--------------------------------------------------------------------


                T H E  N E T W O R K  O B S E R V E R

  VOLUME 2, NUMBER 8                                   AUGUST 1995

--------------------------------------------------------------------

     "You have to organize, organize, organize, and build and
      build, and train and train, so that there is a permanent,
      vibrant structure of which people can be part."

          -- Ralph Reed, Christian Coalition

--------------------------------------------------------------------

  This month: Privacy and authoritarian culture
              The GII as a library
              Consumer guides on the Web
              The dynamics of HTML standards

--------------------------------------------------------------------

  Welcome to TNO 2(8).

  This month's issue features two articles by guest authors.  Chris
  Borgman's article is based on her powerful presentation at the
  Conference on Society and the Future of Computing.  Her thesis
  is that the Global Information Infrastructure is best understood
  as a library that raises all of the human and technical issues
  of information indexing and retrieval that librarians have been
  working with on a large scale every day forever.

  Rich Lethin, who has long filtered the cypherpunks newsgroup for
  the Red Rock Eater, has contributed an article about the BBN Auto
  Mechanics List, a voluntary cooperative venture in rating Boston
  area auto mechanics.  We all want to think that this successful
  model for mutual assistance on the net can be generalized, but
  it's not that simple.  The central question for me concerns
  critical integrity: what are the conditions that encourage people
  to try in good faith to provide honest evaluations of things?
  In a world of publicists, the answer to this question might be
  depressing.

  And I've written a brief article analyzing another argument
  against a broad right to privacy.  (I've already collected
  a batch of such arguments in TNO 1(10).)  A larger theme is
  the return of authoritarian cultural forms.  We know all about
  authoritarian government, and we don't like it very much.  But
  I fear that we've forgotten about the seductions and oppressions
  of authoritarian culture.  For a crash course in the subject,
  let me suggest Morris Shechtman's "Working Without a Net" (a
  book for people whose angry authoritarian fathers have convinced
  them that a steady diet of harsh criticism is a sign of love,
  it comes highly recommended by the Speaker of the House) and Cal
  Thomas' book "The Things That Matter Most" (with its completely
  unabashed celebration of censorship).  I'd like to suggest that
  you: (a) figure out in detail what's wrong with the arguments in
  these books; (b) figure out why decent, intelligent people might
  nonetheless regard them as necessary responses to the world as it
  is; and (c) write down what you've learned.

  About the quote from Ralph Reed that has become TNO's permanent
  motto: for the next several months I'm going to offer some brief
  commentaries on it.  I hope these commentaries don't seem too
  didactic; it's just that I really want the full meaning of Reed's
  statement to get across.  This month let us notice that he is
  talking about a "structure".  He means a membership organization.
  In his case, of course, it's the Christian Coalition, but the
  underlying principle applies widely.  He's not just talking about
  getting everyone on the net.  He's not just talking about getting
  his views out to an abstraction called "the public".  He's not
  just talking about sending out political action alerts to the
  ether and hoping that someone somewhere will act on them.  He's
  talking about building an organization.  What does that mean?
  It means having chapters and membership lists.  It means giving
  everyone a chance to discover their own strengths and passions
  and the support to enact those things within the framework of the
  organization.  It means creating a sense of belonging, productive
  activity, personal growth, successes, and shared goals.  Have
  you been involved in such an organization?  Have bad experiences
  convinced you that organizations are necessarily boring, static,
  or oppressive?  Have ever had a chance to learn the skills
  of working with others democratically within an organization?
  These questions are good starting points for defining your vision
  and deciding what you want to be remembered for when you die.

  This issue of TNO brings the demise of "Company of the Month",
  one of the original TNO departments.  It has gotten to be more
  hassle than it's worth.  Maybe it'll be back from time to time.

  A footnote.  What's wrong with American culture these days that
  the only funny comic strip in the newspaper is "Dilbert", which
  concerns the intrinsically hysterical topics of computer nerds
  and office politics?  Do you suppose that Gary Larson would
  start drawing "The Far Side" again if we organized a petition on
  the Internet?  In any event, we really must do something about
  the lame strips that have replaced him.  Maybe we can get Steve
  Bell, who draws the extremely funny comic "If" for The Guardian,
  to come to the United States.  Newt Gingrich and Bill Clinton
  have to be lots more fun to draw than John Major and Tony Blair.

  Jerry Garcia 1942-1995 RIP.

--------------------------------------------------------------------

  Privacy and authoritarian culture.

