Title:

Toward Dialogue Documents as Creative Conversational Tools Manabu Ueda ^*1 Center for Coordination Science Sloan School of Management Massachusetts Institute of Technology July, 1998

Long Abstract:

In this paper I propose the creation of a dialogue document for making available and explaining the knowledge contained within a creative conversation process. A conversation is an important method of collaboration. Usually it is the knowledge which results from conversations that gets described on a document, and it is rare to see recorded the knowledge which only appears in the conversation process.

In a creative conversation where ideas are created, decisions made or problems solved, the knowledge used and created within the conversation process can be as important as the knowledge obtained as a result. As I move toward a more knowledge and information oriented society, the so called knowledge society, it may become important to explain and use the knowledge which appears in the process of intellectual activity.

There are some reasons why knowledge included in conversation processes is rarely explained in a document. First, the result of the conversation is typically more important than the process, because the result drives the subsequent actions. The standard document is used as a resource which comes out of the conversation process to initiate the next stage. Second, actual conversations include subjective or intuitive interactions, and in this way they often seem irrational. These elements run counter to the modern tendency to emphasize the objective rationality of the result of the conversation. Therefore the actual conversation processes are often omitted or revised into a rational story by their interpretations. Third, transcripts of actual conversation are often difficult for observers to understand because actual conversations often strongly depend on the context and common knowledge of the participants. Moreover, the speakers generally don't take into account the people who listen to or read the transcript of the conversation.

I argue that documents in dialogue form, which we call dialogue documents, are the most effective way to explain knowledge included in the conversation process. The reason is that dialogue document allows readers to have an imaginary experience as audiences of the conversation. This means that dialogue documents convey not only explicit knowledge but also tacit knowledge by relying on the reader's active formulation of the experience. Perhaps this is the essential value of dialogue.

The dialogue documents I propose are documents of edited actual transcript for readers. 'For readers' implies that the dialogue document can be edited for conveying knowledge in a conversation process to readers. In that sense, I distinguish between a transcript of conversation and a dialogue document by whether it was edited for the readers' understanding or not. And this definition doesn't limit the media to paper documents.

In a creative conversation which becomes source material for a dialogue document, it would make the editing work of the dialogue document easier if the participants shared the purpose of the conversation and spoke with future readers in mind. However, this is not a necessary condition.

I must also consider how the popularization of IT products and networks through the evolution of IT has made some changes to the activities of conversations. In particular, I can point out three characteristic changes: increasing diversity of conversations, easy recording of conversations, and change in people's reasons for having a conversation.

Finally, I discuss some issues and possibilities from the technological and social aspects for the future research where I am going to design a prototype 'dialogue document editing tool (DiET)' for getting experience through practical research.

*1 Manabu Ueda is a visiting scientist of the Center for Coordination Science MIT from Fuji Xerox Co., Ltd. Japan.

Table of Contents

1. Introduction

2. Knowledge included in a Conversation Process

2.1 Nature of the Knowledge

2.2 Issues of Representation

3. Dialogue 3.1 Definition

3.2 Dialogue as a method of knowledge representation

4. Dialogue Document 4.1 What are Dialogue Documents?

4.2 Dialogue Document Editing

5 Dialogue Document and IT 5.1 Conversation and IT

5.2 Message Quality in Conversation Media

5.3 IT Support for Dialogue Document Production Process

5.4 Social Aspects in Dialogue Document Production and Use

6. Conclusion

1. Introduction The most successful meetings tend to be those where people collaborate to create a document; the meeting is used to do work rather than just talk about doing work. [Schrage, 1990, p. 111] A conversation whether oral or written (e.g., in a Chat Room or e-mail exchange), is one important method of collaboration. It includes a lot of intellectual activities. Usually it is the knowledge which results from conversations that gets described on a document, and it is rare to see recorded the knowledge which only appears in the conversation process. However, the knowledge which appears in the conversation process may be as important as the knowledge in the conversation results, in the same way that patents of a production process are as important as patents in a product.

I can make a transcript for recording an actual conversation; however, actual conversations tend to be fractured, hesitant and ungrammatical. So it is difficult for observers to understand their contents.

On the other hand, there is a form of dialogue which can be used as a method of passing on the knowledge from a conversation process. A document which conveys a conversation in dialogue form is known as a 'dialogue document'. Most dialogue documents, such as a magazine interviews, are edited from the actual conversations. They include supplements to convey missing nonverbal signs, connections, or circumstances by textual expressions as marked by underlining in the following example:

INTERVIEW 'We Have to Innovate' p. 39, NEWSWEEK Mar. 9 1998
...
Newsweek: You've been called upon to make important legal decisions as opposed to technical ones. Is that tough?
GATES: No, I think those things were quite straightforward. [But] I think it's fair to say that [in terms of] contributing to political causes and involvement in Washington D.C. I am - or at least was - naive. The political arena there, we are going to have to do more.
... A dialogue document has the potential for decreasing equivocality, redundancy and noise from a transcript while keeping the conversation process; moreover, I can insert or append some supplementary explanations into a dialogue document to help readers understand it. Thus, here in this paper, I focus on 'dialogue documents' as a method for conveying useful knowledge in a actual conversation process.
In addition, popularization of IT products and networks through the evolution of IT has made some changes to the activities of conversations. For example, in Newsgroups or Chats, the provider retains the previous conversation log files to provide conversational context to newcomers. Also, online customer service and support desks (e.g. some homepages relating to Health, Computer or Government) use their conversation log files to create FAQs (Frequently Asked Questions) databases. They edit the log files (Q&A) to give information to customers who have similar questions. By doing this, they can reduce the cost of dealing with the same question again and again. In addition, IT could support more creative conversations, as I see explained in the following introduction to a recorded conversation in the Radcliffe Quarterly: ... we convened a Radcliffe Roundtable to discuss how technology is changing the way life is lived at the close of the twenties century. Traditionally, our roundtable discussions are held at the Harvard Faculty Club over coffee and croissants with recording devices set up to tape the proceedings. This time, however, we sought a form that would match the content. We experimented with three discussion formats: an online chat room, an asynchronous "threaded" discussion, and, finally, electronic mail, a communications tool familiar to most of us. In the end, the e-mail format was the most conductive to thoughtful discourse. What follows is an edited, composite transcript of our conversations in cyberspace. [Assessing the Revolution, Radcliffe Quarterly, Summer 1998, p. 8] So just as I don't limit the medium of the conversation to synchronous oral exchange, and also I don't limit the medium of dialogue document to paper documents. In addition, I assume that a creative conversation on IT can exist over the various kinds of communication tools.

This paper begins by discussing the knowledge in a creative conversation process and the issue of its representation. I next examine dialogue as a method of knowledge representation. I also propose a dialogue document as a creative conversational tool for conveying knowledge included in a conversation process. Finally I discuss the significant features and issues of the IT support and social aspects of the process of producing a dialogue document.

