typically
build mediating structures between themselves and their readers,
and there is a paper to be written about how those structures are
evolving.  Popular authors have long had fan clubs, but now it not
unusual for them to post letters to their readers on fan Web sites.
Less established authors may respond to their readers individually.
Other texts are presented without authors' names at all, but rather
as part of a structured communication campaign by an organization,
meant either to position the organization in public consciousness
or to affect the content of public agendas and the dynamics of public
debate.  That is another set of relationships around a text, embedded
in another set of institutions.

A simplistic hope would be that digital libraries, by eliminating all
of the technological impediments that separate authors and readers,
will dissolve texts altogether so that they can commune directly.
This scenario is simplistic in part because authors couldn't possibly
commune with all their readers, nor readers with all the authors whose
works they read.  Authors in effect use texts to multiply themselves,
providing low-grade simulacra that compensate for their ability
to explain their ideas to everyone individually.  But the scenario
is also simplistic because it supposes that information technology
dissolves institutions, when in fact it usually just intensifies the
logic of the institutions that are already in place.  The institution
may end up changing, even collapsing or ceding ground to competitors,
but when that happens it a result of the institution's own dynamics
and not because the changes have been dictated directly by the
workings of the technology.  Scientists in a world of digital
libraries will still build professional networks; mass-market authors
will still address themselves to mass audiences; organizations will
still engage in strategic communication; and so on.  What if anything
will change qualitatively as a result remains to be seen.

This does not end the analysis, though.  To the contrary, it defines
an interesting space of problems.  Let us consider again the case
of the scientific paper.  Graduate students soon find themselves
participating in what David Chapman called the "secret paper-passing
network".  This is the professional network through which scientists
circulate drafts of their research papers.  By the time a research
paper appears in an archival journal, after a year or two of editorial
delays, it has long ago been read by most of its core audience
-- the scholars whose opinion the author's career most depends on.
The author's professional friends will have had a chance to offer
comments, and their names may appear in the acknowledgements section
of the finished paper.  The friends, in turn, may never see the
finished paper until they need to cite it, whereupon they turn
to the library to check that they have their quotations and page
numbers right.  In these ways, a scientific paper is already very
much embedded in the social relations of science, and plainly so
to everyone involved, and in ways that leave numerous marks on the
"thing" -- the published text -- that might end up in an outsider's
hands in the library stacks or a newcomer's hands in a first-year
seminar.  The format and conventions of the scientific paper are, as
Chuck Bazerman has shown, very much the historical product of authors'
strategies for dealing with the institutional embedding of their work.

Digital libraries will not eliminate this embedding, but perhaps the
very form of the published work will evolve as the social dynamics
around it are intensified.  The distinction between preprint servers
and online archival journals, for example, is increasingly artificial,
a product more of the outdated institutional arrangements of journal
publishing than of the practical logic of science as the scientists
experience it.  It has also long been suggested that digitally stored
papers might contain extended content such as complete data sets,
working software that readers can run on their own data, much larger
collections of images, appendices, and so on.

At a deeper level, however, the process of circulating drafts itself
changes the nature of the paper.  In many realms, such as politics,
collective writing exercises are frequently organized largely to
compel the authors to agree on what they want to say and how.  Perhaps
scientific papers, which after all are increasingly written by teams
of researchers rather than single individuals, will increasingly work
the same way.  Drafts can be circulated more easily to progressively
wider circles of readers, comments can be obtained more easily,
iterated drafts can be circulated again, and so on, with the paper's
official authors effectively turning into one circle of authorship
among many.  The authors are still claiming exclusive credit for
the work, to be sure, but they are now more openly negotiating that
credit, and their work has its effects on its readers in different,
more interactive ways.  In some cases, the collective discussion
is actually published in the pages of journals in the form of "open
peer commentary" made famous by Current Anthropology and Behavioral
and Brain Sciences.  It also happens less formally in the comments and
responses that many journals publish.  But these mechanisms, while
usefully revealing the dialogical process that normally goes on behind
the scenes, are not iterative.

A dissertation actually has something of the same character; we
make students write dissertations that will gather dust on library
shelves because we want the students to go through the transformative
experience of conducting a full-scale project, relating it to
existing work, and imposing order on the whole sprawling mess in a
form that we can judge.  Many people who have written dissertations
can testify that they are just as happy not to publish the result,
given that it was their first time through a very unfamiliar process.
The most important product of a dissertation-writing exercise is
the dissertation's author, newly minted as a scholar woven into the
community and capable of producing scholarly texts.  Dissertations
might be strengthened if there were better mechanisms, formal or
informal, by which chapters could be reviewed by successively wider
circles of the author's colleagues-to-be in the field.  One approach
would be to have students publish their "related work" chapters as
stand-alone papers in peer-reviewed online journals specially designed
for the purpose; these journals might even employ open commentary on
drafts.  The motivation to referee such papers should be great, given
the political nature and consequences of any survey article.

