Chapter 3 -- The Research Web (3a)

The Research Web (RW) is a social, intellectual, and technological structure devoted to collaborative study of a single-issue domain. The research team uses a set of WWW-based programs designed as a basic tool set for collaboration within the prototypical research project: a widely dispersed, large-scale, long-term scientific collaborative research project. The selection of the WWW as a vehicle for collaboration follows substantial projects of large corporations, for instance NYNEX (Girgensohn, Lee, and Schlueter 1996, 246). Since the universal characteristic is WWW compatibility, additional tools can be added at will to suit the needs of the individual Research Web. The basic tool set of the RW consists of: augmented hypertextual essays, the Research Web Essays; an integrated bibliographic information service, the Annotated HyperBibliography; and an integrated lexicographic tool, the Annotated HyperGlossary. These three productions, plus the tools that support them give the team a means to develop knowledge that is available at a click, and is all annotatable.

The work in this dissertation was begun very early in the history of the WWW and the tools can expect to be eclipsed by new products in time. An example of services that are being created is the CrossRef initiative, designed to facilitate online interpublisher linking of article references to their full texts (Anon 2000) . As a creation of a consortium of publishers, CrossRef is designed to augment their own online journals, not private creations. When mature and largely freed of excessive commercial interest, the tools developed by this initiative may migrate to the scholarly community; and can be expected to replace the HyperBibliography, but the annotation feature will likely never be incorporated into CrossRef tools. Keeping the team's criticism private and accessible will likely be the one of the last features added in large-scale development for collaborative software. The RW's critical apparatus is likely to be the lasting legacy of this work.

3.1 The Concept

The Research Web (RW) is the central concept in this research. The RW can present information and conclusions in a way that cannot be done in conventional literature. A Research Web is the electronic embodiment of the intellectual capital of the network of excellence which develops about an issue domain, the phenomenon being investigated. It is both a social organization and a WWW site that disseminates information, provides communication facilities and an infrastructure for collaborative interaction. A RW can be viewed as a domain-specific information repository (see §3.2.2) and a network of communication channels connecting the collaborators and perhaps sponsors, stakeholders and interested members of the public. This view is compatible with the model of science as a distributed artificial intelligence network. Thagard (Thagard 1997), (Thagard 1993) considers scientists to be nodes in a network connected by communication links. Other links connect the nodes with information repositories (research articles, journals, libraries, web sites). The entire scientific enterprise can be seen as a dense network of scientists (active nodes) and recorded knowledge (passive nodes). The links between the nodes of this network are directional, dynamic and of variable strength. The membership can vary as interests wax and wane or death intervenes, but the cyber-place remains with its knowledge base intact and growing. Ideas, hypotheses, findings, discussions and publications are spun out of the Research Web as portions of the issue domain are transformed into well-structured problems.

A RW network of collaborators adds a layer of specialized interconnection to the existing scientific network. The existing 'web of science' is largely composed of links that are socially frail, weak links (Pickering and King 1992, 357), consisting of awareness and occasional short-term communication. Strong links, links of deep collaboration and friendship, are usually restricted to a small set of single paper collaborators, or to spatially collocated scholars at the same University or department. A RW strengthens existing communication links, usually weak links, or creates direct linkages, strong links, where one existed before. The tools associated with the RW promote interaction and facilitate additions and refinements to the knowledge repository of the RW web site. Since the RW is freely available to the research team, it also strengthens the scientific knowledge and intellectual power of the nodes, the individual scientists. It has been found that once a strong tie exists, it can be sustained through computer-mediated communication (Koku, Nazer and Wellman 2000).

Recognition of a problem creates a need to define the system containing the problem, the issue domain. The issue domain is the body of knowledge, given and earned, together with landmarks and possible gaps, surrounding the problem that initiated the research effort. The goal of the RW is to develop comprehensive knowledge about the issue domain. The issue domain has physical, spatial, temporal and cultural components. Knowledge about the issue domain necessarily must precede problem understanding, structuring, and solution. In order to understand the issue itself, and to provide knowledge sufficient to solve the problem, significant research and exposition is necessary. The exposition takes the form of essays that are placed in an information repository. The essays may initially be taken from position statements, but will evolve into canonical documents presenting the knowledge of the research team and their intellectual support. The RW will have a library of literature representing new knowledge and representations of existing literature, bibliographies of relevant literature with annotations on that literature, databases, and models of the issue domain and its components. The research web will augment the institutional memory of the collaboration by permanently recording essays by the participants, building models, recording the dialog, and maintaining a corporate bibliography and glossary.

The character of the collaborative environment created by the RW is marked by opportunity to access all research materials, and to interact with most research material through critical annotation. Access and opportunity to interact are available anytime from any modestly equipped workstation. All dialog except private e-mail is recorded and indexed, as it almost always adds value by increasing the depth of knowledge. Informal communication among the team members is enabled by e-mail and a team listserver. The listserver is a mailing list with a searchable archive. Tools suitable for these components are widely available. Though the mode of interaction is primarily asynchronous, there are provisions to capture synchronous meetings and dialog in the form of meeting minutes and telephone conversation records. Capture of the synchronous record is the weakest part of the system as it depends on members taking the time to transcribe dialog from speech to text.

3.1.1 Defining the Issue Domain

All inquiries must have a beginning. The thought is articulated by a champion and then elaborated by a core group of researchers who come together to create a proposal. For them the proposal has two purposes, first to gather researchers to support the effort, and second to attract other scholars to the research effort. This core, the conveners, establishes, legitimizes, and guides the collaborative alliance (Wood and Gray 1991). Here, proposal means: a document that explains what the topic of the research is, the scope, purposes, and organization.

Such a proposal may not be a formal proposal submitted to a funding agency, but may be a manifesto presented to professional colleagues. The purpose of this proposal is to gather support. The support needed is the promise of collaboration from colleagues and the promise of cooperation from organizations that can provide resources such as money, labor, and facilities. Grant proposals to fund discrete detailed research efforts may be coordinated by the conveners or an executive committee. Such grant proposals must include some support for the RW, perhaps as management overhead.

