The proposition that documents can be viewed as a negotiation may be perceived by many academics in the social sciences as unoriginal. Indeed, it is unoriginal; I first encountered the idea in John Seely Brown and Paul Duguids article, The Social Life of Documents  and I suspect that this sort of concept has been explored by discourse theory. Nevertheless, I want to introduce this concept more fully because I believe that many readers will have had little exposure to it.
I find the metaphor of documents as negotiation useful because it provides me with some insight into the relationship between author and reader. Authors are like diplomats whose job it is to convince readers to accept a set of terms. In most documents, the roles of author and reader are unequal; traditional documents give authors control over both the proposition to be negotiated and the structure of the negotiation, leaving the reader with only two choices:
At the end of the document, the author hopes that the reader has accepted his or her viewpoint, but this can not be assured; and if the reader wishes to question the author to clarify a point, expansion of the negotiation is limited by the medium.
Hypertext changes the way documents can be negotiated. Much like a choose-your-own adventure novel encourages readers to take an active role in choosing the narratives direction, hypertexts linking capacity encourages authors and readers to renegotiate control over the document. This renegotiation has been met with resistance by many in the publishing business who are accustomed to a producer-consumer relationship. They see hypertext as merely a new container in which to package their product and they discourage writers from linking to outside sites because it is thought that giving readers new narrative choices will reduce sales. To be fair, hypertext is a relatively new medium and one can not expect authors to change over night. It is more reasonable to expect relationships to be renegotiated slowly as people adapt to the new architecture of hypertext.
But what if the architecture of hypertext is not predestined? What if our current actions shape hypertexts design? Then it would seem that the negotiating positions of future authors and readers depend upon how well we understand and shape hypertexts architecture to fit our values. This next section explores the concept of openreferences in the hope that a greater understanding of hypertexts potentials will allow us to better shape hypertexts future. 
Q: What happens when hypertext is taken to its logical conclusion?
A: Everything becomes a link for readers to negotiate.
Currently, when authors create links, they must choose one page as the target of their reference. This makes links a valuable commodity and has given rise to affiliate programs. I believe that the limitation of only one site per link is a problem because, when combined with affiliate programs, commercial interests tend to dominate all other considerations. Amazon.coms "associates" program provides an illustration of this.
Amazon.com would like all narratives about books to lead to book sales at their site. To encourage this, they offer authors who link to amazon.com book pages a commission of about five percent. They sweeten the deal by providing readers with useful resources such as book reviews, related books and links to publishers and authors sites. Faced with an all-or-nothing choice, the decision is obvious: authors who create links to books will overwhelmingly link to sites like amazon.com.
But there is more to books and more to the internet than e-commerce. One alternative to the bookstore is the library. Imagine an internet where all links to books lead to both bookstores and libraries. Imagine an internet where the list of link options can be negotiated by authors and readers. To take the concept of reader empowerment to its logical conclusion, imagine that readers are able to turn any object into a link, regardless of the authors referencing intentions.
Merriam-Webster has already implemented one component of this sort of openreferencing system by creating a browser plug-in that allows readers to look up the definition of any word they select. Imagine what would happen if every application provider created a browser plug-in that allowed readers to:
As individuals and institutions get increasingly sophisticated in their referencing, it becomes apparent that words are becoming objects and that the concept of object-oriented design could provide hypertext with a great deal of added functionality and organization.
Object oriented design treats everything as an object with a set of associated methods. Thus, the earlier amazon.com singular link problem could be solved by assigning all book objects multiple methods. Books could have a buy method, a borrow method, a related method, a review method, and cite method. The interface for object-oriented hypertext could look like a simple Windows menu.
Before I get carried away with parallels between hypertext and technical concepts like objects and APIs, allow me to take a quick detour into the realm of linguistic association by asking the following question:
What do you think of when I say the word "reference"?
Perhaps you think about:
Of the three associations, I think the first two conceptions of references are the most common, and a subset of the third. It is this third category that gets implemented in a hypertext referencing system. Each reference becomes a symbol for an action or set of actions, whether it be a means of citation, a way of evaluating trust, or to use the metaphor of cyber "space", a "place" where one can go to perform new actions.
In the technical sense, references become an Application Programming Interface (API). Each reference object has a set of methods (or programs) that can be performed upon it. Objects are usually divided into classes with a system of inheritance.
To use vehicles as an example, imagine a vehicle class with car, truck, and three-wheeler subclasses. All of these subclasses inherit "wheel" objects, but in different numbers and of different types. Truck tires could be different than car tires; and three-wheelers could be prevented from having more than three wheels. Trucks could possess "bed" objects in which other objects could be placed. All of these vehicle objects could have the "drive" method performed upon them.
As a technical concept, object oriented design is very useful, but when applied to people, it can become dehumanizing. What are the classes of people and what are the methods that can be performed upon them? This is the question that object-oriented hypertext asks.