I dwell on research papers because they are the texts that I know
best, and the community whose infrastructure for connecting readers
and writers is most developed.  I realize that the utopia of freely
available digital libraries of all research publications is far from
inevitable, particularly the part about it being free.  The larger
theme, however, is the potential for innovation in both the form and
process of publication that comes with the technology -- the greater
ease, relatively speaking, with which the social relations around
a text can be allowed to show through.  We will still have texts in
such a world, but it will be much clearer to everyone that texts
inscribe the workings of the world around them.  And the text itself
will stop seeming like the natural unit of analysis.  Whereas a paper
book is, by its nature, a relatively fixed and detachable quantity
whose complex embedding in the larger world cannot be understood
without real thought, new electronic forms have different attributes
-- neither completely unfixed nor completely mired in the details of
relationships, to be sure, but intertwined with the social processes
around them in different, perhaps yet-unimagined ways from the
relatively simple models that have been available so far.

All of this poses certain conceptual challenges for the design of
digital libraries.  Libraries, as we all know, are not basically
about paper.  Rather, they are about managing the diversity of
documents: forming coherent collections of them, representing
them, imposing order on them, connecting people with them, computing
their emergent properties, and so on.  The diversity of documents
is important: documents exist in countless formats, structures,
languages, relationships, and so on, and they do so partly because
they are embedded in so many different institutions and forms
of social relationship.  Libraries are meta-institutions: they
support the work of institutions that work in very different ways.
What I'm suggesting is that, as digital documents evolve into more
complex embeddings in the institutions that create and use them,
digital libraries will be challenged to relate to the institutions
they support in even more complex and varied ways.  Care should be
taken to ensure that implicitly paper-centered attributes of existing
library institutions, understandable though they have been, are not
automatically inscribed into digital libraries whose potential range
of interrelations between documents and users is much wider.  Some
scholars, for example, hold that the very idea of cataloguing the
items in a collection is predicated on those items' permanence, when
the digital world is full of items, such as drafts of papers, that
are important but temporary.  If so then cataloguing will need to be
understood in a broader way.

This does not mean that cataloguing and other traditional library
practices are obsolete -- far from it.  It means that the basic
library way of looking at the world, starting from an assumption
of diversity rather than the computer scientist's typical assumption
of uniformity, will be central to the emerging digital world.
Boundaries may blur or collapse between libraries and neighboring
institutions such as publishing, collaborative authoring, enterprise
computing, peer review, records management, informal paper-passing
and commenting, data capture and archiving, tenure-and-promotion
evaluation processes, personal libraries and filing systems, online
conferencing, and so on.  Design in such a world will need to begin
with institutional analysis, and with a fully drawn understanding
of what documents can be in a world where people can readily reach
through digital representations to pursue the ends to which documents
are a means.


Parallel computing and the structure of the Internet.

Peer-to-peer computing is graduating from its ideological period
(centralization bad, decentralization good) and moving into a period
of rational system design.  You can get a snapshot of this process
by looking at the slides from Nelson Minar's talk at the last P2P

Nelson lays out a first rough taxonomy of peer-to-peer architectures
for distributed computing.  He distinguishes between (1) centralized
architectures, in which one machine maintains relations with many
other machines, none of which communicate with one another, (2) ring
architectures, in which each machine maintains relations with two of
the others, all in a row, so that the system as a whole forms a cycle,
(3) hierarchical architectures, in which the machines are arranged
in a tree structure, and (4) decentralized architectures, in which
the machines are not arranged in any definite topology and may contact
one another arbitrarily or evolve a connection topology as the task
requires.  Which of these architectures is best for a given task,
Minar argues, should be regarded as a question for technical inquiry,
not as an a priori question of ideology.  In this sense the term
"peer-to-peer" is a leftover from the ideological days, since only
two of the four architectures (ring and decentralized) treat their
constituent machines as peers.

According to this more engineering-oriented approach to peer-to-peer
computing, the purpose of a given system topology is not to avoid
seizure by the copyright police but simply to use computing resources
most efficiently.  In that sense, peer-to-peer computing is starting
to rediscover the world of research on parallel computing that has
evolved independently of the Internet.  The language of "topologies"
for parallel computing goes way back, and I am struck by the analogy
between the argument for peer-to-peer computing on the Internet
and the argument that Danny Hillis provides in the opening pages of
his book about the Connection Machine.  The Connection Machine was a
massively parallel architecture that was developed first at MIT and
later at the Thinking Machines Corporation in the 1980's and 1990's.
Hillis observed that the conventional von Neumann serial computer,
though highly evolved, nonetheless makes extremely inefficient use
of its circuitry.  A modern computer might contain literally billions
of circuits, and so it should be capable of billions of processing
operations in each clock cycle.  Unfortunately, nearly all of those
billions of circuits are memory circuits, and nearly all of the memory
circuits do nothing on a given clock cycle except perhaps refresh
themselves.  It follows that an efficient machine needs more of a
balance between memory and processing, so that a larger proportion
of the circuits can be doing useful work at any given time.