  Lately I have been encountering an insidious argument for the
  limitation of personal privacy.  Here is the only written-down
  version of it I've seen:

    Modern Americans enjoy vastly more privacy than did their
    forebears because ever and ever larger numbers of strangers
    in our lives are legitimately denied access to our personal
    affairs.  ...  Privacy, however, makes it difficult to form
    reliable opinions of one another.  Legitimately shielded
    from one another's scrutiny, we are thereby more immune
    to the routine monitoring that once formed the basis of
    our individual reputations.  Reputation ... is a necessary
    and basic component of the trust that lies at the heart
    of social order.  To establish and maintain reputations
    in the face of privacy, social mechanisms of *surveillance*
    have been elaborated and developed.  In particular, various
    forms of credentials and modern ordeals produce reputations
    that are widely accessible, impersonal, and portable from
    one location to another.  *A society of strangers is one
    of immense personal privacy.  Surveillance is the cost
    of that privacy.*  (Steven L. Nock, The Costs of Privacy:
    Surveillance and Reputation in America, New York: Aldine
    de Gruyter, 1993; page 1.)

  So: privacy is a threat to social order; it must therefore be
  constrained or restricted; external surveillance serves this
  purpose; and the result of surveillance is a kind of objective
  publicity that restores social order.  (Note that this is
  a stronger form of another common argument, that pervasive
  surveillance effectively restores industrial society to the
  condition of the agrarian village, where the social order was
  maintained through everyone knowing everyone else's business.)

  This argument contains numerous fallacies; let me just identify a
  few of them.  The first fallacy is the confusion between privacy
  and secrecy: if people have lots of privacy, the argument goes,
  then nobody will know anything about anybody else.  But privacy
  is not the same as secrecy; instead, I have privacy when I
  control which matters are secret and which are disclosed, and
  when, and how, and to whom.  People may have total privacy and
  still choose to tell everyone everything.  In practice, privacy
  permits people to disclose the things they wish to disclose.

  The second fallacy is encapsulated in the phrase "one another",
  which posits a symmetry and equality of individuals that does
  not exist.  The most serious issues of privacy in modern society
  do not concern private individuals' dealings with one another
  on an equal footing; those are regulated reasonably well through
  individuals' right to disclose or conceal what they wish,
  together with their right to choose whom they have dealings with.
  Privacy problems arise, instead, in situations of gross asymmetry
  or inequality in power relations.  We don't worry terribly about
  whether I must disclose my marital troubles to my neighbor, but
  we do worry about whether I must disclose my marital troubles to
  the government.

  These fallacies combine to produce some serious and dangerous
  conclusions: if individuals' ability to conceal and refusal to
  disclose certain information about themselves is construed as
  a threat to social order, then it follows that people must be
  compelled to disclose this information.  And if asymmetries and
  inequalities within society are neglected, then surveillance --
  the systematic coercion of disclosure that powerful institutions
  exercise against individuals -- is legitimated and even morally
  required.

  Lurking within this argument are several subsidiary fallacies.
  One of them is hidden in the term "reputation", which presupposes
  a particular model of information: namely, that you are only able
  to develop trust in me by gaining access to information about
  me that is public -- i.e., accessible to everyone.  If relations
  of trust are held crucial to social order, and if relations of
  trust are held to require access to publicly available personal
  information, then it follows that society must compel public
  disclosure of personal information -- not just disclosure to
  particular parties, but *public* disclosure.  But the second
  step of this argument is clearly false: for you to trust me, you
  don't need *everyone* to know anything about me; you simply need
  to know it yourself.  (And even *that* isn't clear.)

  I could go on, but I won't.  My basic point is that arguments
  about privacy frequently encode, through their conflations and
  omissions and ambiguities, an authoritarian model of culture in
  which people must be actively controlled by outside institutions
  in order for society to hold together.  I think that libertarian
  conservative arguments about the evils of government, whatever
  their merits, have helped us to forget -- or, at least, are
  not helping us to remember -- what an authoritarian culture is
  like.  It's a culture in which most people have been convinced
  that everyone else must be monitored, regulated, and shamed to
  maintain social order.  Let's learn to recognize authoritarian
  cultural forms, because they're coming back.

--------------------------------------------------------------------

  The global information infrastructure as a digital library.