Some readers may still not fully recognize the problem of objectification. To illustrate it further, I will enlist the help of the AmIhotOrNot.com website, a website which exemplifies objectification. AmIhotOrNot.com allows anyone to submit any picture that is on the internet to the AmIhotOrNot.com web site. The picture is then displayed to the sites visitors, who rate the person in the picture on a scale of 1 to 10, according to how "hot" they judge the submitee to be. As visitors rate each picture, they are displayed another persons picture to rate, and the average rating that the AmIhotOrNot.com collective has assigned to the previous persons picture.
The genius of this site is its ability to efficiently exploit humans desire to judge each other and know how others are judged. AmIhotOrNot.com does not store any pictures on its web server. It merely allows readers to direct other readers to servers throughout the internet. These servers send their pictures to AmIhotOrNot.com viewers for display within the AmIhotOrNot.com frame of judgment. The people being judged may have voluntarily submitted their pictures to the AmIhotOrNot.com website or the pictures may have been submitted without their knowledge or permission. Either way, the server on which their picture is stored will send a digital image of them to AmIhotOrNot.com viewers.
To help those readers who are visual learners, I
have devised a menu similar to the one used in the amazon.com book
example. Imagine what would happen if
people could classify and program references to each other in any way that they
wished. The AmIhotOrNot.com method is
one obvious candidate for inclusion in the following menu:
What methods should the reference to Jim Summerset contain? Should the architecture of a referencing system permit some methods and forbid others?
When confronted by the social problems inherent in object-oriented design, I propose that we ask ourselves the following question:
"Can we use an object-oriented design without suffering the problem of objectification, and if so, how?"
One attempt at solving the problem of objectification is to give individuals private control over references to them. Architects of a referencing system could design the software to require that all methods and classifications be approved by the person to whom they refer. This might reduce the potential for abuse, but sometimes there may be legitimate times when people would wish to use methods that others have not approved.
Most modern westerners believe that free speech is a right, independent of whether the audience or subjects of the speech approve of the ideas being expressed. Imagine then, a method in which consumer products could be critically reviewed-- a "Consumer Reports" for the information age.
Suppose that customers shopping in a future store could use a referencing system that allowed them to review companies labor and environmental practices. This referencing system could also display evaluations of products design and construction and a section of customer reviews. Should reviewers be required to obtain the permission of the companies whose products they wish to review? Most westerners would answer in the negative, citing this as a case where free speech rights or public interests overrule private interests.
What rights do authors have? Or perhaps better put, what should be the relationship between authors and readers? It is possible to design a system that puts either in a relatively stronger negotiating position. If the system is designed by authors, it would likely enforce a degree of authorial control over copying, criticism, and other referencing methods. In short, it would retain a large degree of publishers author-ity. If readers designed the system, it would probably allow a much larger number of derivative works to be produced, including stronger forms of criticism and a greater range of other referencing methods.
All of this talk about negotiation assumes that participants are to some degree equal enough to make negotiation possible, for if either of the participants were in a totally superior position, the one could coerce the other. The question I find myself asking next is: "How equal do participants have to be?"
One method of evaluating a negotiation could allow for a great degree of inequality. Suppose for example that I have a stone and you have a stone, and that each of us can throw our stone at the other. Now suppose that I have ten stones and you have one. Or suppose that I have ten stones and nine friends who have ten stones. You could argue that despite our inequality, we still have to negotiate because if you wanted to play the odds of 1 stone against 100 stones, you still might be able to hurt me.
All of this talk about negotiation brings to mind images of "truels" and other game theory examples. It seems to me that game theory could be useful in evaluating possibilities, but it will not provide us with answers to questions of values. What we need to do, in assessing the design of a referencing system, is evaluate the positions of the individuals, groups, and institutions involved and articulate a shared set of values. Then using these values, construct a "game" that all feel they can participate in fairly.
One further concept that I feel needs to be present in a referencing system is trust. In my detour into linguistic association earlier, I noted that the concepts of reputation and trust are built into the English world "reference." Perhaps this is just a coincidence, but I dont think so. I propose that as referencing systems evolve, they will introduce a greater degree of trust (or reputation) architecture.
One method of evaluating trust being experimented with is the idea of a "trust metric." Trust metrics use graph theory to calculate the reputation of nodes in a network. Not having studied graph theory, my understanding of it is rather shallow, but I usually explain it this way:
Imagine that I trust you and you trust Sue.
Can I extrapolate my trust of you to Sue?
My answer to this is "Yes, sometimes, maybe I can." The problem lies in the number of degrees of separation, the similarity of domains being considered, and the accountability of Sue to you, me, and the rest of the world. If I trust your cooking and you trust Sues computer programming expertise, I likely cant determine anything about Sues cooking. Likewise, you may find that you and I have similar tastes in movies, but that doesnt mean that you would trust me to baby-sit your children or manage your stock portfolio.