2. Knowledge included in a Conversation Process

In recent years, the argument 'knowledge becomes a competitive resource' has attracted people's attention. For example, In his book Post-Capitalist Society, management guru Peter F. Drucker says:

The basic economic resource - "the means of production," to use the economist's term - is no longer capital, nor natural resources (the economist's "land"), nor "labor." It is and will be knowledge.[Drucker,1993, p. 8] Similarly, futurologist Alvin Toffler says in Power Shift: Because it reduces the need for raw materials, Labor, time, space, and capital, knowledge becomes the central resource of the advanced economy.[Toffler, 1990, p. 88] Two leading Japanese business experts, Ikujiro Nonaka and Hirotaka Takeuchi, discuss knowledge more concretely in their book The Knowledge Creating Company. They divide knowledge into two kinds, explicit knowledge and tacit knowledge. Moreover, they model an organizational knowledge creation cycle: The two forms of interactions - between tacit knowledge and explicit knowledge and between the individual and the organization - will then bring about four major processes of knowledge conversion which all together constitute knowledge creation: (1) from tacit to explicit; (2) from explicit to explicit; (3) from explicit to tacit; and (4) from tacit to tacit.[Nonaka and Takeuchi, 1995, p. ix] 2.1 Nature of the Knowledge

I focus on creative conversations because a lot of knowledge is used and created in the conversation process. A certain amount of information (formalized data and information) is described in a document such as meeting minutes. However, some of the knowledge in a conversation is seldom described in a document. For example, I suggest that the following processes may occur:

A process of chain reactions (anabolism and catabolism): You explain your idea which has developed from others' ideas, and that idea also reminds others of further ideas.
A process of dialectic: You put up an objection to other people, and you and the others try to get over this conflict in order to arrive at a new higher idea or knowledge.
A process of reflection: You explain your idea to others, and you expand or change your idea according to the feedback of others.

Each of these concepts can be used in isolation, but together they complement each other synergistically.

On another level, I assume that tacit knowledge resulting from experience or intuition which is difficult to explain accurately, is used for interpretation, selection, reasoning and making decisions in a conversation. This knowledge may not belong to just one person but may be shared by the whole group. The members of the group complement each other and complete this knowledge. Also this knowledge varies depending on the combination and number of people. Sometimes I can see patterns in these dynamic sets. In considering the usability of these kinds of patterns, architectural designer Christopher Alexander said the following:

Each pattern describes a problem which occurs over and over again in our environment, and then describes the core of the solution to that problem, in such a way that you can use this solution a million times over, without ever doing it the same way twice. [Alexander, 1977, p. x] He gives the example of how a drugstore, a pedestrian signal and a newspaper stand function as a set. However, the function doesn't independently exist in each object. In one situation, when you are waiting to turn on the signal, you may buy a newspaper from the newspaper stand using change from the drugstore.

Alexander relies on natural language and lots of examples to describe the patterns rather than formal language. I also think that if we try to organize the knowledge included in dialogues, it should be explained by natural language scenarios in addition to formal models such as those that implement the structured conversations of The Coordinator [Winograd, 1988], or the aggregate nodes of gIBIS [Conklin et al., 1988]. Using scenario-based, natural language can provide a fuller and more realistic context for understanding knowledge. Thus a dialogue document may capture some of the tacit as well as the explicit aspects of knowledge.

2.2 Issues of Representation

There are at least three reasons why knowledge included in conversation processes is rarely explained in a document:

the role of ordinary documents
the need to better represent conversation processes rationally
cost of conversation editing

First, ordinary documents generally describe the result of the conversation. The result of the conversation is assumed to be more important than the process, because the result drives the subsequent actions. The ordinary document is used as a resource which comes out of the conversation process to initiate the next stage. The perception of the role of ordinary documents, to convey the necessary information resource to the next action efficiently, may be rooted in the modern information society.

Second, actual conversations include subjective or intuitive interactions, and in this way they often seem irrational. These elements go against the modern tendency to emphasize the objective rationality of the result of the conversation. Therefore the actual conversation processes are often omitted or revised into a rational story by their interpretations. Karl E. Weick points out this kind of imposed rationality. He explains the notion as both an issue-specific rationality and a post hoc rationalizing device in the following excerpts:

The suggestion that rationality is issue-specific is made concrete in Westerlund and Sj?strand's (1979. p. 91) observation that rationality is an honoriflc label "given to the individual or group acting in the manner the evaluator wishes." Rational decisions typically mean managerially rational," which means rational in the eyes of the people on top, the owners, the current dominant coalition. When these stakeholders change, the definition of rational conduct also changes.
...
Organizations use rationality as a facade when they talk about goals, planning, intentions, and analysis, not because these practices necessarily work, but because people who supply resources believe that such practices work and indicate sound management (Pfeffer, 1981, pp. 194-196). The appearance of rational action legitimates the organization in the environment it faces, deflects criticism, and ensures a steady flow of resources into the organization.
...
The final sense in which rationality is used by newer theorists is as a post hoc rationalizing device (Staw, 1980). "Societal ideologies insist that actions ought to be responses - actions taken unreflectively without specific reasons are irrational and irrationality is bad - . So organizations justify their actions with problems, threats, success or opportunities" (Starbuck, 1983, p. 94). The sequence in this quotation is the key point. First action occurs, and then the "reasons" why the action occurred are invented and inserted retrospectively into the organization's history. The action is reframed as a response to a threat, a solution to a problem that becomes clear only after the action was finished, a response to something that no one realized was a stimulus until the outcome became evident.
[Weick, 1985, pp. 110-111] A third reason that conversational knowledge is rarely explained in a document is that transcripts of actual conversations are often difficult for observers to understand because actual conversations often strongly depend on the context and common knowledge of the participants. Moreover, the speakers generally don't take into account the people who listen to or read the transcript of the conversation. Therefore, we have to incur costs in editing the transcript into a form more understandable to potential listeners or readers. The cost must be balanced against the benefit of the knowledge included in conversations. So I choose carefully conversations which have the most potential, because not all conversations have enough benefits to warrant the costs.

In this paper, I suggest that we can use IT to improve the productivity of knowledge documentation included in a creative conversation while decreasing the costs. Such documentation increases the opportunities for conveying and using the knowledge. If the benefits are realized, that makes up for the production costs of the document. Also, as more people become familiar with such dialogue documents, as well as with the limit role of ordinary documents and their imposed rationality, they will influence those around them [Fulk and Steinfield, 1990], spreading the recognition of knowledge contextualized in conversation.

3. Dialogue

3.1 Definition

It is important to consider the implications of "dialogue" before discussing the potential of dialogue documents. As defined in 'Merriam-Webster's Collegiate Dictionary', "dialogue" has the following meanings:

1: a written composition in which two or more characters are represented as conversing

2.a: a conversation between two or more persons; also: a similar exchange between a person and something else (as a computer)

2.b: an exchange of ideas and opinions

2.c: a discussion between representatives of parties to a conflict that is aimed at resolution

3: the conversational element of literary or dramatic composition

4: a musical composition for two or more parts suggestive of a conversation

In this paper, I mainly use this word 'dialogue' to mean " a conversation between two or more persons," especially one involving "an exchange of ideas and opinions. " I focus on creative conversations, in which a lot of knowledge appears in the conversation process. I use the words 'dialogue document' to mean a document presenting edited actual transcript for readers. The dialogue document unlike a transcript of conversation, is edited for readers' understanding. Usually a dialogue document, such as a magazine interview, has a reason for conveying information and it is edited for that purpose.