Thus the Connection Machine, which consists of many thousands or
millions of simple processing elements, all with their own local
memory and all executing the same broadcast software instructions
(single-instruction-multiple-data operation, or SIMD).  The simple
processing elements can exchange data by means of two communications
grids, a single flat plane and a general-purpose packet-switching
network.  The Connection Machine is useful for computational problems
whose inherent structure entails homogenous processing across the
entire data set and whose internal relationships match the Connection
Machine's communication topologies.  A much-reduced Thinking Machines
is part of Oracle now, and it is hard to tell whether the Connection
Machine architecture failed for technical or business reasons.  In any
case, the underlying argument is still partly valid.

The analogy between the Connection Machine and peer-to-peer computing
is this: the memory elements in a von Neumann serial machine are
analogous to personal computers sitting on people's desktops, the
centralized von Neumann processor is analogous to large Web servers,
and the communications network of the Connection Machine is analogous
to the Internet.  So the good news is that the Internet is already
the Connection Machine, without the high-speed planar communication
grid but also without the limitation of SIMD operation.  It is just a
matter of rounding up spare computational cycles and programming them.

Of course, the peer-to-peer community is not alone in viewing the
Internet as a platform for distributed computing.  The scientific
community already works with massive data sets using algorithms,
especially simulation, that involve very high levels of inherent
parallelism, and the concept of grid computing is to build virtual
machines that make the Internet look like a single expansible parallel
architecture for these sorts of advanced computations.  Despite its
undoubted importance, though, scientific computing is in one sense
the least interesting case of parallel distributed computing on the
Internet.  As I pointed out in "Computation and Human Experience",
computations are massively parallelizable to the extent that their
inherent structure maps onto the structure of the physical world
in which the computational elements are arranged, simply because
computation is faster and easier to build when the wires are short.
Our own physical world has three dimensions, and so the optimal case
is a computation whose inherent structure has three dimensions or
less.  And that is precisely the case in simulations of the physical
world, at least the ones whose causality travels at far less than the
speed of light, for which it really is too bad that the Connection
Machine's two-dimensional communications grid has no parallel on the

The hardest cases are the ones whose structure derives not from the
same physical world where the computations will be realized, but from
other worlds, such as the social world, whose structures are quite
different.  This is why research on peer-to-peer computing emphasizes
diversity and taxonomy rather than forcing maximum performance from
a relatively narrow set of computational models.  The computational
worldview will always be on the lookout for mathematically simple
structures underlying the real-world problem that peer-to-peer work
deals with, but in many cases that search will be misguided.  The
social world does have some mathematically simple structures -- for
example, when you get a large enough collection of anything, for
example all the cities in a country or all the books in a library,
things like Zipf's Law always seem to apply.  For the most part,
however, the social world simply has structures of a very different
sort than those that computational methodology is accustomed to.

In this sense, the Internet is curiously allied with its seeming
opposite, the von Neumann serial machine.  The serial machine may
be limited to executing a single instruction at a time while leaving
the vast majority of circuits spinning their wheels.  But in exchange
for this inefficiency, the programmer is freed from the structural
analysis that massive parallelism requires.  Of course, modern
compilers and processors cooperate in discovering small amounts of
parallelism, but this is all done automatically and needn't concern
the programmer.  And programmers do need to analyze the structure of
their problems in some sense, for example for purposes of abstraction
and modularity.  It is just that those sorts of analysis are more
routinely and uniformly rewarding analysis that succeeds only by
flattening a complex computation onto a three-dimensional universe.
The von Neumann processor simply punts on that problem, exchanging
a minimum of efficiency on the average problem for a maximum of
generality across all of them.

The Internet makes a similar compromise.  By providing general-purpose
switching capability, it does not force any a priori topology of
communications onto the programmer or the user community.  In exchange
for this generality, the Internet runs the risk of congestion.  That
is why the Internet is best-suited for applications that make low
demands on latency, and why controversy rages about whether to provide
quality-of-service guarantees for latency-critical applications by
overprovisioning the network (simply attaching more and more routers
to bigger and bigger pipes) or by complicating the Internet protocols
with mechanisms specially suited to guaranteed-latency communications.
Grid computing imposes an even more rigorous set of pressures on the
Internet: minimizing the latency of a large number of data streams
simultaneously, where the data streams have a definite, stable
structure as opposed to the moment-to-moment reconfigurability of
general packet switching.