Some proposals and grants assemble rather loosely associated groups together under an umbrella title under the assumption that each group's research will inform the others. In practice such taxonomic collectives (Sayer, 1992, 250) (united by classification only) almost never come together in meaningful collaboration. The Consortium for Risk Evaluation with Stakeholder Participation (CRESP) is an example of this type of organization. This large organization, funded by the Department of Energy, has operated from 1994 to the present. Originally, it established eight separate task groups that were interdisciplinary and distributed in space. While each of the task groups produced scholarly products, the mandated collaboration between groups was almost totally absent. Such groups need to establish a Web site for the umbrella organization, which may be responsible for generating reports and maintaining a public WWW presence. This web site can refer to Research Webs established for each of the substantive groups. Each group should have an independent RW with its own issue domain.

Like any human artifact, the issue domain must be designed. There are two critical aspects of this design: circumscription and conceptualization. Circumscription defines the scope of the Research Web. If the scope is too large, then the efforts will be too diffuse to gather a critical mass of researchers around related projects (Poole 1994, 23). If too small, then the intellectual content will soon be exhausted. It is best to err on the side of a smaller than optimum scope, as the scope can easily be enlarged.

The process of issue domain circumscription is realized in modeling as a process of simplification and resolution, making the usually fuzzy boundary better defined. The process of issue domain circumscription is realized in modeling as a process of simplification and resolution, making the usually fuzzy boundary better defined. Simplification is a process of distinguishing parts of the topic, defining them and their relationship to the core domain. If the distinguished parts are not essential to the core issue domain, then they may be severed from the issue domain (Sterman 1991, 5). This severance defines a portion of the boundary of the issue domain. The organizing principle may be further clarified by the production of a context diagram that shows the boundaries of the issue domain. Logically, this operation is the differentiation of types within a supertype (Aronson, Harrè and Way 1995, 45). The process of simplification does not include the dismissal of minor contributions to complexity; it consists of a careful paring away of entire regions of investigation that lie outside to the issue domain.

Conceptualization of the issue domain is the process of identifying topics within the issue domain boundary. The property of interdependency must be maintained so the team collaborates on the basis of clear relationships rather than vague associations. Management may coordinate these tasks to maintain interdependence as portions of the domain become known territory. New topics may be added as objects and processes are discovered, provided they meet the requirement of interdependence. Topics that expand on previously identified topics are clearly related in a "drilling down" descent into detail.

The design of the issue domain will be revised as research progresses. Finding and describing the fuzzy boundaries of the issue domain is an ongoing exercise carried out in several ways: arguments carried on in regard to the qualitative models can carry a philosophical component; the simulation model can be used for probing boundaries by examining the sensitivity of the model to variables; and entire areas of issue domains may be severed and abandoned, or may be assigned to RWs working on associated domains.

3.1.2 Determining the Audiences

Above all, the principal audience is the research team. The RW Essays are designed to become canonical documents and are addressed to scholars familiar with the . Works that derive from the essays will use language and arguments that are designed for a different audience. The granting agency and cooperating institutions might be offered the ability to audit the progress of the RW by viewing it, or through periodic progress reports. Sustaining grants may be more likely if the granting agency has access to the RW as well as obligatory reports and published papers. The RW may have a partition that allows public access to some information and to documents specially designed for public use. A similar partition may be set up to allow temporary access to professional colleagues outside the team. Parts of the RW may also be used in academic instruction.

3.1.3 Defining the Vocabulary

Identification of the objects and processes in the issue domain leads to the description and formalization of the language of discourse for the domain. In the context of a multidisciplinary issue domain words may take on multiple meanings (Chen 1994). Rachel and Woolgar report an extraordinary constellation of meanings and implications for the term 'technical' (Rachel and Woolgar 1995). The author once sat through three hours of meetings devoted to defining the meaning of 'hazard.' Meanings of terms need to be articulated by collaboratively developing a glossary. The glossary must be a dynamic document that can be annotated at will by members of the research team; thus the glossary becomes an Annotated HyperGlossary. Glossary entries can be referenced in any RW essay, or other HTML document, and displayed in popup windows without leaving the essay.

Interactive software is necessary to develop the glossary and to display the information assembled. The program designed to assist in the assembly of the glossary is called Lexicon (see §4.8). A geographic analog of the glossary, an Annotated HyperGazetteer, could provide the important ability to define describe and display regions, features and linear objects of interest in the issue domain. These features might be dynamically displayed on maps from the gazetteer, or be referenced from maps to the gazetteer.

3.2 Five Aspects of a Research Web

The Research Web is designed to be a vehicle for research collaboration, information dissemination, and model building. Five prominent aspects of the Research Web have emerged. These aspects serve to validate the design and explain how the RW works.

3.2.1 The Research Web as an Application of Critical Social Theory

The Frankfurt School developed critical social theory, or critical theory, in the mid 20th century. The principal thinkers associated with the School are Adorno, Horkheimer, Marcuse, Fromm, Benjamin, and later Habermas. The primary goal of critical social theory is to find "alternatives to existing social conditions which more adequately address human desires." (Ngwenyama 1991, 268). The social condition we seek to find an alternative for is the current environment for development of scientific knowledge and its publication. We can apply the basic principles of the Theory and then appeal to Habermas' Ideal Speech Situation to show how the Research Web meshes with this tradition.

The Frankfurt School proposed a revision to Marxism that put forth five basic assumptions (ibid. 269). Here is the list of assumptions and how the RW works with them:

3.2.2 The Research Web as an Information Product

The Research Web (RW) is an information product, software that manages data, information and knowledge. An information product is a complex of hardware, software, procedures and data that is designed to serve a particular purpose for a set of audiences. The purpose of the RW is to provide an environment to investigate the nature of a phenomenon, the issue domain, using scholarly processes. The audiences are: the research team composed of scholars, collaborators, and research assistants; stakeholders, the people and organizations who are affected by the research; the sponsors of the research, those organizations and individuals who provide the resources for the research; and finally, perhaps, the public, composed of students and interested parties who may benefit from learning about the phenomenon under investigation.