3.2 Dialogue as a method of knowledge representation

In this section, I discuss the features of the dialogue form from the point of view of knowledge representation, and argue that the documentation of dialogue form as the dialogue document has a potential to convey tacit knowledge included in a creative conversation process.

Dialogue has a long history of use in communicating knowledge such as philosophy and religion,: including in the Socratic Method in the Dialogues of Plato, the Bible, the Talmud, Old Chinese philosophies (The Tao of Dialogue), Buddhist teachings (The Zen of Dialogue), interactive preachings, and so on. Many kinds of dialogue documents exist, from magazine interviews to dialogues which are not based on actual conversations but only use a dialogue form for explaining the author's idea. Bakhtin interprets Plato's understanding of the similarity between actual dialogues and simulated dialogues in one's head, as follows:

Plato, for example, understood thought as a conversation that a man carries on with himself (the Theaetetus, the Sophist). The concept of silent thought first appeared only with the mystics , and this concept had its roots in the Orient. Moreover, in Plato's understanding of the process, thought conceived as a "conversation with oneself" did not entail any special relationship to one's self (as distinct from one's relationship to others); conversation with one's own self turns directly into conversation with someone else, without a hint of any necessary boundaries between the two.[Bakhtin, 1981, p. 134]

In contrast, an interview extracts information or knowledge from a person in an actual conversation. An interview is a conversation between interviewer(s) and interviewee(s). Skilled interviewers coordinate the conversation process and translate tacit knowledge to explicit knowledge, as in this example described by Nonaka: The Creative Lifestyle Focus Center started various activities that enable Sharp to create new concepts based on inputs from consumers. One such example is the "Trend Leader System," which brings together about 600 "Leading consumers," ranging from junior high school students to senior citizens in their seventies. Depending on the nature of the information needed for developing new products, relevant people clustered into "focus groups" are called into the center. Skilled interviewers elicit information from them that may remain tacit otherwise. This system which started in 1985, enables Sharp to predict consumer trends ten years into the future. [Nonaka and Takeuchi, 1995, pp. 189-190] Dialogue develops sequentially. Things change along a time line. For example, the changes may include adding and creating new information, constructing logic, creating understanding, and making decisions. Basically, a conversation has a linear structure which is its time line, but the content includes semantic and contextual relations that, create a lattice structure like spaghetti or textiles. Well-edited conventional documents such as reports take the outcomes of one or more conversations and construct a consistent hierarchical structure by teleological summarization. A dialogue form has a backbone of a time line. But it is edited from a conversation as a lattice structure into a multi-interpretable structure that may be called a semi-lattice structure.

A dialogue with a semi-lattice structure has a polymeric structure in which people can interpret many different structures which are sometimes interconnected. Speech can be seen to contain a variety of structures (e.g. deductive, chronological, geometrical, and comparative). A person might think of these structures as being organized into one big conglomeration. However we can only present them one by one in discourse. In addition, in a dialogue, the discourse might be disturbed, extended, or reinterpreted by the interactions among other speakers. Dialogue form may be able to coordinate the lattice structure of conversation into the semi-lattice structure for making sense for the listener or reader.

	Dialogue Form	Well-edited conventional Form
Backbone Structure Semantics Interpretations Understanding	Time line Semi-lattice Dialogic Multiple Comprehensive	Rational logic Hierarchy Monologic Single Logical step

Table 3.1

Dialogue form guides the listener or reader by its time line. The reader can trace the same development of the dialogue as the people who are communicating through it because the reader is receiving information in the same order as those people. A reader can share much of the conversational context of the participants in the dialogue.

In a sense, dialogue form is a methodology for allowing listeners or readers to get a sense of some tacit knowledge underlying the conversation by allowing them to have an imaginary experience as audiences of the conversation. Polanyi emphasizes that the knowledge which can't be explained by words is formed as a result of activities which actively formulate your experience when you explore that knowledge.

Gestalt psychology has demonstrated that we may know a physiognomy by integrating our awareness of its particulars without being able to identify these particulars, and my analysis of knowledge is closely linked to this discovery of Gestalt psychology. But I shall attend to aspects of Gestalt which have been hitherto neglected. Gestalt psychology has assumed that perception of a physiognomy takes place through the spontaneous equilibration of its particulars impressed on the retina or on the brain. However, I am looking at Gestalt, on the contrary, as the outcome of an active shaping of experience performed in the pursuit of knowledge. This shaping or integrating I hold to be the great and indispensable tacit power by which all knowledge is discovered and once discovered, is held to be true. [Polanyi, 1983, p. 6]

Allowing listeners or readers to have an imaginary experience as audiences of the conversation might mean that dialogue documents allow access to same tacit knowledge by relying on the listeners or readers active formulation of the experience. Perhaps, this is the essential value of dialogue form.

4. Dialogue Document

4.1 What are Dialogue Documents?

I use the words 'dialogue document' to mean a document representing edited actual transcript for readers. Dialogue documents are very different from transcripts of conversation, as well as from formal reports summarizing results of a conversation.

'For readers' implies that the dialogue document is edited from the original conversation in order to convey knowledge in a conversation process to readers. I distinguish between a transcript of conversation and a dialogue document by whether it was edited for readers' understanding or not.

Of course, it would make the editing of the creative conversation into a dialogue document easier if the participants shared the purpose of the conversation and spoke or write, in the case of electronic media such as chat or e-mail with future readers in mind. However this is not a necessary condition.

A transcript made on a recording tape (including a videotape) can record the whole of the conversation as a physical phenomenon. A rigorous transcript of conversation such as conversational analysis in ethnomethodology [Sacks, 1979, pp. 23-53] can describe the fine details of a conversation. These transcripts preserve to the extent possible the actual situation in a conversation. Also, they allow listeners or readers to have an imaginary experience as audiences of the conversation. Speakers in actual conversations, however, don't take into account the people who listen to or read the transcript of the conversation. Actual conversations often strongly depend on the context and common knowledge of the speakers. Therefore it is difficult to understand the contents of an actual conversation from its transcript for observers. Anthony Giddens says:

It is a sobering experience to hear a tape recording or read a transcript of conversation to which one has contributed. Conversations are much more fractured, hesitant and ungrammatical than most people realize. When we take part in everyday talk, we tend to think that what we say is fairly polished, because we unconsciously fill in the background to the actual words; but real conversations are quite different from conversations in novels, where characters speak in well-formed and grammatical sentences.[Giddens, 1997, p. 74] It is reasonable to edit a transcript to a dialogue document for effectively conveying the understanding to the listeners or readers. In fact, most interviews recorded in articles are edited. The editorial actions which shape a dialogue document includes not only the post-conversation process but also acttions within the ongoing conversation. For example, if you plan to record your conversation for use as a message or a memorandum, it may be that while you are talking you will have an awareness of future users of the transcript, and you may therefore talk carefully in order to reduce equivocality and redundancy, or summarize a complicated interaction before proceedings. These kinds of actions are included in the editorial actions for a dialogue document.

We see from Table 4.1 that a dialogue document has some of the advantages of both a transcript and a conventional document in the case of conveying knowledge in a conversation process.