I believe that our ability to build methods of trust into a referencing network is critical to the networks success. Parents and politicians are currently demanding that computer software filter the internet. This is wildly unreliable. The computer software uses algorithms that search for words, pixels, or word and pixel combinations that the user has forbidden. Thus, the computer must distinguish between the pornographic use of the word "breast" and "breast cancer" and between skin tone colored pixels comprising a backdrop for Governor Bush and his wife, and pixels composing a pornographic photo. Getting humans to agree with each other on values is difficult enough; getting humans to agree with computers seems nearly impossible. Thus, I conclude that computer methods of trust evaluation are woefully primitive and that a successful method of allowing humans to share evaluative responsibility is needed.
By this time I suspect that at least a few readers have concluded that all this negotiating requires too much work. The cynical will say that the majority of the population would be content to let others make decisions for them as long as they remain comfortable and entertained. I agree, in part, but in my optimistic youth, I believe that a large number of well-educated members of the younger generation, who are growing up in the hypertext environment of the internet, are beginning to expect more negotiable media.
The metaphor of documents as negotiations provides a useful framework for thinking about media. When applied to a system of radically expanded hypertext, which I call "openreferences", this framework allows for the prediction of a growing tension between the roles of authors and readers. How this tension gets resolved is up to us.
Agre, Phil. "The Internet and Public Discourse." FirstMonday 1998. <http://firstmonday.org/issues/issue3_3/agre/index.html>
Andersen, Trond. "Consumer Power via the Internet." FirstMonday 1999. <http://firstmonday.org/issues/issue4_1/andresen/index.html>
Brown, John Seely and Paul Duguid. "The Social Life of Documents" FirstMonday 1996. <http://firstmonday.org/issues/issue1/documents/>
Cameron, Robert D. "A Universal Citation Database as a Catalyst for Reform" FirstMonday 1997. <http://firstmonday.org/issues/issue2_4/cameron/index.html>
Lessig, Lawrence. Code and other laws of cyberspace. New York: Basic Books, 1999.
Spinello, Richard. "Ethical reflections on the problem of spam" Ethics and Information Technology
 Jakob Neilsen disagrees. He argues that creating useful links gains authors valuable reputation capital. Neilon has written extensively about usability on the web. He puts out a column titled Alertbox <http://useit.com/>
 Lawrence Lessigs Code and other laws of cyberspace argues that technological architecture is not neutral. Architecture is not destiny, but architecture enables some actions and limits others. Which actions are allowed and which are forbidden is determined by the architects.
 Phil Agre notes the tendency of computer designs to objectify and methodize (my words) in The Internet and Public Discourse. FirstMonday 1998. <http://firstmonday.org/issues/issue3_3/agre/index.html >
 Technically savvy readers may see a resemblance between my sample book menu and the Windows Sendto menu. Thats because what you are viewing is a screenshot of a modified Windows Sendto menu. Unix devotees will recognize that the Sendto menu is nothing more than a graphical interface for the pipe command. Thus in unix terms, references becomes an output file that can be piped to any program that accepts file inputs.
 Trend Andresen describes a system like this in a 1999 FirstMonday article titled "Consumer Power via the Internet." Andersen argues for government creation of such a system. <http://firstmonday.org/issues/issue4_1/andresen/index.html>
 Spam is another example where absolute conceptions of free speech and privacy conflict. Distributers of unsolicited bulk e-mail argue that they have speech rights, whether they be commercial or not. For more on the ethics of unsolicited bulk e-mail, see Richard Spinellos article "Ethical reflections on the problem of spam" in Ethics and Information Technology.<http://www.wkap.nl/sample.pdf?240815>
The most popular form of email filtering currently being used relies upon the MAPS Real-Time blackhole list, a cooperatively assembled blacklist.<http://mail-abuse.org/rbl/>
The X Naming System proposes a system of contracts to address the problem of spam. They propose that all e-mail servers prompt incoming mail senders with a contract, promising that the content of their mail is not spam. Such a system could logically be extended to many other domains, including telephone service and pornography filtering.
 I share Lawrence Lessigs view of copyright as a bargain between the public and artists. I think this fits well with my concept of documents as negotiation. <http://www.thestandard.com/article/display/0,1151,16071,00.html>
 Ted Nelson proposed a referencing system in the 1960s called Xanadu Docuverse It was designed to allow readers to navigate a wide range of content, while retaining publishers control over quotations and copyright revenues. The Xanadu project is owned today by an Australian company. It has yet to be fully implemented. <http://xandadu.com.au/>
 This picture of Governor George W. Bush and Laura Bush was blocked by the "pornsweeper" software filter. The likely explanation is that the pixels forming the background were too close to the skin tones. <http://dansdata.com/pornsweeper.htm>
 For a classic on the topic of human freedom and responsibilty, see Fyodor Dostoyevskys "The Grand Inquisitor" in The Brothers Karimazov.
 Perhaps a few members of the older generation might be interested too ;-)
Comments on this article would be appreciated. Please e-mail email@example.com