The last section outlined how intellectual resources are wasted today and
noted that hypertext publishing systems promise to reduce that waste significantly,
with quantitatively great benefits. This benefit involves doing what we
already do, but more efficiently. Harder to quantify are the effects of
extending the range and improving the quality of debate.
The result of arguing complex policy questions in modern media has typically
been caricature, polarization, and paralysis. Recognition of this problem
has led even thoughtful analysts to advocate simplistic policies: if ideas
will be caricatured or ignored, it makes sense to cast them as slogans from
the outset. In a sense, these questions are beyond the range of what can
practically be discussed in present media. Why is this, and what effects
might a hypertext medium have?
Breadth, complexity, and qualification
Many discussions seem to operate at or beyond practical limits related to
breadth, complexity, and the need for qualifying remarks. Some are interdisciplinary,
requiring backgrounds broader than real people have. Examples are arms control
and missile defense systems, topics that combine issues such as weapons,
sensors, space systems, orbital dynamics, software, strategy, diplomacy,
and much else. Topics of this sort are up against the breadth limit.
In addition to sheer breadth, some topics are complex, in that any non-trivial
statement about a single part may depend on its relationship to a complex
whole. Again, arms control and missile defenses are examples, with the emphasis
this time on the complex relationships among their parts. A statement on
either subject may make sense only in a certain set of scenarios (involving
assumed technologies, measures and countermeasures, relative costs, goals.
. .) but each individual scenario may easily be too complex to allow brief
description. Topics of this sort are up against the complexity limit.
Finally, some topics are highly controversial, requiring that statements
be carefully qualified. Examples, again, include arms control and missile
defenses. This time, the emphasis is on the difficulty of stating how a
particular proposal might be improved or how it might fail (subject to the
assumption that other technologies, costs, goals, etc., etc., are such-and-such)
while not being seen as taking a stand either on those assumptions or on
whether there is or isn't any sense in the general idea of controlling arms
or of shooting down missiles. Topics of this sort are up against the qualification
We face many other big, messy problems where discussion is up against one
or more of the above limits. Examples include acid rain, ozone depletion,
nuclear winter, genetic engineering, nanotechnology, economic policy, and
military strategy. Many of these issues are cross-disciplinary, involving
chemistry, physics, biology, ecology, economics, political science, and
so forth. All are complex, involving economic systems, ecosystems, multiple
technologies, international politics, and so forth. All are subjects of
contention. In all of them, an improved chance of avoiding major mistakes
could be of enormous value.
These problems involve complex tangles of facts, values, and policies. With
more accurate facts known and available, and with better means for discussing
policy choices, we could hope to find policies that would better serve our
values (such as survival). A better medium for representing and debating
these problems could help substantially.
Today, discussion of big problems tends to be simplistic, polarized, repetitive,
and ineffective. In part, this is because we lack effective ways to debate
technical issues point by point and to make the results available as factual
building blocks for further arguments. In part, this is because we lack
effective ways to represent complex problems and overlapping scenarios they
spawn. In part, this is because we lack ways of representing large contexts
to which people can easily append small, incremental insights. Hypertext
debate promises to be more detailed (hence less simplistic) and more cumulative
(hence less repetitive). This, in turn, should make it more effective and
perhaps less polarized. Thus, it should help us better discuss big problems
Basically, hypertext publishing will let people express ideas, relationships,
and criticisms more effectively. But at a higher level, this will help groups
represent whole networks of beliefs, including broad summaries rooted in
detailed evidence and argument. At a yet higher level, this will help groups
do battle over their worldviews, enabling direct point-by-point comparison.
Overall, this process of expression, transmission, and evaluation will aid
the evolution of knowledge in society by knitting minds and ideas together
There are many open questions, ranging from how best to represent links
in machines to how best to represent abstract argumentation structures on
screens. But even now, we can see how to improve on paper publishing in
crucial ways. Even now, we can foresee great benefits from systems that
seem within our grasp.
Improving problem solving
A hypertext publishing medium will have abilities beyond supporting improved
critical discussion. Since it is computer-based, it can naturally support
software for collaborative development of modeling games and simulations
 (and enable effective criticism of published
model structures and parameters). Social software could facilitate group
commitment and action: individuals could take unpublicized positions of
the form I will publicly commit to X if Y other people do so at the
same time. Once Y people take a compatible position, everyone's commitment
(to making a statement, forming a group, making a contribution, etc.) could
be automatically published. The possibilities for hypertext-based social
software seem broad.
But let us focus on the narrower value of hypertext publishing in debating
facts and policies. It will drop the costs of such operations as publishing
an idea and following a reference by factors of ten to a thousand. It will
greatly improve our ability to see what is and isn't in the literature,
and to find the best available arguments for and against a point. By speeding
work (and redirecting effort from square to round wheels), these characteristics
of hypertext publishing will make intellectual effort more efficient.