Needless to say, the purpose of a dialogue document is different from that of an ordinary document such as a rationalized (well-interpreted and structured) document. The ordinary document is a resource which is usually handed over from the conversation process to the next stage. In this situation, the result of the conversation is assumed to be more important than the process, because the result drives the subsequent actions. However, as I have pointed out, if we want to represent most of the knowledge in a natural conversation, it is essential to include the process whereby we arrived at the result ? that is, the process whereby the decision was made, the idea created or the problem solved.

A dialogue document has a potential for decreasing equivocality, redundancy and noise from a transcript while keeping the semantics of the conversation process (some of the structural dependencies of discourses); moreover, we can insert or append some supplementary explanations into a dialogue document to help readers understand it.

	Transcript	Dialogue Document	Ordinary Document (e.g., a report)
Knowledge of the conversation process	raw	refined	minimal
Summarized or interpreted knowledge as conversation results	none	a little	a lot
Conversation process	preserved	refined and enhanced	mostly lost
Semantics	fractured, hesitant and ungrammatical	refined	usually modified
Context-dependency, redundancy, equivocality and noise	preserved	coordinated	suppressed
supplementary explanation	none	some	some
Benefits for user	low	medium ^*1	medium
Information retrieval cost (e.g. time for getting an outline)	high	medium ^*2	medium
Editing cost	none	medium ^*3	high

Table 4.1 From this table we can see that dialogue documents are more effective than the transcripts of conversation for presenting the knowledge in the conversation process, and the capability of dialogue documents makes up for the deficiencies of ordinary documents.

In a sense, from an economic viewpoint, if a dialogue document were to be used more generally, two things would need to be changed: the benefits of dialogue documents for the user (*1 in Table 4.1) would need to be increased and the costs of dealing with a dialogue document (*2 and *3 in Table 4.1) decreased. The benefits of the dialogue document depend on the creativeness and importance of the conversation and on the effectiveness of the dialogue document's use, but in many instances, the benefits may make up for its production cost. The costs have a great potential for decreasing through IT. I discuss this potential in the next section.

4.2 Dialogue Document Editing

In dialogue document editing, I recommend refining a transcript using supplementary explanations, or getting rid of meaningless redundancy while keeping the semantics. However revising should not introduce new ideas the transcript using new ideas which arise from the editing work in to the transcript.

In general, oral conversation and editing of the transcript are not happening simultaneously; however, it is possible to stop and rewind the taped conversation to re-record a different opinion. Also in a text-based conversation (e.g., over e-mail or chat), it is easy to rewrite the original messages in the log. In these cases, an interaction occurs between conversation and editing. We might be able to prevent such rewriting in revising a transcript by assigning as editor someone not involved in the conversation. It is, however, inefficient to do this when refining a transcript using supplementary explanations or getting rid of meaningless redundancy; moreover an arbitrary conflict between speaker and editor might occur.

In another way, it is easy to prohibit altering of discourses by allowing the addition of clearly indicated supplementary explanations; then the editor sees that the prohibition stands on the presupposition that real discourses should be kept in any case. If, however, we take the presupposition that discourses should be edited to convey effective understanding to the readers, such editing activities may be necessary. I guess this situation has similarities to the issue of the authenticity of news, between reporting of facts and elaborate editing of news, or the issue of the recording of an improvisation of jazz music.

Another point I need to consider is that the activity for editing a conversation process may become a narrative activity. Put another way, the activity pulls up another activity which makes the person editing adapt the conversations to a narrative structure to the conversation. Using narrative structures in the editing can be useful for helping readers' understanding; on the other hand, it carries the risk that an actual conversation process will be distorted by fitting it into a narrative structure^*2.

For my purpose here with the dialogue document, it is necessary for the transcripts to be refined, adding supplementary explanations and getting rid of meaningless redundancy while keeping its semantics. However I don't allow the revision of the semantics of the transcript.

*2 Prince calls this "Narratized Discourse" [Prince, 1982]. Shank points out that after the editor has interpreted the meaning from a conversation, he uses 'a skeleton story' for representing the meaning which he has interpreted.

We have a great many words and phrases in English that are really the names of complex stories and thus serve to standardize particular situations. These words are stories or more accurately the names of stories. Although useful for telling stories quickly, they are very dangerous because we have standard sets of reactions to them. [Shank, 1990, p. 148]

Narrative structure in a conversation has been much discussed: [Rumelhart(1957)], [Propp(1958)], [Labov(1972)], [Danes(1984), Functional Sentence Perspective, Thematic Progression Pattern], [Van Dijk(1980), Superstructure, macro-structure], [de Beaugrande and Dressler(1981), Network Model], [Tannen(1984), Conversational Style, Story Round, Constructed Dialogue].

It has been shown in most studies on this subject that narratives have some archetypes that are used to interpret and represent events or scenes. I can see these structures in all sorts of representations: dialogue documents, conversations, and utterances. In sociology, I presume there is the same issue of "documentary method of interpretation", a notion which Garfinkel borrowed from Mannheim in his book 'Studies in Ethnomethodology':

The method is recognizable for the everyday necessities of recognizing what a person is "talking about" given that he does not say exactly what he means, or in recognizing such common occurrences and objects as mailmen, friendly gestures, and promises. It is recognizable as well in deciding such sociologically analyzed occurrence of events as Goffman's strategies for the management of impressions, Erickson's identity crises, Riesman's types of confomity, Parsons'value systems, Malinowski's magical Practices, Bale's interaction counts, Merton's types of deviance, Lazarsfeld's latent Structure of attitudes, and the U.S. Census'occupational categories. [Garfinkel, H. Studies in Ethnomethodology, Chap.3, Common sense knowledge of social structures: the documentary method of interpretation in lay and professional fact finding, 1967 Prentice-Hall]

5. Dialogue Document and IT

5.1 Conversation and IT

Popularization of IT products and networks through the evolution of IT has made some changes to the activities of electronic conversations. In particular, I can point out three characteristic changes:

Increasing diversity of conversations
Easy recording of conversations
change in people's reasons for having a conversation

The most popular conversation tool has been the telephone, but because of the advent of computer networks and high-speed lines, some other types of mediated conversation have become popular. For example, asynchronous conversations such as those produced by e-mail tools, telephone conversations with more than three participants, videoconferences, or synchronous written conversations using chat tools have all become common.

The general public can easily record audio-visual information because of the wide ownership of Video Cameras and VCRs. On the other hand, with tools that deal with written messages such as e-mail tools and chat tools, the conversation is automatically recorded as a communication log file. Technically, any conversation via computers has the potential to be recorded and transmitted as an electronic file, but it is not always desired.

The borderline between spoken discourse and written discourse has gradually become blurred. Especially, short e-mail messages are usually similar in language to spoken discourse. For example, when people receive a short e-mail message which was sent 5 minutes ago, and reply to it within a couple of minutes; the e-mail messages are often more similar to spoken discourse than to a collection of letters.

In conversations on bulletin boards, DL (Distribution Lists) or Newsgroups, communicators indirectly converse with a lot of observers. That is, their messages are sometimes intended not only for other communicators but also for the observers as the audiences.