  Christine L. Borgman
  Department of Library and Information Science
  University of California, Los Angeles
  cborgman@ucla.edu

  The National and Global Information Infrastructures (NII and
  GII) offer the promise of creating a global digital library
  in which anyone, connected anywhere on the network, can search
  for information independent of time, place, or form.  Public
  discussions of the "information superhighway" suggest that the
  global digital library nearly exists already, or that it soon
  will be accomplished.  Even technical and policy documents
  suggest that we are close to achieving universal access to
  information resources, such that anyone can find what they want
  or need in the glut of information that exists already.  While
  such claims attract public support for building the computing
  network, research funding for those proposing technical
  solutions, and customers for computing network services, they
  obscure the complexity of the information retrieval problem.
  They also obscure the role of libraries in providing access to
  distributed information resources.  This short paper summarizes
  four issues that need to be addressed if the GII is to serve as
  the Digital Library of the Future.  We discuss these issues in
  more depth elsewhere (1).

  1.  The Global Information Infrastructure should be viewed as a
  single Digital Library with access for all.

  In the ideal case, the GII will be a decentralized, distributed
  "virtual" library that interconnects all the databases and
  other resources on the Internet and subsequent computing and
  communications networks.  The international library community
  already has created an institutional framework for such a
  system through shared cataloging databases, interlibrary loan
  agreements, document delivery services, and other forms of
  access to information resources held elsewhere.  By utilizing
  the technical framework of the GII and the institutional
  framework of library cooperation, it should be possible to
  search the GII as a single Digital Library to identify, locate,
  and obtain information resources, no matter where or in what
  form they exist.  In theory, the global digital library could
  increase international equity in access to information and offer
  the freedom to read, a privilege often denied within individual
  countries.  However, freedom of information is not one of the
  basic tenets of the GII policy proposals, despite the efforts of
  various human rights groups.

  2.  The Digital Library should provide pointers to information
  resources that exist in all media, whether online or offline.

  Discussions of the Global Information Infrastructure and Digital
  Libraries often implicitly assume that digital libraries consist
  entirely of digitized content and that the full content of all
  information resources soon will be online.  The value of online
  catalogs and indexes that point to offline materials receives
  little recognition outside the library community, which we
  attribute to misconceptions about the nature of communication
  technologies, information resources, and information
  organization.  First is the misconception that digitized
  information will supplant, rather than supplement, information
  resources in other forms.  New technologies create and fill new
  niches, while prior technologies often continue to evolve and
  fill other niches, as the history of communication has shown.
  The centuries of human knowledge that are stored in non-digitized
  formats will continue to be valuable, and only a very small
  portion of these resources are likely to be converted to digital
  form.  Paper and other durable hard copy formats will complement
  digitized formats.

  Second is the misconception that the information that exists
  on the Internet is an adequate substitute for the holdings
  and services of libraries.  The volume of information on the
  Internet pales in comparison to the holdings of the world's
  research libraries, most of which has been carefully selected.
  Very little of the "free" information on the Internet has
  passed through an authoritative review process -- much of it
  is self-published or otherwise ephemeral in nature.  The reader
  or user of such information bears the burden of determining
  what are accurate or credible sources, lacking the imprimatur
  of reviewers, editors, and publishers, or the judgement of the
  librarians who select the materials.

  Third is the misconception that catalogs of information resources
  lack value unless the full content exists online.  In searching
  for information, one must first identify the existence of
  information resources and their location before they can be
  obtained, whether online or offline.  The greatest value of
  the GII as a Digital Library will be to provide pointers --
  catalogs, indexes, abstracts, document surrogates, and other
  representations of content -- not only to online information
  resources but to the centuries of information resources that will
  continue to exist only offline.

  3.  Implementing the social policy to create the Digital Library
  will be even more difficult than implementing the technology
  policy.

  Information technology will enable the global digital library --
  it will not create it, or necessarily even promote it.  A single
  Digital Library with access for all will be realized only through
  the efforts of individual countries, institutions, and people.
  The slogan "think globally, act locally" applies to the Global
  Information Infrastructure as much as it does to the environment.

  The open systems and interoperability principles stated in
  the NII and GII proposals are necessities for a global digital
  library that provides access to information resources in all
  formats, in all languages, and on systems operating on all
  technical platforms.  Setting, promoting, implementing, and
  enforcing interoperability standards is very difficult even
  within one country.  The Internet achieved interoperability
  through cooperation among the government, education, and
  non-profit sectors; these sectors now must co-exist with
  competitive commercial ventures.  Creating interoperable systems
  between countries with technology policies that rest on different
  political, economic, social, and cultural traditions is even more
  difficult.  Telecommunications connectivity is far lower in most
  parts of the world than in the United States, policies for access
  and usage vary widely, and the range of hardware and software
  platforms varies even more.  The principles of open, unmediated
  access to information and the freedom to read that Americans
  take for granted do not apply in all countries that are connected
  to the Internet.  We must account for conflicting standards and
  policies in creating a global network, for what works in the
  United States does not necessarily work elsewhere.