A RW closely follows the architecture of information products defined by Meyer and Zack (Meyer and Zack 1996) (See Figure I, below). It contains three major abstract divisions: a product family, documents presented on demand to the information consumers; a repository, the information store for the RW; and a refinery, which contains the means (software and management processes) to manage and add value to the information in the repository. The repository is a database, whose architecture can assume a structure most suited to produce the product family. The product family consists of several document types carefully designed to serve the needs of the research team. The refinery is the heart of the RW. It serves five major functions: acquisition, the accumulation of information about the issue domain; refinement, the processes used to add value and quality to the information; storage and retrieval, the means to manipulate the repository's data; distribution, the means of taking the products to the information consumer; and presentation, the means of displaying the information to the consumer.

Given the basic assumption that the RW exists in an environment established by the Internet and WWW, some of the processes outlined by Meyer and Zack fall out or are modified (see Figure II, below). The Internet manages the distribution process. That assignment of responsibility constrains the information system (RW) to limit its information to files structures compatible with the WWW. That restriction of file types also limits the product family to WWW compatible file types and their printed representations. The user must be equipped with a modern personal computer and an Internet connection. The Internet browser wholly manages the presentation process. That assignment forces the user to assume the responsibility of installing a browser on his or her personal machine. The user must also acquire the training required to use the browser and personal machine. The storage and retrieval process becomes a distributed system, with files stored in any server machine connected to the Internet.

The process platform of the RW architecture is thus simplified to acquisition process, criticism process, and editing process. The acquisition process and editing process are human-centric processes managed by the author, scientific leader, and facilitator. The criticism process is managed by the user and by software on the server machine. The criticism process and editing process together constitute the refinement process of the Meyer and Zack architecture. The product platform is simplified to the repository and a combined distribution and presentation process managed by the browser under the direction of the user through an interactive screen display.

Abstracting from the Meyer and Zack architecture, it can be seen that the RW is an Information Product constructed to support a knowledge transformation process (see Figure III above). The research team can, from two types of input (Information Sources, and Tacit Knowledge) produce the ultimate output (Information Products) through three intermediates: Identified Knowledge, New Local Knowledge and the Information Products of the RW. Four interlinked processes support this transformation:

In order to design a good set of tools to facilitate collaboration, great attention must be given to the genres of communication currently in use by similar organizations. Each established form of communication is a genre (Orlikowski and Yates 1994), (Agre 1998), (Bazerman 1988). These genres must be compatible and mutually supportive in order to be successful. Such a group of genres form a genre system (Orlikowski and Yates 1998) (see Table I, below).

Our genre system is composed of genres developed for general use in the RW environment, and specialized genres appropriated by the research team. The genres may be adapted to the peculiar demands of the issue domain. It is possible that a genre might be designed specifically for an issue domain-specific purpose, for instance a species list for a biological or ecological issue domain.

Table I The Genre System of the Research Web

GenreName	Function	Description	Reference
Models	Provides a synoptic overview of a process or set of objects. May serve as a clickable index to discussions of its constituent parts or submodels.	A tabular presentation of the characteristics of objects and how the characteristics may be operationalized. Or a graphic presentation usually consisting of nodes and links (bubbles and arrows) showing the relationship of its components. Or a computer simulation of the system or a portion of the system.	§3.2.4.2.1 §3.3 §3.5.2.4.1
RW Essay	a treatise, usually scholarly	A highly augmented annotatable HTML document with hypertextual links to bibliographic information, definitions, notes, and related documents.	§3.4 §3.5.2.4.2
DocReview	a critical apparatus	A suite of programs that allows the user to make annotations to predefined segments of text and graphics, and allows those comments to be read by any user.	§4.3 §3.4.2
Annotated Hyper-Bibliography	provides bibliographic information	A hybrid software system that allows the user to input data in a PBM (personal bibliographic manager) then utilize that data in an annotated bibliography and in popup windows in RW essays. Supports user annotation.	§4.4 §3.4.4.1
Annotated Hyper-Glossary	provides definitions and technical commentary	Definitions appear on a web page that allows the user to comment or add a gloss to the definition.	§4.5 §3.4.4.1
Discussions	provides a venue for threaded discussions of a topic	WWW site that allows annotation of an initial topic statement, and allows annotations of the annotations indefinitely.	§3.5.2.2.4 §4.2.5
FAQ	answers commonly asked questions	A WWW page listing frequently asked questions with answers.	§3.2.4.3
Personal Home Page	introduces a team member to peers	A WWW page that provides links to interests, CV, and publications.	§3.2.4.4
Topical Home Page	introduces a major topic in the web site	A WWW page that briefly describes a topic and provides links to more information.	§3.2.4.4
What's New	informs the members of revisions or additions	A WWW page that allows the user to click directly to new or revised documents.	§4.7
Listserver	allows members to message the entire team	E-mail lists for general discussion, contains an archive of past messages.	§4.1.3
Calendar	scheduling	A page that lists scheduled events.	§4.2.4

Each genre selected or designed must also be modified to conform to the demands of the electronic environment. Hypertext alone will force numerous adaptations. International teams will need to consider cultural preferences and taboos.

3.2.4.1 The Essay Genre

Agre says that the designer of a new genre needs to work with the existing forms used in the activity, but then do more (Agre 1998). The Research web essay does more by placing the essay's annotation in context: there is no need to thumb through dictionaries and bibliographies, or to flip to a section of endnotes. All that information may be had at a click.

Scholarly annotation includes a wide variety of supplemental information: information on sources, clarification of obscure statements, references to related works, alternative meanings, definitions, glosses, editorial notes, etc. Over the centuries of representation of knowledge in codex form rules have evolved for the treatment of such information, with occasional revisions due to minor changes in technology, or changing fashion. With the introduction of hypertext and the WWW, those rules can now be challenged. A web page is not a codex document; it is a scroll. With the power of hypertext and the ability of the web browsers to create new pages, the rules can be reinterpreted to great advantage. In particular the editorial rule of "clean text," the minimization of interruption of the reader's view of the primary text, can be applied even more stringently. The RW Essay is a genre inviting application of new rules for annotation.