Transcripts of conversation become used for other purposes, not just as logs. For example, in Newsgroups or Chats, the provider retains the previous conversation log files to provide conversational context to newcomers. Also, online customer service and technical support desks (e.g. many software company's home pages) use their conversation log files to create FAQs (Frequently Asked Questions) databases. They edit the log files into the form of voice questions and answers to give information to customers who have similar questions. By doing this, they can reduce the cost of dealing with the same question again and again.

These above situations produced by the evolution of IT have a potential to facilitate the production of dialogue documents from both the technological and social points of view.

5.2 Message Quality in Conversation Media

I would like to discuss message quality in conversation media. In short, the difference between face-to-face conversations and other conversations is editing time. In a typical conversation, each participant can meet and talk with others face-to-face. They perceive the same contextual world, though from each of their different viewpoints.

Many researchers of conversation analysis point out that in a conversation, message sending and message receiving occur simultaneously for a participant. A speaker often coordinates his/her on-going message using feedback from other people's reactions, the atmosphere, and his/her own voice (e.g. intonation & word usage).

Daft and Lengel classify communication media using a framework of equivocality and uncertainty on information requirements. They discuss information processing in organizations from a viewpoint called 'Media richness' or 'Information richness':

Information richness is defined as the ability of information to change understanding within a time interval. Communication transactions that can overcome different frames of reference or clarify ambiguous issues to change understanding in a timely manner are considered rich. Communications that require a long time to enable understanding or that cannot overcome different perspectives are lower in richness. In a sense, richness pertains to the learning capacity of a communication.

Communication media vary in the capacity to process rich information (Lengel and Daft 1984). In order of decreasing richness, the media classifications are (1) face-to-face, (2) telephone, (3) personal documents such as letters or memos, (4) impersonal written documents, and (5) numeric documents. The reason for richness differences include the medium's capacity for immediate feedback, the number of cues and channels utilized, personalization, and language variety (Daft and Wiginton). Face-to-face is the richest medium because it provides immediate feedback so that interpretation can be checked. [Daft et al, 1986,Pp560]

From the viewpoint of editing, in same space-same time conversations, although the message is strongly influenced by feedback, the editing time for reflecting on feedback is very short. So the messages are often incomplete and ambiguous and shared context and common knowledge support them. However, in other circumstances, when feedback is not present, voice messages become more similar to text messages. For example, when you telephone someone, but that person is not there, you can leave a message on the answering machine, but you have to make (i.e. edit) the point of your message quickly without any feedback. If this is your first call, you might call him/her back later. As we see in this example, message editing is more important in no-feedback or low-feedback conversations.

Voice communications without real time feedback lose an important advantage. El-Shinnawy and Markus compare V-mail (Voice mail) and E-mail (Electronic mail) using media richness theory. E-mail is chosen rather than V-mail for reducing uncertainty. However, contrary to the expectations of media richness theory, V-mail isn't preferred to E-mail for solving equivocality [El-Shinnawy et al, 1992]

In a sense, feedback in real time conversations shapes the ease with which the conversation can be repaired: if you utter an irrelevant or imperfect message, you can follow up with a second message immediately. In contrast, with different place-different time conversations like e-mail and v-mail, it is difficult to follow up immediately. Irrelevant or imperfect e-mail messages are not easily and immediately repaired, which can lead to trouble. In addition people are changeable. For example, in the following case, one e-mail message may go through several stages:

The response you write just after receiving a criticism by e-mail.
The message as revised after you read your reply.
The message as revised just before pushing the send button.

This means it is not human nature to be consistent and comprehensive. So it is important to edit carefully when the message is recorded by durable media in different time (long interval of feedback) conversations. And the sender should take full advantage of the editing time available in different time conversations. (See table 5.1)

	Time for Message Editing	Required Message Editing quality
Same Time	short	low
Different Time	long	high

Table 5.1: Editing time and Quality of Message

The transcript of a written conversation is potentially a good source for dialogue documents; in other words, it includes better edited messages than a transcript of voice conversation.

In comparison with voice conversation, written conversation may be troublesome, because it requires special literacy for typing, editing, and transmitting. However the voice message is often volatile and not available for re-using. Given that the message is intended not only for conversation but also documentation for readers of a dialogue document, these typing and editing costs may be offset by the relative benefit.

5.3 IT Support for Dialogue Document Production Process

To model the dialogue document production process I adopt Coordination Theory [Malone and Crownston, 1994] because coordination theory focuses on the activity rather than the actor who executes the activity. Using this theory, it becomes possible to detect a lot of similar processes over different disciplines, scales, and points of view. Coordination theory provides a basic framework for analyzing a process. In this theory, the word 'coordination' means "managing a dependency among activities. "(B

The activity and dependency are essential elements of the process ontology, and are basic parts of the process description methodology. An activity is named by a verb, and the subject of the verb is not specified. That means the activity is potentially executed by any actors, such as: individuals, organizations, computer programs etc. A dependency indicates a situation where multiple activities handle a common resource. I believe it is especially important to focus on the activities rather than the actors in considering the possible use of IT, because the actors can be replaced and reorganized but not the basic activities.

A dependency gives the activities a topological meaning as resource producers and consumers. There are three dependency types, Flow, Fit and Share, which correspond to the balance of producers and consumers. In this section, however, we would like to look into the basic activities rather than different possibilities of dependencies. So I focus on flow dependency.

The dialogue document production process is constructed of four activities: conversation, recording, editing, and management. (See Fig 5.1). I would like to discuss the process according to these four activities.

Conversation

The conversation activity is a core activity for creating knowledge. The conversation is based on interactive discourse, whether at the same place and time or not.

Today there are various IT communication tools potentially utilizable for conversations. Some tools can be regarded as augmenting one or more the five senses; other tools can be seen as adding value to non-electronic communication such as written or symbolic (semaphoric) communication. However, there is a danger of focusing entirely on the functional aspects of communication technology without considering the organizational context in which the communication takes place. For example, in business situations, I can see that information technology such as communication tools change the structure of work which in turn may require new functions from the technology. While I am going to discuss those social aspects of communication technology in some depth in the next sections, I will make some preliminary comments here:

JoAnne Yates suggests that the history of management systems has a lot of implications for the existence of IT; moreover, to study and interpret historical events may illustrate current problems and issues. Changes in the social communication system, managerial theory, and communication technology express their mutual dependency rather than derive from the invention of new technology. She says the following:

Recent innovations in computers and telecommunications have been so spectacular that contemporary commentators tend to focus solely on the technology, seeing it as the driving force causing changes in other parts of the organization. The case studies in this book, however, illustrate some of the problems with simple technological determinism. Technologies were adopted, not necessarily when they were invented, but often when a shift or advance in managerial theory led managers to see an application for them. Moreover, technologies were often adopted simply to facilitate existing managerial methods; potentially more powerful applications, such as the use of the telegraph for railroad dispatching, were ignored for long periods. The technology alone was not enough - the vision to use it in new ways was needed as well. [Yates, 1989, pp. 274-275] IT communication tools have already been used for conversations. Even though these tools have some constraints in comparison with face-to-face communication, they don't interfere with the contents of conversations. On the other hand, using 'Semi-structured messages' is an important method for mapping communication structures into a format which can be processed by computer. Malone and his colleagues demonstrated the effectiveness of this 'semi-structured' approach, by applying it to automatic information filtering for e-mail and news. In this case the computer filters and routes human messages using the syntactic structure introduced by users. There are also communication protocol support systems: gIBIS [Conklin et al., 1988], Strudel [Shepherd et al., 1990] and the Coordinator [Winograd and Flores, 1986]. In these systems, users construct a message using a set of semantic structures which reflect established and efficient routine patterns of communication. For example, the Coordinator suppresses irregular activities, "breakdowns", taking out deviant communication structures from routine communication patterns.