As these arguments suggest, there is a difference between building
a distributed virtual computer on top of a million far-flung Internet
hosts, which is easy, and building a distributed virtual computer that
uses resources efficiently.  Some computations, such as animation, can
be decomposed into a large number of independent processes (one frame
per processor), and in those cases the simple Internet-as-computer
metaphor works very well.  But the physical world and the social world
are both highly connected places, so most computations are not like
that.  As it is, an abstraction barrier separates the generic Internet
from the diverse computational structures that are built upon it.  At
some point that abstraction barrier is going to come under pressure,
and when that happens we will have to decide whose Internet it really


Toward a global campaign against organized irrationality.

We ought to start a global campaign against organized irrationality,
such as the assaults on rational thought that are conducted by the
use of public relations methods in politics.  We would start by naming
various types of distortion -- projection, for example -- that are
pervasive in professionalized public debate.  By explaining in plain
language the methods and motives that produce those distortions, we
would help citizens to protect their minds against the irrationality
that pelts them.  The need is profound: democracy will be impossible
until civil society can delegitimate organized irrationality.  I know
that organized irrationality may seem like too overwhelming opponent.
But people have defeated other moral outrages in the past, and they
can defeat organized irrationality too.

To think about how a campaign against organized irrationality might
work, let us compare and contrast it with the human rights movement.
Precursors aside, the human rights movement as such begins with the
UN's Universal Declaration of Human Rights that Eleanor Roosevelt
negotiated fifty years ago.  Those sorts of international declarations
are often misunderstood; after all, they have little or no legal
force.  For this reason they are often called soft law, and many
people are surprised to hear that scholars, diplomats, and activists
take soft law seriously.  Soft law serves three purposes: it provides
an occasion for building global networks of proponents, it requires
the proponents to negotiate a common language, and it legitimizes
that language as the basis for subsequent debates within and
among countries.  These purposes interact, with the common language
facilitating networking and the growing legitimacy of the language
creating a greater motivation for networking.  The resulting discourse
has its effects indirectly: once the language of human rights becomes
entrenched in the public discourse of a country, so that ordinary
citizens use human rights language to reframe their own local issues
and build new structures of civil society, initiatives to give the
language legal force can get under way.  Those initiatives will be
strengthened both by the international legitimacy of the discourse and
by the assistance that human rights activists in different countries
can offer one another as the concrete political work begins.

I will compare the proposed campaign against organized irrationality
to the campaign against violations of human rights in several ways.

(1) Naming.  A phrase would be needed to provide a common banner for
movement activists in different countries to march under.  A phrase
is not enough in itself, but it can focus the project of developing
a more extensive discourse.  The ultimate goal is to entrench this
discourse in everyday public discussion, so the necessary intellectual
work must proceed on both scholarly and vernacular levels.  I suggest
the phrase "organized irrationality" partly because it resembles an
existing phrase, "organized violence".  A drawback is that "organized
irrationality" is negative, not positive; names the problem, not
the values that a solution would realize.  Unfortunately the word
"rationality" is burdened with all kinds of distracting and spurious
connotations.  The demagogues who make their living destroying public
discourse often use anti-intellectual rhetoric to mock anyone who
stands up against organized irrationality, with the result that many
ordinary people believe that only trained specialists have the tools
and authority to identify the distortions.  Trained specialists who
can't explain themselves in vernacular language don't help either.

(2) Prior identification.  Ordinary people can understand the idea of
human rights straightaway: when someone tortures you, you are quite
clear that something wrong has happened and that it shouldn't happen
again to you or anyone else.  And once the language of human rights
finds a foothold in response to those gross violations, little effort
is required to extend it to a wider range of issues, such as access
to medical care.  Achieving global consensus on that sort of extended
usage might be hard, but comprehending in general terms what it would
mean is easy.  Whereas the goal of torture is to produce a certain
mental clarity about the reality of the situation -- we can do this to
you -- the practitioners of organized irrationality have the opposite
goal -- to produce a confusion that persuades people to wander around
in a state of fragmentation or, better, to waste their days screaming
incoherent slogans ("victim!") at imaginary enemies.  People who have
been reduced to physical wrecks by torture may be afraid to stand up
against the evil, but they understand its nature.  People who have
been reduced to intellectual wrecks by organized irrationality will
have been immunized against all attempts to help them, and it would
obviously be wrong as well as futile to compel them to accept help.
Thus, whereas a campaign against violations of human rights logically
begins with the most extreme victims, who are the most motivated
to write letters and lobby international bureaucrats on behalf of
others, a campaign against organized irrationality logically begins
with the people who occupy a twilight zone: sensitized to the nature
of the evil but not yet lost to it.  Fortunately, such people will
always exist.  People will be found along a spectrum, with the fully
healthy at one end, the raving mad at the other end, and a wide range
in-between of people who employ the rhetorical devices of organized
irrationality simply because they have never heard anything else, but
who have not reorganized their personalities in the disturbed way that
the demagogues intend.