  4.  Information retrieval is a hard problem.

  The paradox of information retrieval is that a person
  must describe the information that he or she does not have.
  Claims that we are close to solving this paradox rest on two
  misconceptions: an incomplete understanding of the information
  retrieval process, and the scaling problem.

  a.  The information retrieval process.

  Information retrieval rarely is a single act of formulating a
  query; rather, it usually is a process that begins with some
  vaguely-felt need of wanting to know something and gradually
  evolves to the point where one can describe some attributes.
  Once the need can be phrased sufficiently to begin searching, the
  question itself may change through multiple iterations of finding
  and using information resources.  Thus people usually approach an
  information retrieval system with a partially-formed query to be
  negotiated.

  When searching for information, a person is seeking knowledge
  or meaning (e.g., what? why? how?) but must formulate a query
  in terms of the content (e.g., words, numbers, symbols) of
  extant information entities (e.g., documents, objects).  As
  an information retrieval system, the GII can deal directly with
  information only as entities with content; meaning must be left
  to the interpretation of each searcher.

  Historically, information retrieval research has focused on the
  most easily computable aspects of the process -- starting with a
  well-formed query and matching that query against the content of
  information entities -- ignoring the information-seeking process
  and the context in which the question is asked.  Information
  retrieval systems are effective only to the extent that they
  can assist in answering question, rather than the extent to
  which they can match queries.  Query matching is a process that
  intelligent agents can accomplish; true information retrieval
  is not.  Query-matching systems were designed for highly skilled
  searchers, usually librarians -- the original intelligent agents.
  In contrast, the global digital library must serve a population
  of information seekers that is heterogeneous in terms of age,
  language, culture, subject expertise, and computing expertise,
  most of whom will be perpetual novices at information retrieval.
  The easy part of the retrieval process may be nearly solved; we
  have barely begun research on the hard part.

  b. The scaling problem in information retrieval.

  The ease of finding information is a function of heterogeneity
  and size of the database, as well as the ability to articulate
  the question in searchable terms.  Finding information is
  simplest in small databases with homogeneous content because
  the meaning of symbols (terms, images, etc.) is constrained
  and the amount of "noise" in retrieval is tolerable.  As the
  heterogeneity of the database(s) searched increases, the variety
  of ways in which each concept might be described increases, the
  variety of meanings for each symbol increases, and the number
  of irrelevant matches (noise) increases.  The global digital
  library must support searching of information resources in
  multiple languages, multiple character sets, and multiple media,
  not just mono-lingual text, further increasing the complexity of
  the searching process.  The keyword full-text search tools now
  appearing on the Internet are being applied to relatively small
  databases, by library standards, and already are encountering
  all the content control problems well known to librarians
  -- variant word endings (e.g., index, indexes, indexing),
  indefinite references (e.g., it, that, which), synonyms (e.g.,
  heat, thermal), homonyms (e.g., Paris, France; plaster of Paris),
  indirect references (e.g., "the matter we discussed yesterday"),
  concepts for which no explicit term appears in the document
  (e.g., history, democracy, social effects, strategy, statistics),
  and difficulties in determining the relative emphasis on each
  concept.  The old programming slogan GIGO applies -- "garbage
  in, garbage out." Information either can be organized as it is
  entered into the system to simplify later retrieval, or it can be
  organized on the way out -- leaving to the searcher the burden of
  sorting through masses of irrelevant information.

  Conclusions

  The technology and policy of the Global Information
  Infrastructure offers unprecedented opportunities -- and
  challenges -- for creating the Digital Library of the future.
  If we view the GII as a single global digital library, it should
  be possible to identify, locate, and obtain information resources
  no matter where or in what form they exist, online or offline.
  To accomplish this goal, we must tackle the fundamental paradox
  of information retrieval -- describing the information that
  the information seeker does not have -- by assisting the user
  in articulating the question.  We have made a start on these
  questions in small and homogeneous databases with skilled
  searchers, but now must address them in the context of
  information resources and user populations that are very large
  and heterogeneous.  The technical problems may be easier to
  solve than the social problems, given the vast range of economic,
  cultural, and linguistic boundaries crossed by the Global
  Information Infrastructure.