In contrast, my idea of conversation support is different. Of course, I admit that there are some domain specific conversation routines (e.g. job interviews, panel discussions, counseling, and diagnosis), and they have some iterative patterns, which could be interpreted as protocols. In creative conversations, those kinds of protocols may be also detected, but I think that conversation protocol support is not necessary. I think that the creativity in conversations may be in the breakdowns, not the routines. Creative conversation is most productive when participants are improvising interactions in response to unpredictable communication structures. Therefore, I have deliberately chosen not to interrupt nor hide these kinds of breakdowns which occur between one person and others.

Recording

The recording activity turns a conversation into an editable artifact.

Needless to say, face-to-face conversation requires numerous forms of non-verbal communication - the exchange of information and meaning through facial expressions, gestures, and movements of the body. Although we routinely use non-verbal cues in our own behavior and in making sense of the behavior of others, much of our interaction occurs through talk - casual verbal exchange - carried on in informal conversations with others. It is generally accepted by sociologists that language is fundamental to social life [Giddens, 1997, pp. 73]; therefore a record of discourse is an important social artifact. In addition we may chose to supplement the conversational record with information about non-verbal cues and some of the materials used in the conversation such as a sketch, a formula or a picture.

Today, some speech recognition tools have already come onto the low price computer market (PC or Mac) such as IBM's ViaVoice. These speech recognition systems are intended as a replacement for typing keyboards. However I cannot say they are easy and useful for recording conversations yet, because typically these tools require the speaker's voice sample by means of time-consuming speech exercises, before they work.. Moreover, they require continuous monitoring and intervention. These tools, however, are evolving rapidly, so in the near future, we might be using speech recognition tools in our daily work.

At the present time, the most realistic choice of IT supported conversation tools for the dialogue document might be written synchronous communication tools such as chat tools or shared editors. If a conversation uses those tools, the conversation and the recording of the text take place simultaneously. Moreover, as I discussed at section 5.2, communicators can take the necessary amount of time for editing messages and they can also keep a measure of real time interaction. The written communication allows not only a parallel processing of a conversation and its recording, but also well-considered message exchanges in comparison with ephemeral messages.

Editing

The editing activity transforms the transcript of a conversation into a dialogue document. A transcript is edited by adding, deleting, and rephrasing to promote better understanding by the reader. We naturally understand that some editing can be seen during the conversation, but here this editing activity edits for a transcript of a finished conversation.

Considering for the moment text based dialogue documents (in contrast with hypermedia and multimedia documents), many recently developed documentation tools can be utilized when editing. Furthermore collaborative authoring systems may be able to support collaborative editing by multiple conversational participants.

Note that in general, structure-oriented editors, such as an outline processor, primarily, are designed for ordinary hierarchical documents. This suggests, therefore, that specialized document editor to support the semi-lattice structure of dialogue documents.

Management

The management activity coordinates the dependencies of the three activities described above. It accomplishes this primarily by coordinating the flow of resources such as time, place, people, tools, and information (e.g. ends and means, methods, conditions, roles) among these activities. Fig. 5.2 shows a typical flow of dependencies and resources.

In a general discussion of the management of conversation recording and editing, Michael Dertouzos introduces an interesting hypothetical scenario for recording business meetings in his book 'What WILL be'. In the scenario, a hyper-document called a "structured hyper-outline form" is made by a tool called ATM (Authoring Tool for Meetings). I surmise from his scenario that the hyper-document includes ordinary minutes, a transcript, query indexes, and reference links for outside data.

Indeed, this following scenario doesn't mention knowledge included in a conversation process, and also it might not assume an improvised conversation which focuses on the emergence of ideas rather than the rationality of the conversation process. However, it gives some hints for the practical use of dialogue documents on IT.

An important meeting on the future of global oil supplies is underway at multinational OIie Energetics BV in the port city of Rotterdam, the Netherlands. You are a twenty-first-century secretary charged with the task of summarizing what the five executives and two consultants are saying and displaying around you so that they can later review key portions of the meeting and so that CFO Wilhelmina Maas, who had to be away at one of the firm's drilling rigs in the North Sea, can later find out what was discussed. Any employee cleared to do so should be able to query your notes without having to read through or view the entire two-hour proceedings. You build your summary in a structured hyper-outline form made possible by the ATM (Authoring Tool for Meetings) middle ware module.

As people speak, you hit different keys on your computer keyboard to record pivotal spoken statements or to index something that was said under one of several categories of discussion that you have already set up. You also direct some of the spoken fragments to a speech-understanding program where they will be transcribed and indexed automatically. You do all this rather well. After all, you are a specialist in hyper-organizing live-meeting notes, and you have spent two years learning and refining the techniques that have landed you this job.

Upon her return late that night, Maas calls up your hyper-summary on her computer and asks, What did Jan say about recent developments in France's nuclear power supplies?" Maas is rewarded with a couple of sentences answering her question and two pointers - one to an audio fragment of Jan's key statements on the topic, the other to an online text version of the same. lf Maas gets her answer in two minutes, and the meeting lasted two hours, then you could rightfully brag that your hyper summary gave her a 60-to-1 leverage, or 6,000 percent efficiency. Bravo! [Dertouzos, 1997, pp. 96-97]

By considering this scenario, I have come up with some conditions for the practical use of my dialogue documents as follows:

Dialogue documents, as well as other kinds of documents, can be regarded as representations of collaborative work. Moreover, the representation is selected by user needs. A document which has polymerized representations and that can be realized on a computer, may be able to correspond to various user needs.
The qualities of the supplementary explanations or reference data in the product depend on the speakers' cooperation more than the secretary's skill. The quality is supported by the shared common knowledge of the meeting members, not the editing skills, however much money you pay.
Dialogue document editing might require special (professional) skills, which might exist in the professional interviewer skills of journalists and editors.

Regarding the production costs, if we use the dialogue document to record some creative conversations in an R&D organization, such scenario production may not justify the cost without a sufficient number of users and user benefits. Dertouzos also says: "Hyper-organizing" and related tools will play increasingly important roles in the Information Marketplace for all kinds of different meetings, for talks and presentations, and generally for key events likely to be revisited after they have happened. [Dertouzos, 1997, pp. 96-97]

5.4 Social Aspects of Dialogue Document Production and Use

In this section I discuss social aspects related to the process of producing dialogue documents. First, I describe my working hypotheses about the implications of the number of participants, assignment of editing role, and intellectual property rights. Next, drawing on the work of Zuboff and Giddens, I discuss the issue of surveillance which arises with recorded conversation. I note empirical evidence for Goffman's perception that surveillance has a tendency to make people perform, and this unnatural behavior might ruin a creative conversation.