(3) Problems of authority.  Democracy requires a strong civil society,
and civil society is founded on an extensive network of individuals
who have stepped forward to articulate positions on particular issues
and mediate between ordinary citizens who agree with those positions
and the public authorities who would be responsible for implementing
them.  That much is a commonplace.  It should be recognized, however,
that authoritarianism also creates something of that same description,
identical in all formal ways to the workings of a democracy except
that the "positions" and "issues" are systematically distorted.
A democratic society is impossible unless people are immune to
organized irrationality, and that will not happen unless they can
name the distortions.  As a result, a movement for democracy is
necessarily didactic in nature, at least when starting from the depths
of organized irrationality that most "advanced" countries now suffer.
The human rights movement must also teach its language, but the
contrast between human rights and its opposite is much easier to
explain than the contrast between rational thought and irrationality.
The problem is that the demagogues of authoritarianism also pose as
teachers, and they too "help" their victims to name various phenomena
using phrases (e.g., "political correctness") that twist reality.
The work of freeing people is thus superficially similar to the work
of enslaving them, and it is hard to make the difference clear to
people whose minds have been infected by the madness.  The difference,
very importantly, is not just one between opinions, but between
irrationality and the wide range of rational opinions.  In the case
of the phrase "political correctness", for example, the irrationality
consists in part in the systematic blurring of two ideas: (a) opinions
other than the speaker's own, and (b) the practice of forcing one's
opinion on others.  In this way, heterodox opinion is portrayed as
ipso facto repressive, a reprehensible suggestion in any democracy.

(4) Cultural meaning.  New political meanings are generally made
by drawing on the reservoir of meanings that a culture inherits from
its history.  In that sense human rights means something different
in every national context, even as its values of universality stand
in tension with particularism of nationalist cultures.  Human rights
campaigners in every country must search through their respective
cultures to find language that fits the global human rights discourse
with the language of their own society.  This language may come from
religion, or from historical leaders who are remembered as especially
just.  Campaigners against organized irrationality have the same
task.  They are looking for a language not of justice but of reason,
or at least of opposition to unreason.  This is hard because of
the aforementioned strands of anti-intellectualism that afflict many
cultures; campaigners against unreason are too easily labeled as
pretentious authorities whose academic niceties are far removed from
the robust plain-spokenness of the common people.  Campaigners against
unreason probably have no alternative to positioning the demagogues as
the corrupt authorities that they are.  In practice this will usually
mean drawing upon and elaborating in a positive way the inchoate
populism that can be found in most cultures.  The idea is to offer
people an identity that extends traditional elements and draws out
the forces of sanity that can presumably always be found within them.
To be an American, for example, is surely on some level to laugh at
doubletalk.  The problem is that authoritarianism goes to enormous
lengths to colonize popular identity -- it has no alternative, given
that the beneficiaries of authoritarian culture are greatly outnumbered
by its victims.  This is why academic talk does not suffice.

(5) Law.  The goals of the human rights movement are clear: to
make violations of human rights illegal, and to make laws against
human rights violations enforceable.  The goals of the campaign
against organized irrationality are less straightforward.  Much as
organized irrationality is the enemy of democracy, making organized
irrationality illegal would be antidemocratic, not to mention
impractical.  The campaign against organized irrationality, then,
seeks to instill certain norms in each society.  It does this not
by writing new laws, but by publicizing, explaining, and entrenching
certain patterns of thought that identify and reject the twisted
rhetorical devices that characterize organized irrationality.  The
contrast with human rights, however, is not as sharp as it may appear.
Laws do not write themselves, and they do not enforce themselves.
A revolution in law such as the codification of human rights will not
occur, or at least will not have consistent practical effect, unless
the underlying principles of human rights are legitimated throughout
the society -- not unanimously, perhaps, but beyond any possibility
of overturning through an overt campaign or coup.  A social campaign,
then, whether for human rights or against organized irrationality, is
fundamentally going to be won or lost on the battlefield of people's
minds.  And it is won only when the patterns of reasoning that it
promotes are institutionalized in the society, meaning that they are
woven into taken-for-granted daily discourse, written into textbooks,
appealed to by all sides in public debates, and finally interpreted
as synonymous with national identity.  Organized irrationality, like
all social pathologies, can never be eliminated entirely.  It will
always remain latent in the culture, waiting for an opportunity to
return to the surface.  But it can be driven into its cave and kept
there so long as society continues to affirm democratic values.  The
consequences for law will surely be numerous, but literally outlawing
irrationality will not be one of them.