  Note

  1.  A series of papers and a book are forthcoming from this
  research.  The following papers are in print or in press as of
  this writing:

  Borgman, C.L.  International issues in access to information,
  or Can the Internet bring democracy to closed societies with few
  telephones or computers?.  Proceedings of the Computers, Freedom,
  & Privacy Conference, March, 1995, Burlingame, CA. pp. 66-70.
  New York: Association for Computing Machinery.

  Borgman, C.L.  (in press).  Information Retrieval Or Information
  Morass?  Implications Of Library Automation And Computing
  Networks In Central And Eastern Europe For The Creation Of A
  Global Information Infrastructure.  Proceedings of the Annual
  Meeting of the American Society for Information Science, Chicago,
  October 9-12, 1995.  Medford, NJ: Learned Information.

  Borgman, C.L.  (in press). Will the Global Information
  Infrastructure be the Library of the Future?  Central and
  Eastern Europe as a Case Example.  61st International Federation
  of Library Associations General Conference, Istanbul, Turkey,
  20-26 August, 1995: Libraries of the Future. The Hague,
  Netherlands: International Federation of Library Associations
  and Institutions, POB 95312, 2059 CH.  IFLA.HQ@IFLA.NL

--------------------------------------------------------------------

  Empowering the consumer: The BBN Auto Mechanics List.

  Rich Lethin
  lethin@ai.mit.edu

  Do you worry about marketing wizards using their databases
  of intimate personal information to manipulate you?  Maybe
  prices will change in the supermarket aisles as you walk by;
  maybe highway billboards will customize a tailored pitch as you
  round a turn.  If you find these possibilities spooky, there's
  a technological way to fight back: developing databases to
  help consumers make informed choices.  The general concept is
  illustrated by the BBN Auto Mechanics List.

  The list is a few pages on the world wide web.  It rates auto
  repair and body shops in the Boston area based on feedback that
  users email to John Bowe, who maintains the page.  Each shop has
  an entry with a grade, from A+ to F, and a few tersely edited
  comments describing the service and satisfaction that the user
  got at the shop.  Here are two of the 160 entries:

    [A] : ABJ FOREIGN MOTORS
    91 Marshall St, Somerville - 625-6632
    1/95: 1988 Subaru GL wagon. Sears said CV joints needed.
      ABJ said no, just protective booties, saving lots of
      money. Friendly and busy, yet quick.
    8/91: friendly, honest, know their stuff. Had to keep the
      pressure on them to get [the work] done by the end of the
      day. Still.. [an] "A".
    12/90: The guys there are smart, reliable, honest, excellent
      mechanics. What more can one ask?

    [D-] : CENTURY TIRE
    Beacon St, Cambridge
    2/95: "Rude and sleazy". Put on cheaper tires than were
      paid for. Hesitant to credit visa for difference, tried
      to push overpriced accessories instead.
    2/95: "Classic bait-and-switch game". Quoted good price for
      what he wanted over phone (Nokia Hakipolita Snows for BMW),
      but pushed junk on him in person. Slow service.

  I was familiar with these shops, and the descriptions matched
  my experiences.  So this past winter, for repair of damage from
  a small accident, I consulted the mechanics list for a good auto
  body shop.  I chose this one:

    [A] : MIKE'S AUTO BODY
    Malden
    4/95: Rave review, and a better than expected price (and
      below Dick's). Happy to drop customer at the T. Even cleaned
      road grime. Mike's is mildly associated with a local Porsche
      car owner club, so Porsches are a specialty.
    8/94: Suspension and related work on Porsche 911 Turbo.
      Excellent work. (The Porsche folks in Germany would be proud.)
      Genuinely interested in satisfying customer.
    8/91: Very good quality, and price is right. Priced job WAY
       below Dick's.

  and saved about $300 versus the competitor's quote.  The 4/95
  entry is mine.

  How much impact is the list having on Boston repair shops?  John
  helpfully supplied me with the list's access log, to allow a some
  guessing/estimating.  Last month, 1300 unique machines accessed
  the page, up from 550 the month before.  Some of those were from
  foreign countries and should be disregarded; to quickly estimate
  the proportion of accesses that were local, I looked at the 1000
  easily-classified accesses from the EDU domains.  Half were from
  local schools (most from MIT).  So, roughly at most 650 owners
  used the list to help choose a repair shop last month.

  The list includes 160 shops. If each shop services 20 customers
  a day, then these shops serviced 96000 customers in a month.
  If each of the 650 accesses to the list resulted in a customer
  choosing one of the shops on the list, then less than 1% of all
  customers were list-informed.  39 of the shops on the list got
  an A, if each of the 650 browsers went to one of those A shops,
  those shops' business increased by 2% during the month.  So, the
  impact on the shops is pretty small now, though at least one is
  aware of the list.