By how much? A factor of ten? Given all its advantages, perhaps so, but
this seems too much to count on. By a mere ten percent? This, too, may be
an overestimate, but for a well-established system it seems more like a
gross underestimate. Therefore let us consider what it would be worth to
improve the problem-solving ability of various research and policy communities
by this modest (and ill-defined) ten percent.
The value of solving problems
More efficient application of intellectual effort to a wider range of problems
will mean more useful knowledge. More useful knowledge will mean many things.
Knowledge generates wealth. Take our ten percent figure to mean a ten percent
faster rate of advance in technology (and in the understanding of how to
use it wisely). The value of this would rapidly mount into the billions
Better economic knowledge would shape better economic policy. Take our ten
percent figure to mean a ten percent less ridiculous economic policy (by
some measure), and again the benefits rapidly mount into the billions of
Better knowledge would improve health care. Take our ten percent figure
to measure an improvement in the rate of biomedical advance, or in the effectiveness
of applying existing knowledge to the problems of medicine. The number of
lives saved would rapidly mount into the millions.
Better knowledge can aid survival. Take our ten percent figure to mean a
ten percent lower chance of making a major mistake regarding arms control
or missile defenses. What is the value of a significantly lower chance of
In short, an estimate of the long-term value of a hypertext publishing medium
can be conservative, yet enormous. Knowledge is a primary asset of our civilization,
crucial to all our goals. Hypertext publishing promises to speed its evolution,
bringing broad and great benefits.
Why hasn't this been obvious?
Claims of wonderful benefits within reach are rightly suspect; the more
wonderful, the more suspect they they should be. The heuristic here is that
great opportunities are apt to be visible and appealing, and hence already
exploited; those that seem to be sitting around within reach are usually
illusory. It seems wise to consider how the benefits outlined here might
be real and yet underappreciated - lest we suspect that hypertext publishing
must violate some law of conservation of (social) friction, dismiss its
value, and pursue wheels with corners instead.
The chief value of hypertext publishing lies in how it can aid the evolution
of knowledge - but evolution, like spontaneous order in general, is a thoroughly
misunderstood subject. People tend to think that order must result from
orders and that progress must result from design. Schools teach evolution
(if at all) as an allegedly controversial theory in biology. They do not
speak of the evolution of economic systems (but who designed markets or
corporations?) or of technologies (but who designed the modern automobile?)
or of science (but who invented its practice and content?) or of language
(but who designed English?). Each of these achievements emerged through
many trials, many errors, and the slow accumulation of what works. This
describes evolution (including the evolutionary process we call 'design').
Ignorance of evolution makes it hard for people to see the value of improving
society's ability to express, transmit, and evaluate a myriad of small,
People share other cognitive blind spots ,
and the value of a hypertext publishing medium falls into several of them
at once. People pay more attention to the tangible than to the intangible,
and seek direct solutions to their problems more often than indirect solutions.
Accordingly, they tend to undervalue basic research (with its unknown benefits)
compared to research aimed at particular problems. Basic research is an
indirect approach to the intangible product of useful
knowledge. Hypertext publishing is likewise an indirect approach to this
intangible product, but it isn't even aimed at a specific field, like physics
or molecular biology. Its benefits, being more diffuse, seem less concrete.
People also tend to underrate the importance of media, thinking that thinking
depends just on the quality of individual minds. Further, a hypertext publishing
system (supporting an open-ended set of user interfaces and information
structures) is better seen as a medium for the evolution of media than as
a medium in its own right. Finally, people tend to think that only majorities
matter, encouraging them to neglect the value of better intellectual tools
in the hands of minorities. And as for developing a better medium
for evolving media for a minority - what good could that
be, in a world that has achieved color television?
Might it be that the goal of hypertext publishing is achievable, valuable,
and yet radically undervalued? It seems so, and that presents us with a
If hypertext publishing is promising enough, we should consider its implementation.
Questions include how this might be done, when efforts might bear fruit,
and what reasons there are for making a focused effort.
Various paths could lead to a hypertext publishing medium. Incremental paths
might start with existing hypertext systems and electronic mail; these paths
run risks of setting standards that are incompatible with long-term needs.
Another kind of incremental path would start with a design aimed at meeting
long-term needs but would first implement functions having independent,
short-term value. The resulting system will provide markets for many profitable
activities: library services; telecommunications; publishing; input and
output of paper documents; the development and sale of hardware, software,
and integrated systems for businesses and laboratories.
Discussions with researchers reveal some common confusions and misconceptions
regarding hypertext publishing. Some of these lead to underestimation of
its difficulties, others to overestimation.
Underestimation of difficulties chiefly results from failures to understand
the difference between a distributed publishing medium with two-way links
and the distribution of hypertexts joined (at best) with one-way references.