(6) Interests.  The impossibility of outlawing organized irrationality
has a nonobvious side-effect: campaigns against it do not provide
lawyers with ways of making a living.  Human rights campaigns are
largely staffed by lawyers, and one of their motivations is that human
rights law gives rise to controversies in which lawyers are employed.
The point is not that lawyers are entirely mercenary; people in
many professions seek ways to align their careers with their values.
The problem is simply that it is unclear who can align their careers
with the campaign against organized irrationality.  This is one
reason why such a campaign should include expanded norms of practice
for journalists and editors: confronting and rejecting irrationality
should no longer be viewed as a breach of the reporter's duty of
objectivity -- treating all "sides" to a controversy equally whether
they make sense or not -- but quite the contrary should be part of
the reporter's job as an upholder of democratic values.  Crusading
against organized irrationality -- not just random mistakes by
ordinary citizens, but the systematic practices by which legions of
professional advocates twist language and subvert reason -- should
be one way that a journalist can advance professionally.  Principled
journalists surely do not enjoy having to quote spokespeople who
dissolve serious issues into blurry associations, and they should have
legitimated grounds, and even material incentives, to refuse to do so.

(7) Asymmetries.  Human rights campaigners have the advantage that the
most serious violations of human rights occur in the least powerful
countries.  The leading industrial countries, for all their failings,
nonetheless uphold high standards of human rights domestically.
Human rights campaigns generally do not threaten them, and they are
happy to wield the language of human rights selectively in support
of their foreign policies.  While this does threaten to delegitimate
that language, in practice its cynicism is clear enough in contrast to
the principled stands of organizations that criticize all governments
equally.  In the case of organized irrationality, by contrast, the
worst offenders are found in the leading industrial countries, and
the worst offender by far, the society in which organized irrationality
has been most intensively developed and professionalized, has been
the United States.  This is, of course, not the received understanding.
Even though the term "propaganda" was once routinely used as a synonym
for public relations in the United States, Cold War propaganda stuck
the word "propaganda" exclusively on the communist governments that
the United States opposed.  While the propaganda of communism was
surely reprehensible, we should understand how ineffective it was
in comparison to the propaganda of the First World.  Vaclav Havel
wrote extensively of the emptiness of official language in communist
Czechoslovakia, where it was repeated by everyone but believed by
no one.  Just as the economic system of capitalism proved itself
much more capable of manufacturing automobiles and computers, its
superiority in the propaganda realm was just as great, and for the
same reason.  Private propaganda is of higher quality than public
propaganda, and the export of American political technologies is
one of the gravest dangers to democratic values globally -- even
as the labelling of those technologies *as* American is the most
straightforward ways of building societal immunity against them.
At the same time, the Cold War also provides considerable grounds
for optimism.  Human rights campaigners once confronted a world
power, the Soviet Union, whose hostility to human rights was vehement
and overwhelming.  Yet the Soviet Union's fell in large part because
it was delegitimated through its signing of the Helsinki Accords.
Now a global human rights campaign is accelerating in the Chinese
diaspora as well.  The evil of organized irrationality in the United
States should not be equated with the evil of the Gulag Archipelago --
confusing people is not as bad as killing them -- but the magnitude of
the challenge is comparable even so.


Beyond scientism.

Having disparaged irrationality, I also want to talk about what Hayek
and others have called scientism.  Scientism is not science; in fact
it's the opposite of science parading as science for the benefit
of people -- scientists and non-scientists alike -- who uncritically
treat science more as symbol than substance.

Here is an example.  I know someone who regards science very highly.
He holds some strong beliefs: that science is the only possible source
of knowledge; that a modern society depends on such knowledge; that
the scientific foundations of knowledge are constantly under mortal
attack by forces of irrationality that include religion, mysticism,
and bad philosophy; and that these onslaughts of irrationalism are so
powerful that science -- and thus civilization -- are in grave danger.
I was aware of these beliefs, but I hadn't realized their intensity
until one evening when I happened to mention that I found plausible
the widespread idea that a person's emotional state could have
some effect on their physical health.  When I said this, my friend
looked at me in slack-jawed amazement.  He then underwent a long
bout of stammering, starting and stopping various bits and pieces
of sentences, until at last he was finally able to explain what was
happening to him.  He told me that he regarded my suggestion about
emotional states and physical health as so bizarre that it would
be literally immoral even to discuss it, lest the bizarre belief
be given a civilization-endangering respect that it does not deserve.
He had been trying to explain this to me, except that he believed that
explaining it would be immoral, thus the stammering.  Finally, though,
he persuaded himself that the risk to civilization of explaining the
problem to me was less than the risk to civilization that I would
pose by spreading my error to others.  He was certain that I had been
joking, or at best just irresponsibly spouting off without considering
the consequences, and he tried to get me to recant.