Every genre is a template for social action (Orlikowski and Yates 1994, 542); in this case the social action is the transmission of knowledge. The RW Essay gives the reader the ability to criticize the content of the essay, the language used, and the quality of the references. In other words the RW Essay not only informs, but also contains its own critical apparatus. Criticism is a major method of creating knowledge, and the best way to provide a research team with the means to participate in that creation. In the RW, commentary is a form of scholarly dialog. Commentary in the RW Essay is clearly secondary text, the essay's content is not altered until the author or authoring team chooses to incorporate or disregard the commentary in a new edition.

The Research Web Essay is a very highly augmented Web page. Like any other web page (a very well known genre), it reacts to clicks on hot spots on the page. The clicks pop up new windows displaying well-established features of scholarly literature (Cronin et.al. 1998, 1319). Features implemented include marginal notes, sidebars (see alsoes), bibliographic information (including full text if available), glossary definitions (and discussion), a link to a means to annotate the essay (DocReview), and general notes (analogous to sticky notes).

These features are available to those that use the computer to read the material. Footnotes and references both have return links to the text. If the reader is curious as to where in the text an author is cited, all that needs to be done is to click on a return link icon (there is one for each time the work is cited). As a service to those who wish to have hard copy of the essay, footnotes, glossary of terms, and references for the essay are printed following the text of the essay. The printed essay is of course badly disabled by loss of the hypertext links.

The RW, in order to present a theory about the phenomenon under investigation must represent all the expressed models of the issue domain, and should discuss the unexpressed mental models as well. Mental models are the filters by which we interpret the reality we perceive (Sterman 1991). Our mental models are constructed on observations and learning which lead to a set of hypothetical generative mechanisms (Harré 1978, 275) that account for our experiences.

Mental models are naturally unexpressed, and therefore become expressed as their description takes shape. I cannot imagine a description of a mental model being anything other than an evocative work of prose or art that shows the coherent nature of a set of ideas and their interactions. The closest formal genre that might fit would be an essay. This essay is bound to be intensely personal and full of conjecture and very weakly supported theory. It would very likely need to be kept private, as part of a notebook or a journal. An edited version would be very helpful as a basis for beginning the collaboration. The edited version would be a position paper, perhaps published in the RW on the team's biosketch page.

Since all but the smallest, or atomic, models contain submodels (Ziegler 1990, 29), larger models such as the descriptive, explanatory, and simulation models must be hierarchical. Those hierarchical structures are naturally expressible in hypertext documents. The models are process models, models of objects known as type or class definitions, and diagrams showing the relationships between objects (entity-relationship diagrams). All these models can be backed with dictionaries and catalogs.

The descriptive model forms a very large portion of the RW. It will contain a series of essays, organized hierarchically, that captures the nature of the phenomenon. As the essays are refined through criticism, they will uncover more detail giving the description breadth and depth. Each essay will produce two products: models and essays of details uncovered in the parent essay. Models take many forms, usually diagrammatic, but often textual. Submodels developed in the descriptive model are: process models that describe the temporal progress of actions; type descriptions, that list and describe the objects and processes that the objects participate in; entity-relationship diagrams that describe how the objects are related; and catalogs and dictionaries giving definitions of the terms used in all the models. In realist terms, each essay will deal with objects at its level and the emergent qualities of that level. Models are the glue that holds the essays together.

The explanatory model will be hierarchical and will resemble the descriptive model in form. The nature of the essays and models will be quite different, as they will deal with hypothetical causes and must describe the basis for experiments to support or reject the hypotheses. The essays produced for the explanatory model will be much more abstract than those of the descriptive model. The explanatory model must go beyond the description of the phenomenon in operation to the reasons why the phenomenon operates as it does. Causal models require the identification and elucidation of the mechanisms that presumably cause the observed behavior of the phenomenon, its objects and processes. The explanatory model expresses both hypotheses and theory. Those hypotheses will be examined in experiments, and the theory will be expressed in the simulation model.

The simulation model is ultimately represented by a computer program. That representation is, by itself, inadequate because of the incomprehensible detail and rapidly changing content and computational recipes (algorithms). The ruling representation for the simulation model is the description of the design. The design description captures the scientific intent of the model, leaving its implementation to the technicians. The design description is a model, usually graphic, that are defined as finite state machines (Hendricksen 1989), or petri nets (Peterson 1977), (Benwell 1991).

3.2.4.2.1 Representation of the Models

Models have been inaccessible to many social scientists due to poor modeling practices and the cloaking of explanatory and simulation modeling in mathematical notation. The mathematical notation frequently alienates those inclined to be a bit skeptical about the applicability of mathematics to the human issues in our lives. The RW makes modeling accessible to human geographers, sociologists and other social scientists by a restructuring and expansion of the usual modeling processes. The Research Web is designed to make knowledge as models easy to accumulate and access.

The Source Model is an unexpressed but very important model that cannot be represented symbolically. There are sentential models (Harré 1970) that form part of the bridge from the Source Model to the Descriptive Model and to the Explanatory Model. Sentential models are often simply collections of textually expressed knowledge. The RW Essay may be considered a sentential model. The Simulation Model is expressed in a computer language algorithm that is inaccessible to non-specialists, but it is modeled in a variety of expressed models.

The Descriptive Model is a suite of models of the objects of the issue domain, and models of the processes that relate the objects within the issue domain and to the rest of the world. The models of objects are maps of their attributes and the top-down hierarchies of related objects (Aronson, Way and Harré 1995, 36). Our descriptive models are composed of lists of objects, their attributes, and the values that their attributes can assume, including much of the information found in the auxiliary model (see Table II, below). For instance, operationalization information can be added to each attribute and process; for instance an attribute such as a person's age can be operationalized in several different ways: by date of birth, by common designation (years), by developmental age (zygote, fetus, postmenopausal ...), by life stage ( infant, child adolescent ...), and others. Each of the operationalizations has characteristics such as data type (integer, real ...), value range, precision, and of course a description. Processes may include several alternatives.