  But users who do use the list can save a bunch of money.

  John told me that maintains the list because he had found it
  useful a few times when he worked at BBN; he sees it as a way
  to contribute back to the net.  It doesn't take much of his time
  (he only receives about 4 or 5 messages a week), and he doesn't
  have any plans to expand it or form a company around it.  He's
  busy with his real job.

  The mechanics list resembles a restaurant guide book which
  gathers its data from cards mailed by selected diners.  What
  distinguishes it, and how has technology enabled it?  One key is
  that the net has reduced the cost and work of distributing and
  collating the data.  In contrast to restaurant guides, there was
  no established market for mechanic lists, so a publisher would
  take a risk investing in it.  Sourcing the page on the web is
  effectively free, though: email is free and the incoming text
  is handy for incorporation into the web page.  The electronic
  distribution also changes the character of the list.  The
  mechanics list gets updated regularly, which improves information
  quality (users can give quick pointers to inaccuracy) and also
  makes it fun to send in information and see it incorporated.

  Some characteristics specific to the Boston auto repair market
  probably help make the list work.  Boston seems to be the right
  size: big enough to make finding a good mechanic a challenge, but
  small enough that the number of shops on the list is manageable.
  Lots of people are on-line in Boston, and they're a relatively
  homogeneous bunch of students, engineers and computer scientists,
  so their expectations and experiences of auto repair are likely
  to correlate.  The student population is particularly transient
  and thus unfamiliar with the Boston mechanics.

  Can the service work with more Bostonians on-line?  Would it
  work in other markets?  Maintaining the list will become more
  than the small distraction it is for John Bowe right now.  Maybe
  volunteers will help or perhaps the list would be supported
  by donations from happy users.  Or, perhaps this process could
  be automated to scale into a general net-based consumer voting
  scheme over companies, products, and manufacturers.

  There ARE problems with scaling.  Beyond mentioning problems
  of "efficiency", such as dislocation and obsolescence of real
  people, I'll neglect the consequences and focus on the problems
  in its workings.  There's lots of room for esoteric economic
  models and vigorous hand-waving here, so I'll use the BBN Auto
  Mechanics List to try to ground my comments.

  The list is vulnerable to abuse.  There's nothing to prevent
  a garage from contributing bogus raves about itself, slamming
  competitors, or hiring an advertising agency to do this for them.
  John hasn't noticed any bogus reviews coming in and the content
  seems accurate so things seem to be working now -- probably
  because of the list's obscurity.  But I'm skeptical that it
  can continue.  In other net forums, such as those discussing
  new musical groups, people are being paid to hype specific
  artists.  Investment newsgroups have had shills promoting penny
  stocks. Similar things could happen to the BBN list - though the
  relative permanence of mechanics (compared to the musical scene
  or stock markets) makes manipulation a bit more difficult.

  Is libel a problem?  It looks like the list owner is protected
  now, with his disclaimer about passing the information on with
  no guarantees about accuracy.  However, the recent Prodigy case
  provides a precedent for considering small bits of editing to
  confer responsibility.

  Scaling leads to potential for inadvertent degeneracy.  It
  might avalanche toward extremes with bad reviews influencing the
  objectivity of later reviews, or move to irrelevance with a large
  variance in perceptions decaying most of the shops' grades toward
  "average".  There are many, many other ways for information
  exchanges of this sort to fail.  (One aspect of the list that
  fights these trends is that the ratings of garages are not
  limited to a single grade.  The well-edited descriptions help
  readers make their choices in a more informed manner.  The
  mechanics list occupies a nice position in the representational
  spectrum, with letter grades available but more descriptive data
  also available.)  Can systems methodologies for gathering this
  information be designed which fight degeneration?

  Why scale up services like this?  Recently (6/30/95) the New York
  Times profiled Providian Bancorp, which provides credit cards to
  consumers.  Providian mails credit solicitations with unspecified
  interest rates.  If a consumer responds, Providian can access
  their credit history and use statistical techniques to tailor
  the highest possible interest.  The techniques might notice that
  the consumer has been insensitive to interest rate in carrying
  a large balance.  In this negotiation, the consumer is being
  put at a disadvantage by the records kept of his past behavior.
  However, if the consumer could access a database of credit
  card companies, interest rates, and background on the consumer-
  unfriendly practices of Providian, they'd be comparably leveraged
  in the negotiation, and would probably get a better rate.
  Informed consumers can make better decisions, and this principle
  works for many other misleadingly advertised products.