I was amused as heck by all this, and I carefully verified that he had
understood me correctly.  I then proceeded to torment him by walking
through the argument why such a belief would be plausible.  After
all, I said, scientists regard emotions as electrical and biochemical
states in the brain, and the brain is connected to the body.  My
friend looked as though I had sworn allegiance to Satan, went through
another round of stammering, and finally explained to me that it
was not possible for mental states to influence the physical world
-- that would be magic, and magic is the opposite of science.  It
soon developed that my highly scientific friend believed in a radical
version of mind-body dualism, so radical that the mind could not have
the slightest causal dealings with the body.  I found this belief
nonsensical to the point of delusion, but he refused to discuss any
further what he regarded as a gross assault on civilization.

This really happened.

I will give you another example.  When I worked at UC San Diego,
a professor in our sociology department named Steve Shapin published
a book entitled "A Social History of Truth".  It's an account of
the social context in which science arose, and particularly the role
of trust among the aristocratic class that founded the Royal Society.
Steve was being playful with his title, which he intended as a
provocative invitation to scholars of early modern science to read his
book and consider the fine points of his argument.  Little did he know
that he would be subjected to a right-wing campaign portraying him as
an example of liberal relativism.  This campaign made an utter mockery
of his argument, ignored it altogether to be honest, and proceeded
purely from his title to spin all sorts of intellectual slander about
him.  This was shameful enough, but what was really shameful was the
willingness of some actual scientists to join in.  I had lunch with
the most vocal of these, a biologist who lectured me at tremendous
and very tedious length about the scientific method.  It didn't
seem to have occurred to him that I had attended eighth-grade science
class, not to mention sophomore experimental physics class, and was
well-informed on the subject.  Indeed it do not seem to have occured
to him to take the slightest interest in anything that I had to say.
Instead, he "knew" perfectly well what was going on: an attack on
science.  He ascribed to Shapin the kind of philosophical idealism
that some people gloss by saying that "reality is just a social
construction".  Never mind that most theories that use the phrase
"social construction" have nothing to do with idealism.  This guy did
not have the faintest idea of what Shapin had said; even though Shapin
had written an entire book about the social origins of the scientific
method, he carried on as if Shapin (in his words) "simply failed to
understand" such-and-such basic facts, namely the scientific method.
The irony of the situation was pretty serious: this scientist was
preaching to me the central importance of empirical inquiry in the
fight against dogma, while rejecting in the most dogmatic fashion
a serious attempt at  empirical inquiry into the practice of science.
He was not alone, and I got the impression that a rumor had spread
through the local scientific community that a sociologist on their
own campus was one of "them"; their stereotyped expectations were so
perfectly confirmed that empirical inquiry into the nature of their
colleague's views was not required.

I have seen this pattern on numerous occasions: when you burrow into
the logic of society's most vocal defenders of science, you routinely
encounter the most howling antiscience at its core.  Having said this,
experience shows that my life will now become much harder unless
I hasten to add that I am not myself hostile toward science; indeed,
it seems to me that the advocates of scientism are the ones who do
not believe in science.  And I have a theory about where the pattern
comes from.  New institutions rarely become established without a
fight, and the institution's participants routinely keep the fight
going, consciously or not, for years and centuries after the rest of
the world has moved on.  For example, traditional healing practices
once represented serious competition to scientific medicine, and
reasonably so, since medicine had not yet invented sterilization.
The point is not just about the comparative efficacy of the two
systems.  As traditional social systems broke down, the cultural
context of traditional healing practices broke down along with
them, and scientific medicine was better fitted to the new social
systems.  In a social sense the most basic claim of science is not
that it works better, but that its methods are public and defeasible.
If a traditional healer makes a claim, it is basically a matter of
reputation and authority, so traditional healers are regulated by the
community's long-term experience and not by its ability to evaluate
particular claims.  If a scientist makes a claim, on the other hand,
the warrant for the claim is, it is said, out in public where anyone
with sufficient training can evaluate it.  In a modern society where
knowledge-claims have consequences for the distribution of power,
this kind of public defeasibility is crucial for social institutions
to function at all.  That is why my friend was so appalled at my
seemingly unscientific views, and why he explained his objection
in political terms.  From his perspective, claims about the emotional
basis of physical health -- which he regarded as nondefeasible
by their nature -- were a short step away from social collapse and

This theory explains many otherwise strange phenomena.  Consider,
for example, the attempts by Congress and the National Institutes
of Health to establish an Office of Alternative Medicine to perform
scientific tests of various unconventional treatments that have
been offered as medical therapies.  They got a guy to run it who
was a trained scientist who had been raised in a traditional Native
American society.  This initiative seemed like perfect common sense
to me.  If you have large numbers of people believing that they can
treat illnesses by strapping magnets to their bodies, it would seem
like a valid function of government to support controlled scientific
experiments to determine whether these procedures have any medical
effect.  After all, such experiments are the most routine, most
straightforward kind of science in the world.  This view, however,
was not shared by a vocal group of scientists who were adamantly
opposed even to performing the tests.  They couched some of their
arguments in terms of objections to the methodology or credentials
of the people who were chosen to perform the experiments, but it was
clear that their objection was in principle.  They were not interested
in fine-tuning the experiments but in preventing them.  Often their
arguments were disturbingly circular: no scientific proof existed that
these therapies had any benefit, and so it would be wrong to search
for scientific proof.  By those rules, of course, no science would
ever be done.  Alternatively, they argued that the therapies should
not be tested because they had no theoretical basis -- the purest
dogma, as if theory were the test of reality and not the other way