Object models are usually presented in tabular format. For each characteristic of the object (attribute) there is a row; and for each property of the attribute there is a column. In the RW, tables are presented in web pages, and each cell in the table may be annotated by using DocReview. Object models may be extended to include operationalizing details such as how measurements are made and with which protocol.

Process models are usually presented as node-and-link diagrams. In the RW, the diagram may be image-mapped so that each element, node or link, can be annotated. Clicking an element will display a document that fully describes the element. In complex processes, a node may be expanded into a sub-model. Clicking such a node can link directly to the sub-model, or the submodel can be part of a document. The documents referred to in a process model are annotatable in DocReview.

Relationship models are presented as node-and-link diagrams. In the RW, the diagram is image-mapped. Clicking a node, which is always an object, refers the reader to the object models for that object. Clicking a link will display a document that describes the link. These documents are annotatable in DocReview.

Physical Body

Attributes	Operation- alization	Description	Data Type	Values/Range	Precision	Notes
Age	Common	The number of units of time that have passed since birth.	Integer	Units: weeks, Months,Years	minus 0 to plus 1 Unit	In Japan, people are assumed to be one year old at birth.
	DOB	The date of birth.	Abstract: Year of birth; month of birth; day of birth.	month: 1-12; day: 1-31	one day	In Moslem culture, years date from the Hegira (AD 622).
	Developmental	Stages of life marked by physical milestones and characteristics.	Nominal	Zygote, blastula, embryo, fetus, neonate, infant, child, adolescent, adult	fuzzy
	Gestational	The age of a fetus since presumed conception	Integer	weeks	plus or minus 2 weeks	There are many clinical proced- ures often used in combination.
	Life Stage	Stages of a model life in a given culture.	Nominal	Infant, child, adolescent, student, married, retired, senescent	fuzzy	Could be elab- orated with any milestone, or rite of passage (e.g. bar mitzva, confirmation, divorce, promotion).
Sex	Reported	The sex that the person assumes in life.	Binary	Male, Female
	Preferential	The sexual roles that a person assumes.	Nominal	Female, Male, Lesbian, Gay, Bisexual, Asexual, Chaste
	Chromosomal	The makeup of the sex chromosomes.	Nominal	XX, XY, XXY, XXXY
Object Model	Auxiliary Model

The process models of the RW are almost always directed graphs, or node and link models, backed with RW Essays describing the operation of the process. These models generally have text labels on nodes, which may be circles or boxes, and links that are lines connecting the nodes. The lines representing the links may be augmented with arrowheads or other devices to indicate the directional character of the link. Mathematical models can be qualitatively transformed into graphs for better understanding: input, process, and output.

The hypertextual nature of the RW allows us to use each model as an index to its component parts, both nodes and links. The models are realized as "image maps." In an image map, each element, node or link, is invisibly surrounded with a "hot spot" area that may be clicked on the computer display. When the element is clicked information about the element is "popped up" in a new window. That new window may contain a submodel, or a textual explanation of the element. It should be noted that the label associated with the element is simply a sign, and the element has to be clicked to obtain the information about the element.

Hierarchical hypertextual models have the monumental advantage of allowing the reader to request explanation of any element of a model in increasing detail, perhaps even until the element is irreducible. This ability to "drill down" allows exploration of just those parts of the model that interest the reader. This type of modeling has been employed for several years in maps as "hypermaps" (Milleret-Raffort 1995, 208).

So, we have foundation models expressed in simple graphics and plain natural language. This organization allows the presentation of the model to be quite simple, since the labels simply suggest the information about the element. The brevity also allows extremely complex models to be reduced to an absolute minimum of complexity. Burying the information about a sign in a subsidiary document is called information hiding. One great advantage to information hiding is that, for the novice, the signs become links to knowledge; and for the expert, the signs become knowledge.

3.2.4.2.2 The Process of Modeling in the Research Web

All modeling in the RW starts with an exhaustive analysis of the objects and processes operating within the issue domain. This analysis is referred to as systems analysis, and it is the basis for all modeling that follows. System analysis is not only accessible to the social scientist, but in fact is their strength. Any scientist who has described a social phenomenon, from migration to learning, has engaged in system analysis. The system analysis performed by the social scientists resides in the repositories we call the descriptive and explanatory models.

Description of prototypes, description of the data, and description of behavior is the heart of social science. The team may start by describing the objects and processes captured in the Annotated HyperGlossary. Objects are described in annotatable tables with hyperlinks to associated essays and other informative documents. Empirically observed processes are described in node and link diagrams that may be incorporated in RW essays.

Explanation is theory-building based on description and probable causal mechanisms connected through abduction. The explanatory model will include abstract objects and processes based on the real objects and processes of the descriptive model. As causal hypotheses are developed, they can be assembled in the explanatory model. The explanatory model will be the source of experimental hypotheses that will be examined in the later stages of work. The experiments of the past and those performed in the course of the research provide inductive proof of the theory.

When the system analysis and explanatory modeling has reached sufficient maturity or detail, then simulation modeling can start. It is at this point that operationalization takes place. Operationalization is normally considered part of experimental design, but it is also a set of instructions for simulation modelers. The descriptive and explanatory models through operationalization form the auxiliary model, and the auxiliary model informs the experiment and the simulation model. Social scientists who choose to involve themselves with simulation modeling and computer programs may certainly do so, but can also defer to computer scientists or hired programmers.

Whoever does the simulation modeling must be held responsible for firmly linking the programming with the descriptive, explanatory and auxiliary models. Programmers are well known for disdaining commentary linking their procedural code to the reasoning of the analysts who have designed the theory. The team must not allow failure to maintain the chain of reasoning to degrade the modeling effort. If the simulation modelers are puzzled by the foundation models, then they must bring this to the attention of the social scientists, for such puzzles are feedback pointing to weakness in the foundation models.

Throughout the modeling process, new knowledge and criticism will surface. This new material will require the expansion and revision of the models and essays that formed the basis for the new material. Change will constantly ripple through the models and essays, requiring a steady stream of revision. Management of this process will determine the quality of the work of the team.