  How can automated tools be structured?  Are there systems and
  algorithms that can be developed and deployed to increase the
  quality of this type of information, and to protect against
  abuse?  Maybe.  Game theory might be employed to design
  mechanisms for voting schemes that are robust against shills
  using authentication schemes such as digital signatures.  Privacy
  of respondents needs to be protected, and the technologies for
  anonymity on the net that have recently been developed seem a
  good starting point.  Representational schemes for agents are
  in development; these might be used as a language of reputation.
  Finally, learning models and prediction mechanisms such as
  automatic collaborative filtering might be used to better-tailor
  preferences for the consumer.  It's a complicated system and it
  would be a cool experiment.  Let me know what you think.

  However, it's probably not necessary to wait for the deployment
  an automated huge system right now to have an impact.  The BBN
  Auto Mechanics List demonstrates that a small effort can do a lot
  of good.

  --

  The BBN Auto Mechanics List is at
  http://web1.osf.org:8001/faq/bbn-auto-mech.html

--------------------------------------------------------------------

  Wish list.

  Strange and instructive things are happening these days in the
  world of the WorldWide Web.  Take Netscape.  Having given away
  eleven bazillion copies of its version 1.0 web client and stocked
  its coffers with IPO cash that it didn't really need, it's time
  for the Netscape company to start making money.  But why should
  Netscape make any money?  After all, the client they've given
  away for free is perfectly good.  The answer is that the client
  they've given away only works with a certain set of features.
  Let's focus on HTML.  Everyone knows that HTML is not a great
  programming language, particularly if you want to create complex
  things like tables.  So HTML is going to get some more features.
  And as the creators of web pages start using those features, the
  old web clients will slowly stop working.

  Now, this process could happen in one of two ways: (1) the W3C
  (the WorldWide Web Consortium) could define some new standards
  for HTML and everyone could then go out and support them; or (2)
  companies like Netscape could start defining their own features
  without any regard for the W3C process.  Of course, these two
  things will both happen, and they will probably interact with
  one another as well.  The dynamics of the process will be shaped
  by the interesting special properties of HTML and the Web, but
  they will also exemplify the larger market dynamics of technical
  standards.  The most interesting of HTML's special properties
  is that the source code for Web pages is public.  If you like the
  look and feel of someone's Web page, just pull down a menu, grab
  their source code, and modify it to insert your own content and
  produce the look and feel you want for yourself.  Copy-and-modify
  programming is important throughout the computer world, but the
  Web has taken it to new heights.

  But here's the catch: if the page you copied uses non-standard,
  Netscape-specific features, and if you use Netscape yourself,
  you're unlikely to find out about the problem until much later.
  If you really like the non-standard feature then you may not even
  care about the problem, figuring that most people use Netscape
  anyway and other browsers probably won't crash too badly.  HTML
  features and programming cliches can travel like viruses through
  this dynamic, copied from one neat page to another.  It's a good
  dynamic in many ways.  It lets people get up to speed in HTML
  programming very fast, since they never have to start writing
  code from a blank screen.  On the other hand, it might cause
  some havoc for the process of defining standards.  Standards are
  good because, among other things, they keep people from being
  locked in to a particular supplier's products.  Suppose that a
  Netscape-specific dialect of HTML somehow arose and became widely
  used.  And let's say that other Web browser companies develop
  their own distinct dialects.  Then the Web will slowly break
  into separate regions, each with its own dialect of the language.
  Someone who wanted to use a certain subspace of Web pages would
  have to acquire the Web client that can read those pages.  The
  result would be a fragmented market, with each supplier receiving
  a high margin but with the total market greatly depressed because
  the benefits to buyers of entering the market would be much
  less.  Some argue that companies have an incentive to encourage
  this situation; if Web features are completely standardized then
  barriers to entry in the market for Web clients will be low, and
  profits will be low accordingly.

  What can we do to prevent such a situation?  I would suggest that
  the W3C, or some other friend of standards, produce a Web crawler
  that checks pages for compliance with the standards.  The pages
  are all public and various search tools already sweep over them
  on a regular basis, so nobody should mind a standards-checking
  tool doing the same thing.  Every site and every user could set
  a switch indicating whether they wish to receive an automated
  commentary on their HTML style.  (The switch would be set to "no"
  by default.)  Also, it would be possible to ask for a commentary
  on a page right away by feeding a URL to a site which has an HTML
  commentary demon running full-time.  The commentary might include
  things like "warning: <SO-AND-SO> is a Netscape-specific feature
  and not part of the standard" or "please note: <SUCH-AND-SUCH>
  is part of the draft next version of the standard, but is not yet
  officially a standard" and so on.