By considering Giddens' 'structuration' perspective on surveillance, I can suggest that users of dialogue documents will always need to be aware of and attempt to guard against the dangers posed by the potential use of the documents for surveillance, as well as of the dangers posed by impression management in the creative conversation and its editing. Finally I suggest some ideas for encouraging participants and editors to be creative in their use of the dialogue document.

Number of participants: In a creative conversation, people generally agree that fewer participants might be better. Philosophers' dialogues often involve two or three participants, magazine interviews are usually between two people, a collection of letters is usually between two correspondents, and in a chat on the Internet (e.g. Yahoo chat), two speakers often dominate the conversation though a lot of people come and leave that chat room.

Thus, although the creativity of conversations is affected by many things, such as the skill of a interviewer, leadership, or group cohesiveness (including such things as shared knowledge, aim or culture), I recommend that the number of participants in the type of creative conversation that I anticipate be kept to as small a number as possible, optimally only two..

Who edits: Each speaker may be the best editor of his or her own part of the dialogue because it is efficient to do this when refining a transcript using supplementary explanations or getting rid of meaningless redundancy. In this case, it may be important to achieve consensus among all participants about the contents of the dialogue document. Unfortunately, this kind of group editing sometimes leads to arbitrary conflicts among editors. In such cases, computer supported collaborative authoring tools may be useful. [Michailidis and Rada, 1996, pp. 9-43]

Intellectual property: Existing copyright laws (for example those relating to corporations, the academy or publishing) can be adapted to joint authorship; however, an absolute method for estimating each collaborator's contribution may not exist. It is true that conversations rest on individual behavior because we can distinguish each speaker's utterance. On the other hand, conversations have emergent properties, that is, characteristics that are produced when individuals interact, but are not reducible to individual contributions.

Surveillance: Previous research in social psychology has documented the impact of surveillance on the behavior of those observed. For example, if the dialogue document is used for personnel evaluation by estimating each speaker's contribution to a conversation, the speakers may dislike such records or alter their behavior in a manner that ensures good evaluations. On the other hand, if you are in a position of evaluating the members who conversed in the dialogue document, you can find some useful information for the evaluation whether it was intended or not. Giddens explains two forms of surveillance in modern organizations: visual observations and documentation. He argues that persistent documents about people's lives can be used for the second type of surveillance, through files, records, and other forms of documentation [Giddens, 1997, pp. 290-291].

In the case of the dialogue document, the invisible audience, that is the anticipated readers, may therefore influence the behavior of the participants. Shoshana Zuboff calls this kind of surveillance structure created by IT the 'Information Panopticon'^*3. The influence of this surveillance (what Zuboff calls "panoptic power") depends not on physical visibility (as in architectural settings) but on the informational visibility in the IT setting. [Zuboff, 1988, pp. 315-361].

An observer (or imaginary observer) causes people to behave in different way than they would if alone. Goffman explained that people manage the impression they make on other people in conversations. In this sense, he argued that this behavior can be analyzed from the same perspective as a dramatic performer, a performance, and a dramaturgy. [Goffman, 1959] From this self-presentation perspective, people are well informed about the roles and expectations associated with each situation. He assumes that the performer and the situation (which includes other performers and audiences), may be in opposition in cases where the performer wishes to appear to surpass others, and also the performance may be a scheming, tactical ploy, designed to deceive others and to gain power over them. I presume that impression management may occur in the editing stage, as well as the conversational phase.

*3 Panopticon:
According to Foucault, the modern prison has its origins in the Panopticon, an organization planned by the philosopher and social thinker Jeremy Bentham in the nineteenth century. 'Panopticon' was the name Bentham gave to an ideal prison he designed(.)
...
The Panopticon was circular in shape, with the cells built around the outside edge. In the centre was an inspection tower. Two windows were placed in every cell, one facing the inspection tower and the other facing outside. The aim of the design was to make prisoners visible to guards at all times. The windows in the tower itself were equipped with venetian blinds, so that while the prison staff could keep the prisoners under constant observation, they themselves could be invisible. [Giddens,1997, pp. 291-292]
Note that the power of Panopticon is rather like the mechanism which leads (at least some) prisoners to behave politely when observed by prison staff. With the Panopticon it is the possibility of such observation that makes the prisoners internalize the prison staff's eyes. The prisoner's consciousness then splits in two.

In creative conversations I assume, however, that relationships among speakers and those between speakers and readers (a social structure), may not always be opposed. Giddens [Giddens, 1979] argues in his structuration theory that social structures should not be seen solely as barriers to action and as repressive of the agent's ability to act, but are also intimately involved in the production of action. I presume the social structures at play in a dialogue document conversation might be better understood from this structurational perspective. Yates suggests that Giddens' structuration, as a non-positivist social theory, can be used as an analytic framework to describe the influences of institutional structures without losing sight of the individual actors. [Yates, 1997]

A hypothetical danger is that by using a dialogue document as a surveillance tool, it may affect people's impression management. In an organization, if evaluators use a dialogue document as a surveillance tool to evaluate each participant's contribution to the conversation, conversely the participants (speakers and editors) use the dialogue document for their impression management. Then, the usage of dialogue documents becomes instituted for evaluating within the organizational structure, while suppressing the capability of dialogue documents to convey knowledge included in a creative conversation.

This kind of structurational perspective does not eliminate the danger of surveillance, and in creating and using such dialogue documents, individuals should be aware of, and attempt to guard against, their use for surveillance whenever possible. So I can suggest that users of dialogue documents will always need to be aware of and attempt to guard against the dangers posed by potential use of the documents for surveillance, as well as of the dangers posed by impression management in the creative conversation and its editing.

Guidelines and processes for editing, for example, may help. In addition, individuals involved in the creative conversation as well as readers of the dialogue document may be encouraged to enact collaborative roles such as the ones I describe below.

One set of collaborative roles for creative conversations involves a social structure between the speakers and a reader that can be compared to that of basketball players and coach. For example, a group of younger engineers might chose a dialogue document for conveying not only the discussion results but also the process to a senior engineer, because they can expect to get key advice, sometimes based on tacit knowledge, from that engineer.

Another possible social structure to encourage among participants is that of jazz players and their fans. Membership in a task team constituted by interdisciplinary professionals gives full play to each player's individuality while working with other players cooperatively, as in a jazz improvisation. In addition, the fans' feedback may also make them into more creative artists. Thus, the utility of dialogue documents depends on how speakers and readers enact social structures.

These are ways of dealing with some of the problems presented by surveillance and impression management, but users of dialogue documents will always have to be aware of the potential dangers of these problems.

I fully realize that the preceding discussion is highly simplified, and glosses over some of the most hotly contested issues in social psychology, such as language theory and speech theory. But my purpose has, of course, not been to survey these issues, but to consider the social aspects of the dialogue document in practical use, particularly those aspects related to the potential drawbacks posed by surveillance in its use and impression management in its creation.