3.2.4.3 The FAQ Genre

The FAQ (Frequently Asked Questions) was created early in the history of the Internet, probably as a defense from repetitious questions from newcomers to open discussion groups. Defense from the naive was the true motive, but service to the learner is a fortuitous and more important outgrowth. It is the nature of members of a group to bully newcomers who have naive questions, so not only was the discussion burdened with the questions, but also from many rude responses from the group members. Customs have now developed to control this behavior: members now send private e-mail messages to the "offender" suggesting how to get to the FAQ.

The FAQ begins as a compendium of naive questions and answers to them, but soon actually becomes an index to succinct topics of group knowledge covering the entire nonspecialized issue domain (Crowston and Williams 2000). FAQs are the place to go if one needs a quick and correct answer on a topic. There are even Indexes to FAQs or "FAQ on FAQs").

The fully developed genre organizes the information in a very carefully designed hierarchical index, perhaps headed with a "top ten" set of questions. Each section of the FAQ then simply contains a set of questions with answers. There may be several answers to a given question and there may be hypertextual links to more complete discussions of the question. The FAQ frequently contains a search engine so the user can rapidly view the knowledge relating to a set of keywords. The FAQ usually contains an introduction to the community of interest, including listserves, appropriate canonical documents and URL references.

In an interdisciplinary Research Web, the FAQ might contain background knowledge from each of the specialties. Many of the questions regarding terminology will be discussed in the Annotated HyperGlossary, and principal references in the Annotated HyperBibliography. Since the RW's working area is usually closed to the public, the FAQ is often not required there, but if there is a large public partition, a FAQ can be a real service.

3.2.4.4 The Home Page Genre

The home page is an index page (Furuta and Marshall 1996, 185) for a project (topical home page or THP), or the personal home page (PHP), a narrative of self-evaluation (Roberts 1998, 85), of the account owner. The home page is a new genre created as an outcome of technology and may be considered the first truly digital genre (Dillon and Gushrowski 2000). It came into existence with the WWW, and provides a means to navigate the structure of the web site. Every file system needs a "root" directory containing any number of files and subdirectories. Located at the root, the home page will provide the information necessary to navigate the web site.

Any team has the need to know each of the members. Social as well as informative purposes are served by the personal home page. Personal interests and expansive descriptions of research interests provide a window to view the personality of the subject. Dillon and Gushrowski (Dillon and Gushrowski 2000, 204) found that at least half of their experimental subjects agreed on each of ten features that should appear on the PHP: a title, an e-mail address, an update date, a table of contents, a create date, external hypertext links (perhaps research project pages), a welcome message, from one to four graphics, photographs, and a brief bio (for academics, a CV or a link to a full CV). As time passes, the PHP has become an increasingly stable genre. The personal home pages of the team members must be linked from an index page (see §3.5.2.2.1). The index page (a topical home page, see below) might include a brief description of the position that each person fills on the team. The PHP should be linked by URL, as it can be stored on any WWW server. The facilitator can assist any member in preparation of a PHP.

The home page for the RW, a THP (Crowston and Williams 2000, 208), is basically an index to the other major topical pages in the site. The content of this page is devoted to an introduction to the issue domain of the team's research. As a navigation device, this page should provide one click access to major sections of the site, such as the index to Research Web Essays, and to frequently used pages such as "What's New" and the MailRoom. The site needs separate THPs for each partition implemented: public, the working area, and the guest partition. A topical home page has not only links to subtopics, but also introductions to them. The topical home page can introduce and subject in the logical hierarchy of the site. In a strongly hierarchical issue domain, the connections of the topical home pages may be shown in a diagram, the organizing model, with each link and node in the diagram or model serving as a clickable link to other pages. These "clickable images" are a powerful navigation tool.

3.2.4.5 The Meeting Minutes Genre

Groups will find it necessary to meet synchronously from time to time. In these meetings, members discuss important issues and make decisions that may affect other members. Minutes of these meetings need to be recorded and opened to annotation. The attendees of the meeting need to see that the minutes are both complete and correct. The minutes then can be archived and become a searchable part of the team's research record.

In the RW, meeting minutes have evolved into an interactive genre that to date has been a very successful application of this research. After the scribe renders the minutes in an electronic format, the minutes are forwarded to the facilitator who mounts the minutes in DocReview (see §4.3). Each item in the minutes is made into a separate review segment. The facilitator then sends a notification of the posting of the DocReview to all attendees. The attendees can review the minutes item by item at their convenience any time well before the next meeting. Objections, corrections, elisions, and amplifications become DocReview comments, visible to all immediately.

Just before the next meeting, the meeting chair may print out the DocReview from the WWW in a format that interleaves comments after the review segment (the "on-the-bus" format). Copies are distributed to the attendees along with the agenda. The first item on the meeting agenda is the correction of the previous meeting's minutes. The attendees discuss the recommended substantive changes and the scribe notes the dispositions. The last item on the agenda is usually discussion of outstanding action items. The action items are tagged with the date assigned and are carried over until completed or removed. Like all items in the DocReview the action items may be annotated.

After the meeting, the facilitator will correct the old minutes, archive the DocReview (thus preserving the annotations), and post the minutes to the web site. The approved minutes may then be e-mailed to a distribution list. The archived minutes are then linked from an index page for all meeting minutes. The minutes are searchable documents.

All knowledge is socially constructed. This is especially true of social science. The body of knowledge in every discipline is the result of an ongoing multilogue between all scholars, present and past, who have contributed to the literature. The multilogue can be expressed as argumentation.

Realist philosophy recognizes that hypothetico-empirical proofs will seldom be found in social science due to the open system nature of human society. Rather than depending on certainty to establish validity, as is possible in closed systems sometimes seen in the natural sciences, validity in social science is established by other standards such as relevance, cogency, and truth (Dunn 1982) . All these standards may be judged by the principle of practical adequacy (Sayer 1992, 65), or by the degree to which the model of the system resembles the system itself, verisimilitude (Aronson, Harré and Way 1995, 12).