These thoughts come to mind, as you might have guessed, in response to
the recent news report on scientists in the Netherlands (where else?)
who are seeking scientific evidence for supernatural explanations
of near-death experiences (NDE's).  Now, I think it is perfectly
proper to subject these phenomena to scientific tests.  What bothers
me is the assumption, apparently by all parties both pro and con,
that the stakes are nothing less than the foundations of science.
It seems to me that culture and science are engaged in a pointless
battle here, with scientists -- meaning, now, adherents of scientism
-- assuming that the phenomena are going to be easily be explained
away with a squirt of brain chemicals, and non-scientists -- meaning,
now, adherents of a kind of simplistic anti-science -- gunning for the
most basic foundations of the mechanistic worldview.  This drama, it
seems to me, gets in the way of any serious inquiry into the matter,
which in turn ultimately strengthens the hand of the anti-scientists
and undermines the place of science in culture.

(I'm told that you can find the original article about the NDE work
at , free registration required; look
for "near death" and it's the first article that comes up, with Pim
van Lommel as the first author.)

Let us consider another case: the phenomena of people who believe
that they can talk to the dead.  Talking to the dead is a widespread
practice, indeed nearly universal, in shamanistic cultures throughout
the world, and it is also an experience that many people in modern
societies have had.  The firefighters at the World Trade Center
routinely say that they can hear their dead buddies directing
them through the rubble.  It will not suffice to accuse these
people of pretending.  Some people do pretend about such things,
of course, just as some people pretend that they can fix plumbing.
But it strains credulity to think that cultures the world round have
built enormous ritual systems around pretending that they see images,
hear voices, engage in conversations with those voices, and so on.
People do have those experiences.  The question is what's going on
with them.  The scientistic impulse is to blow off the phenomenon
with some trite explanation: they're crazy, they're ripping
people off, they've eaten moldy grain, etc.  Followers of scientism
routinely claim to know the truth about such matters, even though
they believe that knowledge follows from experiments, and even though
no such experiments have been performed.  Scientism is not science.

I'm not going to take a position about the reality of talking to the
dead.  I do want to argue, though, that it is far from implausible
on scientific grounds that people actually do have experiences that
deserve to be called "talking to the dead".  Science often starts
from the metaphors of the age, so let us start from metaphors of
software.  Imagine that our minds are software architectures that
happen to include fairly general provisions for mobile code.  Imagine
that modules of mobile code can move easily from one person's mind
to another, transferred through a wide range of subliminal signalling
mechanisms.  (Sam Shepard's early plays are based explicitly on this
premise.)  Perhaps the modules resident in different people's minds
can even communicate with one another through the outwardly innocuous
arrangements of objects in a room.  In such a world, it is entirely
imaginable that our selves are distributed, and that each of us,
far from being confined to our own heads, is actually spread out in
several people's heads.  This idea is not far distant from what Freud
called "introjection" and it is even closer to what later Freudians
called "projective identification".  When we die, the theory suggests,
we do not instantly depart.  Rather, large parts of our psyches remain
distributed throughout the minds of the people we knew and encoded
in the physical arrangements we left behind.  Someone who talks to
the dead, on this view, is simply making contact with these remnants
of the psyche.  It's not a complicated idea.  And anyone who knows how
computers work could fill in the details.

Am I saying that this theory is true?  No, of course not.  I have
no proof.  What I'm trying to explain is the unfortunate cultural
dynamic by which this entirely plausible, entirely thinkable theory
nonetheless remains unthought -- or at least unsaid, even unhinted,
in any respectable public discussion.  Even to suggest it is to open
the doors of the most dangerous conversations in Western society.
It is immediately evident, for example, that the distributed psyche
theory could be used to explain telepathy, ghosts, and some kinds of
clairvoyance.  It could provide a scientific basis for the afterlife,
metempsychosis, and many other religious beliefs, and especially
for the ritual practices of shamans.  It is hard to know who would
find this theory more threatening, the scientists -- who would have
to admit that vast ranges of human experience lie entirely outside
the artificial boundaries that they've set for their theories -- or
the anti-scientists -- who are probably smart enough to see how many
claimed phenomena the mobile-code theory *doesn't* explain.  My real
point is that science as we know it today is too caught up in cultural
double-binds for its own good, or ours.