The issues raised by true hypertext publishing are explored at length in
Overestimation of difficulties stems in part from the assumption that hypertext
publishing, to succeed, must reach a large fraction of the population and
contain a corpus of knowledge on the scale of a major library. These grand
goals are inappropriate for a new medium (though one should seek system
designs that do not preclude such achievements). This paper has argued that
a hypertext publishing medium could reach the threshold for usefulness and
growth with only a small community of knowledge workers, and that it could
be of great value while used by only a minute fraction of the population.
With this realization, the fear that hypertext publishing must be an enormous,
long-term undertaking seems unmotivated. No positive arguments have been
advanced to support this fear.
Overestimation of difficulties also stems from the notion that the challenges
of hypertext publishing include all the challenges of lesser hypertext goals
- that a publishing medium would be of no value unless isolated hypertexts
had proven their competitiveness with books, magazines, movies, schools,
or whatever. This seems mistaken. Isolated hypertexts compete with authored,
organized documents; a hypertext publishing system would compete with the
disorderly tangle of material found in journals and libraries. One can imagine
that linear textbooks are always superior for organized presentations of
established knowledge, while simultaneously believing that the linked, non-linear
organization of a scientific literature would greatly benefit from computer
support. This shows the difference of the goals, and the lesser challenge
of certain aspects of hypertext publishing.
It seems that the technology is in hand to develop a prototype hypertext
publishing system, but how long will it take for such a medium to grow and
mature to the point of practical value? If one's measure is number of users
and one's standard of comparison is telephone, radio, television, or the
(failed) goals of videotex, the answer is a long time, perhaps never. It
would be hard to reach a readership as large as that of serious books, much
less that of books in general, or of newspapers and magazines.
A better measure of value is the evolution of knowledge - and knowledge,
once evolved, can have a broad impact through conventional channels. By
this measure, payback can begin while the system is small. What would it
mean to improve the effectiveness of 1,000 competent people by 10%? In the
course of a few years, it would more than repay the person-years required
to implement the system, giving a good return on the investment of intellectual
resources. The value of hypertext publishing does not depend on its becoming
a dominant medium, or even very large (though in time it may do both). Its
value and richness will grow as software evolves and the literature expands,
but its value may be substantial while the system is still small and unpolished.
Why bother trying to make this happen, rather than saying 'Let's wait and
see'? True, it might happen anyway, sooner or later - but with billions
of dollars and millions of lives seemingly at stake, a little sooner is
far better than a little later. And a blind, incremental approach could
lead to bad initial choices, setting inadequate standards and blocking progress
for a long time. It seems that we can best serve our interests by focusing
on the goal of an adequate system and trying to move toward it with all
Knowledge evolves, and media are important to the evolution of knowledge.
Hypertext publishing promises faster and less expensive means for expressing
new ideas, transmitting them to other people, and evaluating them in a social
context. Links, in particular, will enable critics to attach their remarks
to their targets, making criticism more effective by letting readers see
Hypertext publishing should bring emergent benefits in forming intellectual
communities, building consensus, and extending the range and efficiency
of intellectual effort.
ease of tracing references: machine support for link tracing means
that all references are equally easy to follow to their referent;
ease of creating new references: users can grow their own networks,
or simply annotate someone else's document with a comment (without changing
the referenced document);
information structuring: both hierarchical and non-hierarchical organizations
can be imposed on unstructured information; even multiple hierarchies can
organize the same material;
global views: browsers provide table-of-contents style views, supporting
easier restructuring of large or complex documents; global and local (node
or page) views can be mixed effectively;
customized documents: text segments may be threaded together in many
ways, allowing the same document to serve multiple functions;
modularity of information: since the same text segment can be referenced
from several places, ideas can be expressed more modularly, i.e., with less
overlap and duplication;
task stacking: the reader and writer are both supported in having
several paths of inquiry active and displayed on the screen at the same
time (this is also a feature of window systems in general). In addition,
each 'digression' occurs in a separate window, leaving the database and
display in essentially the same state until the digression is ended;
collaboration: several authors may collaborate, with the document
and comments about the document being tightly interwoven.
For full, fine-grained, filtered public hypertext systems - hypertext publishing,
in the terminology of this paper - we can add the following, emergent advantage:
speeding the evolution of knowledge in society.
Knowledge is a primary asset of our civilization, crucial to all our goals.
By improving the quality of debate and speeding the evolution of knowledge,
hypertext publishing will not only further our goals, but help us choose
them more wisely.
I have benefited greatly from discussions on these matters (some spread
over many years) with Jim Bennett, Roger Gregory, Robin
Hanson, Kirk Kelley, Mark Miller, Theodor
Nelson, Chris Peterson, Phil Salin, and Randy Trigg.
Richard, The Selfish Gene. Oxford
University Press, London (1976).  Radnitzky, Gerard and W.