Practical adequacy and the measurement of verisimilitude are arrived at by a process of social interchange. This interchange, at present, takes place formally in the literature. The rules of interaction are designed to insure that scholars present rebuttals or confirmation in published papers. Classically, each paper must present a complete valid argument rather than an attack or confirmation of a single small point, though disagreements are frequently noted en passant in the review of literature. Discussion of minute points is seldom published because such communications do not meet the minimum publishable unit requirements. Argumentation on a single point or subsection of a published paper is carried on in often unindexed minutia such as Discussions, Technical Communications, and Letters to the Editor, or remains unpublished. The RW can have a much more intense and rapid social interchange of any item regardless of importance.

Scientific arguments, both regular (supportive) and critical, have a very clear structure (Toulmin, Rieke and Janik 1979, 249). Eisenhart and Borko propose that following Dunn and others, research, especially action research in education, can be considered a form of argumentation (Eisenhart and Borko 1993). Dunn suggests that knowledge and practice can, through a process of reasoned argument and debate, be successfully advanced (Dunn 1982). He uses the metaphor of jurisprudence as a process model. Gross also refers to courtroom argumentation (Gross 1990, 7). Representation of adversarial positions held in conflict over environmental impacts of chemical weapons disposal by argumentation has been demonstrated by Liebow, et.al. (Liebow, et.al. 1998). Both Dunn and Liebow suggest using the Toulmin method of argumentation. Toulmin's typology of the elements of argumentation includes claims, grounds, warrants, backing, qualifiers and possible rebuttals (Toulmin, Rieke, and Janik 1979, 25).

3.2.5.1 Argumentation Capabilities in the Research Web

Information contributed to the argument is collected as a byproduct of normal operations of the Research Web. Submission of e-mail, DocReview comments, HyperBibliography comments, HyperGlossary comments, discussion forums, meeting minutes, and above all RW Essays provide the elements of argumentation. A formal argumentation structure may be built by using a web-based tool, The Landscape of Reason, after which the structural layout may be directly manipulated. This tool presents the argument in a hierarchical hypertextual format that may be read by anyone on the WWW. Construction of the argument is an intellectual edifice built by organizing information extracted from the contributions of the team. Additions, deletions or revisions are to be managed by the facilitator, but may be opened for direct user manipulation in technologically facile teams.

In the RW Essay, each review segment may be critically annotated publicly thus directly contributing to not only content but to the valuation of adequacy. Such public argumentation approaches Habermas' Ideal Speech Situation much better than current practices of scientific peer review (Gross 1990, 129). The quality of references and the accuracy of the vocabulary of the issue domain can be argued in the Annotated HyperBibliography and Annotated HyperGlossary.

The Annotated HyperBibliography supports argumentation by allowing public evaluation of the references used as backing for an element in the argument. Commentary contributed to the AHB can point out other works that amplify, conflict, or rebut the cited work. Appropriate comments can be incorporated in the formal argumentation structure.

The vocabulary of the RW's issue domain is defined in the Annotated HyperGlossary. The annotation capabilities of this tool allow the users to gloss each definition, and discuss the phrasing of that definition. Should an alternative meaning emerge, that new definition may be added just as dictionaries carry alternative meanings. Hypertextual links to the proper alternative definition can be made from the RW Essays.

E-mail of course has been used for decades in dialog and will continue to do so as it is a completely general critical tool. In the RW, specific e-mail messages can be directly cited by reference. In the Landscape of Reason, a summary of the major point will be entered, but there will be a hypertextual link directly to the message.

3.2.5.2 Measurement of Quality of Argumentation

Progress in research is directly proportional to quality of argumentation. There are two forms of assessing the rationality of a discussion: material rationality or procedural rationality. Material rationality is a measure of the acceptability of the contributions to the argument; this is of course specific to the topic, and thus inaccessible to general methods of evaluation. Procedural rationality, how the dialog is conducted, is accessible on a general level, as rules for the dialog can be expressed. Pröpper has devised such a set (Pröpper 1993) (see Table III, below).

In the RW there are two major tools that express argumentation: The Landscape of Reason, and DocReview. All contributions to a discussion in the Landscape of Reason are likely to be valid elements of argument. In DocReview, many annotations may not contribute to the content of the argument, but to the form of the document. DocReview annotations need to be screened to remove the contributions that do not contribute to the content of the argument. Once the contributions have been weeded, they can be analyzed or coded. After coding, several evaluative measures can be extracted using methods developed by Pröpper (ibid.).

The purpose of evaluation is to determine the effectiveness of an enterprise. The effectiveness is a combination of many factors, principally the quality of the team members, the quality of the tools, and the quality of the assembled body of evidence. While the expense of performing such evaluation may be far beyond the resources available, it may very well be useful in some cases. Comparisons in the effectiveness of different tools may be of interest. Given a large body of participation, the effectiveness of individual members or teams could be measured. In a very large scale RW, such a study may benefit both the conduct of the research, but also contribute to research about the RW concept.

3.3 Models as an Organizing Principle for Navigation

The organization of the team's working area must be driven by the nature of the issue domain as represented by the organizing framework, a model. To organize the working section by any other method is a grave error. In a distributed RW, to organize the working area by location is to invite provincialism and competition. To organize by disciplines in a multidisciplinary team is bound to fragment the team and stultify any attempt at interdisciplinary research. Organizing by individual or authoring team is likely to encourage elitism and competition rather than teamwork.

The organizing principle of the Research Web's web site is a model that provides a framework for the pieces of the research effort. This model, the most abstract of the models, is the top level of a hierarchy of submodels. Every object in the issue domain will have to be securely placed within one of the submodels of the model of the organizing principle. The organizing model is a device that allows the user to navigate content on the basis of hierarchy of level of abstraction. This content-driven organization may be supplemented by any number of indices organized on other bases, even alphabetical.