6. Conclusion

In this paper I have proposed the creation of a dialogue document for making available the knowledge contained within a creative conversation process. I have discussed three main issues: the role of ordinary documents, the need to better represent conversation processes rationally, and the costs of editing conversation. I have looked at the reasons why we rarely see the knowledge from the conversation process recorded in documents, even though this knowledge is in same cases as important as the result of the conversation.

The dialogue documents I propose are documents of edited actual transcript for readers. My argument is that such documents in dialogue form, are the most effective way to provide access to the knowledge included in the conversation process, because the dialogue documents allow readers to become virtual audiences in the conversation. This means that dialogue documents convey not only explicit knowledge but also allow access to some tacit knowledge by relying on the reader's active formulation of the experience. Perhaps, this is ultimately the essential value of dialogue.

To crystallize my notion of the dialogue document, I discussed its features in contrast with those of a transcript of conversation as well as an ordinary document. I analyzed the dialogue document from the perspective of 'production costs and benefits' and 'message quality and editing time'.

Finally, I considered the possibility of IT support for the dialogue document production process and I discussed the implications of both the technological and social aspects of dialogue documents production and use.

It is important to note my epistemology and approach to knowledge representation. I argued that a dialogue document is edited into a multi-interpretable structure that may be called a semi-lattice structure. This means that a conversation can be interpreted as a semi-lattice structure, even if this isn't always the nature of the actual conversation.

I don't intend to argue that there is a semi-lattice structure that is unobservable but which generates observable discourse. That kind of structuralist framework ultimately leads to rationalism, while the notion of dialogue documents comes from an epistemology which enables the process knowledge included in a conversation to be interpreted.

As I said, my hypothesis regarding the capability of the dialogue form came from Polanyi's re-interpretation of gestalt psychology. Gaetano Kanizsa [Kanizsa, 1979] is another gestalt psychologist who worked in visual gestalt perception, and I would like to explain my approach to tacit knowledge transmission using the metaphor of Kanizsa's Triangle.

In Fig.6.1. the invisible triangle is tacit knowledge, and the surrounding symbols are a dialogue document. This tacit knowledge representation approach is different from Nonaka's idea, because Nonaka makes the conversion from tacit knowledge to explicit knowledge. However, my approach is not to draw the triangle directly, the triangle is organized in your mind by your active formulation. I might say that this approach is a means of conveying tacit knowledge directly.

My proposal for the dialogue document is based on a simple, interdisciplinary study, my everyday experience, and a conceptual analysis rather than on systematic empirical studies. My goal in this proposal is to introduce the concept of dialogue document, and present a plausibility argument. Of course a conclusive test of my notion will require empirical results as well as further analytical work. As a next step in this study, I am presently designing and plan to test a dialogue document-editing tool (DiET) which will be make possible such empirical work.

Acknowledgements

The ideas expressed in this paper have benefited from extended conversations with JoAnne Yates, Kaoru Ueda, and various other colleagues at the Center for Coordination Science, Massachusetts Institute of Technology. Special thanks also to Tom Malone, John Quimby, and George Wyner for their insightful comments on a draft of this paper, and to the Fuji Xerox Co., Ltd. for the financial support that made possible the research on which this paper is based.

References

Alexander, C., Ishikawa, S. and Silverstein, M. A Pattern Language, Oxford University Press, 1977, p. x

Bakhtin, M.M., Forms of time and chronotope in the novel. In M. Holquist (Ed.) The Dialogic Imagination by M. M. Bakhtin, University of Texas Press, 1981, p. 134

Conklin, J. and Begeman, M. gIBIS: A hypertext tool for exploring policy discussion. In Proceedings of the Second Conference on Computer Supported Cooperative Work, ACM, 1988, pp. 140-152.

Daft, R. L. and Lengel, R. H., Organizational Information Requirements, Media Richness and Structural Design, Management Science Vol. 32, No. 5, 1986, pp. 554-571

Dertouzos, M. L., What Will Be: How the New World of Information Will Change Our Lives, HarperCollins, 1997, pp. 96-97

Drucker, P. F., Post-Capitalist Society, Harpar Business, 1993, p. 8

Ei-Shinnawy, M. M. and Markus, M. L., Media Richness Theory and New Electronic Communication Media : A Study of Voice Mail and Electronic Mail,' Proceedings of the 13th International Conference on Information Systems, 1992, pp. 91-105

Fulk, J., and Steinfield, C., A social influence model of technology use. In Fulk, J. and Steinfield, C. (Eds.) Organizations and Communication Technology, Sage, 1990, pp. 117-142

Garfinkel, H., Studies in Ethnomethodology, Prentice-Hall, Inc., 1967, Chap.3

Giddens, A., Central Problems in Social Theory, London, Fontana, 1979

Giddens, A., Sociology: Third Edition, Polity Press, 1997,

Goffman, E., The presentation of self in everyday life, Anchor Books, Doubleday, 1959

Kanizsa, G., Organization in Vision: Essays on Gestalt Perception, 1979, Fig.3.2

Malone, T. W. and Crownston, K., The interdisciplinary study of coordination, ACM Computing Surveys. 1994

Markus, M. I., Toward a critical mass theory of interactive media: Universal access, interdependence and diffusion. In Fulk, J. and Steinfield, C. (Eds.) Organizations and Communication Technology (). Sage, 1990, pp. 194-218

Michailidis, A. and Rada, R., A Review of Collaborative Authoring Tools. In Rada, R.(ed.) Groupware and Authoring, , Academic Press, 1996, pp. 9-43

Nonaka, I. and Takeuchi, H., The Knowledge-Creating Company.,Oxford University Press, 1995

Polanyi, M., The Tacit Dimension, Gloucester, Mass., Peter Smith, 1966, p. 6

Prince, G., Narratology: The Form and Functioning of Narrative, Walter de Gruyter & Co., 1982

Sacks, H., Hotrodder: A Revolutionary Category. In Psathas, G. (ed.) Everyday Language: Studies in Ethnomethodology, Irvington Publisher, 1979, pp. 23-53.

Schank, R. C., Tell Me A Story: Narrative and Intelligence, Morthwestern University Press Evanston, 1990, p. 148

Schrage, M., Shared Minds, Random House, 1990, p. 111,

Shepherd, A., Mayer, N. and Kuchinsky, A., Strudel - An Extensible Electronic Conversation Toolkit in CSCW 90 Proceedings, ACM, October 1990, pp. 93-03

Toffler, A., Powershift, Bantam Books.1990, p. 88

Weick, K. E., Sources of Order in Underorganized Systems: Themes in Recent Organizational Theory. In Lincoln Y. S. (Ed.), Organizational Theory and Inquiry, Sage publications, Inc., 1985, Chap.4 pp. 110-111

Winograd, T. and Flores, F., Understanding computers and cognition: A new foundation for design, Norwood, NJ: Sblex., 1986

Winograd, T., A Language / Action Perspective on the Design of Cooperative Work, Lawrence Erbaum Associates, 1988

Yates J., Control through Communication: The Rise of System in American Management, Johns Hopkins University Press, July 1989, pp. 274-275

Yates, J. Using Giddens' Structuration Theory to Inform Business History, MIT Sloan School of Management Working Paper, 1997

Zuboff S., In the Age of the Smart Machine : The Future of Work and Power, BasicBooks, 1988, pp. 315-361