  Such an HTML style crawler might have a variety of uses.  For
  example, it could track the spread of new features, producing
  statistics that would be available on Web pages for anyone to
  read.  Some difficult design decisions would also be necessary.
  For example, does one crawl the Web from the same starting-points
  as the standard Web search tools, or from a different set of
  points?  Does one crawl only those pages that can be reached
  through links in pages already crawled over, or does one also
  thread the search through other pages in the same directory as
  pages that are crawled over?  Could Web client producers create
  their own crawlers that search for features that are specific to
  their competitors' clients, and then send direct mail to those
  pages' authors suggesting features that are more compatible with
  the standard?

--------------------------------------------------------------------

  This month's recommendations.

  H. Landis Gabel, ed, Product Standardization and Competitive
  Strategy, Amsterdam: North-Holland, 1987.  The computer industry
  doesn't come close to obeying the laws of supply and demand from
  neoclassical economics.  Why?  Because issues of compatibility
  generate all kinds of strange incentives and strategies.  This
  book contains the best account I've seen of these phenomena.
  The first chapter, by Joseph Farrell and Garth Saloner, could
  have been the textbook for Bill Gates' march to eleven-figure
  wealth -- not by making better products, but getting software to
  market quickly and then leveraging various compatibility effects
  to consolidate and expand his market dominance.  We'll see a
  lot more of this sort of thing, and I think it's important for
  everyone to learn more about it.

--------------------------------------------------------------------

  Follow-up.

  Robert Putnam <rputnam@cfia.harvard.edu>, whose article about the
  decline of associational ties in the United States I discussed in
  TNO 2(3), wrote to register some disagreements and clarifications.
  First of all, he was a little disgruntled that I referred to the
  Journal of Democracy as "generally conservative", lest anyone
  think that *he* is conservative, which he is not.  He accurately
  points out that an interest in associational ties is a bipartisan
  matter at the moment.  We went back and forth about his concept
  of "social capital", which I had asserted isn't really a kind of
  capital.  The issue is complicated, because a narrow construal
  of the term -- roughly, the stock of relationships of trust that
  I have built up -- is indeed a kind of capital, and by summing
  up the relationship capital possessed by individuals in a region
  we can come to something that deserves to be called "social
  capital" and ascribed to a region rather than to an individual
  or organization.  But his concept goes beyond this to speak
  of a general climate of trust and a general assumption of
  trustworthiness (something that can be found in northern Italy,
  as he argues in his book "Making Democracy Work", but not in
  southern Italy), and this is the part that doesn't seem like
  capital to me.  He also expressed skepticism that computer
  networks can support "communities" of the sort required by
  his argument -- networks of trusting relationships that form
  the basis of a healthy and vigorous civic and economic life.
  I don't much care about this myself, though, since I'm much
  more interested in viewing computer networks as integrated
  with regional and professional communities, supporting their
  existing dynamics and perhaps helping to change those dynamics.
  Although I probably didn't make this clear in TNO 2(3), see my
  longer discussion of the matter in, for example TNO 1(5).

  Web picks:

  The Internet Engineering Task Force, which sets standards for the
  Internet, is on the Web at:  http://www.ietf.cnri.reston.va.us

  The Institute for the Study of Civic Values has a good collection
  of resources on community-building at
  http://libertynet.org/~edcivic/iscvhome.html

  The Londoners who are defending themselves against charges
  of having libeled McDonalds by handing out a leaflet that
  was critical of the company's environmental practices and
  the nutritional value of its food have a web page with their
  original leaflet and other information on the case.  The URL
  is  http://anthfirst.san.ed.ac.uk/McLibelTopPage.html

  Negativland's intellectual property web site is at
  http://sunsite.unc.edu/wxyc/legal.html

--------------------------------------------------------------------
  Phil Agre, editor                                pagre@ucla.edu
  Department of Information Studies
  University of California, Los Angeles         +1 (310) 825-7154
  Los Angeles, California  90095-1520                FAX 206-4460
  USA
--------------------------------------------------------------------
  Copyright 1995 by the editor.  You may forward this issue of The
  Network Observer electronically to anyone for any non-commercial
  purpose.  Comments and suggestions are always appreciated.
--------------------------------------------------------------------

Go back to the top of the file