W. Bartley, III, Evolutionary Epistemology, Theory of Rationality, and
the Sociology of Knowledge. Open Court, La Salle, IL (1987). Popper,
Karl R., Objective Knowledge: An Evolutionary Approach. Oxford
University Press, London (1972).  Conklin, Jeff, A Survey
of Hypertext. MCC
Technical Report No. STP-356-86, Rev. 1 (February 1987).  Trigg, Randall, A Network-based
Approach to Text Handling for the Online Scientific Community. PhD
Thesis, University of Maryland (University Microfilms #8429934) (1983). Nelson,
Theodor, Computer Lib/Dream Machines. Published by author,
available from The Distributors, 702 S. Michigan, South Bend, IN 46618 (1974). Nelson,
Theodor, Literary Machines. Published by author, version 87.1,
1987, available from Project Xanadu, 8480 Fredricksburg #138, San Antonio
TX 78229. Available on disk from Owl
International, 14218 NE 21st St., Bellevue, WA 98007 (1981). Hanson,
Hypertext Publishing: Issues and Choices in Database Design. Draft
available from the Foresight Institute,
Box 60158, Palo Alto, CA 94306 USA.  Pool, Ithiel de Sola, Technologies
of Freedom. Harvard University Press,
Cambridge, MA (1983).  Kahn, Ken, 'Language
Design for Open Systems'. In The
Ecology of Computation, Bernardo Huberman, ed., North Holland, Amsterdam
(1988).  Miller, Mark S., Daniel
G. Bobrow, Eric D. Tribble, Jacob Levy, 'Logical Secrets'. Concurrent
Prolog, Ehud Shapiro, ed., MIT
Press, Cambridge MA (1987).  Bush, Vannevar, As
We May Think. Atlantic Monthly, 176:101-108 (July 1945). Englebart,
D. C., W. K. English, A Research Center for Augmenting Human Intellect.
AFIPS Conference Proceedings, Vol. 33, The Thompson Book Company, Washington,
DC (1968).  Delisle, Norman, Neptune:
A Hypertext System for CAD Applications. CR-85-50, Tektronix Computer
Research Laboratory, Beaverton, OR (January 1986).  Delisle, Norman, Mayer
Schwartz, Contexts - A Partitioning Concept for Hypertext. Proceedings
of the MCC Conference on Computer-Supported
Cooperative Work, Austin, TX, pp.147-152 (December 1986).  Trigg, Randall, Lucy
Suchman, and Frank
Halasz, Supporting Collaboration in NoteCards. Available from
Xerox PARC, 3333 Coyote
Hill Rd., Palo Alto, CA 94304.  Lowe, David G., Co-operative
structuring of information: the representation of reasoning and debate.International
Journal of Man-Machine Studies, 23:97-111 (1985). Guide
User's Manual, Owl
International, Inc., 14218 NE 21st St., Bellevue, WA 98007 (206) 747-3203.  Hershey, William, 'ThinkTank'.
Byte, (May 1984). Yankelovich,
Meyrowitz, and Andries van Dam, Reading and Writing the Electronic
Book. IEEE Computer, pp.15-30 (October 1985).  Garrett, Nancy L., Karen
E. Smith, and Norman
Issues, Strategies, and Tactics in the Design of a Hypermedia Document System.
Proceedings of the Conference on Computer-Supported Cooperative Work, MCC
Software Technology Program, Austin, TX (December 1986).  diSessa, Andrea, Harold
Abelson, Boxer: A Reconstructible Computational Medium.Communications
of the ACM (September 1986). VanLehn,
Kurt, Theory Reform Caused by an Argumentation Tool. ISL-11,
Xerox PARC, 3333 Coyote
Hill Rd., Palo Alto, CA (July 1985).  Steele, Jr., Guy L.,
LISP: The Language. Digital Press, Bedford, MA (1984).  Malone, Thomas W., Kenneth
R. Grant, Franklyn A. Turbak, Stephen A. Brobst, and Michael D. Cohen, Intelligent
of the ACM, 30:390-402, (May 1987).  Rumelhart, David E.,
James L. McClelland, and the PDP Research Group. Parallel
Distributed Processing: Explorations in the Microstructure of Cognition,
Vol. 1. MIT Press, Cambridge
(1986).  Drexler, K. Eric, Engines
of Creation. Doubleday, New York (1986).  Drexler, K. Eric, Technologies
of Danger and Wisdom. Directions and Implications of Advanced Computing
(Symposium Proceedings) (DIAC-87), Jonathan Jacky and Douglas Schuler, eds.
Computer Professionals for Social
Responsibility, Seattle, WA, pp. 182-186 (12 July 1987).  Kelley, Kirk, Xerox
PARC, personal communication.  Nisbett, Richard, and
Lee Ross, Human Inference. Prentice
Hall, Englewood Cliffs, NJ (1980).