A good organizing model uses a dominant metaphor of the issue domain. For instance, the process of plate tectonics is made abundantly clear by using maps showing the location of spreading centers, transform faults, triple points, hot spots and subduction zones. Sections across a plate relate the cycle of crust generation, plate collision, plate motion and consumption. Migration can be described with behavioral diagrams showing how an individual or family makes a decision to move and then resettles. A lifecycle timeline can be used to show the times of life when migration is more or less likely.

The conveners of the Research Web must have a conception of both the character and scope of the issue domain from the beginning. The character of the issue domain is the target of the research. The character of the issue domain is described in a model, the descriptive model; and the theory behind the behavior of the issue domain is described in an explanatory model. After the work begins, the issue domain must continue to be described and circumscribed. Defining the scope of the issue domain and its relationship to other domains is an essential and ongoing activity.

There is a definite granularity of the models used in the RW. The organizing model is very coarse, with each element being very general. The descriptive models of the issue domain are fine grained, incorporating many attributes of each object, some of which may seem inconsequential out of the context that identified them. As the modeling of the issue domain progresses, its granularity will grow finer as attributes and subprocesses are added; and will also grow coarser with the creation of supertypes that are the embedding context of previously defined types of objects. Eventually the hierarchy of models will connect the organizational model with the finest element of the described objects and processes.

The modeling hierarchy has both breadth and depth. Depth increases the detail of the objects into finer and more specialized components. A Research Web can grow in breadth, or scope, first by incorporation of topics that lie in the fuzzy boundary circumscribing the issue domain, or by making the organizational principle a submodel of a larger type of object, a great expansion of the issue domain. Breadth increases the diversity of a type of objects, for instance adding a new species to an existing genera, or adding a new land classification boundary to the existing ways of dividing the earth. Projects undertaken within a Research Web generally illuminate only a small part of a submodel of the organizing model. The knowledge created or gathered in the research process is incorporated into the modeling hierarchy as it is found. It may be that the research illuminates a section of the hierarchy that is not well defined, and will not be well-connected to the model. Finding the connections to the model will be a theory-building exercise. Poole cites the problem of unconnected theories in group communication theory (Poole 1990, 239).

In the descriptive model, the objects of the issue domain are described and related. The relationships in the descriptive model are process models created from observation and past research. Tacit knowledge is made explicit and myths and conjecture may also be included in the descriptive model, as part of personal, cultural or disciplinary knowledge (which is not necessarily true).

Explanatory models are attempts to uncover the causes of the observed behavior and form of the objects. Theory is created in the explanatory model. Explanatory models are likely to follow the form of the descriptive model. Processes are explained, and then elaborated by showing how behavior is affected by the state of the system. In attempting to discover and demonstrate cause, the explanatory model may extend into domains not described in the descriptive model. In the building of an explanatory model, theory may uncover new organizing principles that must be incorporated into an expanded descriptive model.

The simulation model is derived from both the descriptive and explanatory models. The theoretical processes that drive the behavioral model of the system must come from the explanatory model. The simulation model's objects are defined by the object's nature as specified in the descriptive model. The simulation model has several potential uses within the RW. It is a powerful means of validating work in the methodological domain. Behavioral experiments may be designed with the simulation model. In a well-developed simulation model the boundaries of the issue domain may be explored through sensitivity analysis.

3.3.1 The Descriptive model

The Descriptive Model (DM) is the model around which the substantive domain of the research is developed. This is where the team will define all the real objects it discovers in the issue domain. There are three main types of models within the DM: object models, process models, and relationship models. The object models define the attributes of each object. The process models define how the objects behave. The relationship models describe how the objects are associated with each other and with the processes.

The DM also may contain data on the objects and processes. While the models describe the characteristics of a class of objects, the data contains information about actual instances of the class. This data becomes the basis for establishing the verisimilitude of the explanatory model with reality. The DM will contain the data in datasets, and each dataset will have elaborate metadata connected to it. The metadata is the quality control for the dataset; if one is to trust the data, one must do so on the basis of the metadata. Metadata therefore is carried in a document or set of documents that are annotatable. If a member of the research team detects a deficiency in the metadata, it is noted and efforts may be made to bolster the quality of the metadata by further research. Sadly, most scientific data is not made public, but even worse, and more often, no metadata such as protocols are attached to the data.

The bulk of the remainder of the DM is a collection of documents comprising what Harré calls a sentential model (Harré 1970, 36). The sentential model is a collection of sentences, or facts (subject to discussion), that are documented in RW essays, e-mail, annotations, on-line text, etc. The sentential model is the basis for argumentation about the DM and the elements of the model refer to them for provenance.

3.3.2 The Explanatory Model

The purpose of the Explanatory Model (EM) is to explain actions within the issue domain. For every process model in the Descriptive Model (DM) there should be a corresponding process model in the EM. The EM shows how a hypothetical generative mechanism accounts for the behavior shown in the DM. The EM will include many elements of theory from the literature of the discipline. In the light of the DM and derived EM, the existing scholarly literature may come under criticism, or may be reinforced.

While an EM's process model is based on the corresponding DM process model, the EM process model will be much more complex. In the DM, it is observed that two nodes are connected by a link; it is the job of the EM to explain how the link works. In practice, it may be that there is more than one mechanism associated with a link. For instance, in the DM for family migration behavior, there might be a link between two nodes called "Active Information Gathering" and "Decision Making." The corresponding link in the EM describes why a family unit that is actively gathering information with respect to relocation will convene a meeting to discuss their migration. There may be many reasons for this change in behavior: one member of the family may have found a critical piece of information that might cause migration to go forward, or to decide not to migrate; some external event such as a legal status change might force a decision. The EM will have to become a directed multigraph; it will have multiple links between the nodes, each representing a different mechanism.

The EM's object models will be identical to the object models in the DM, though the research into mechanisms will likely make some additions to the characteristics of the object necessary. An example of a characteristic that may be added to an object "Dependent Child" in the migration model is student status. If a child is going to graduate from high school, the child might leave the family to go to a University or to take a job. The theory-building that takes place in the construction of an EM may create the need for a new, likely abstract, object that must defined in a new object model.

The hypertextual nature of the models will clearly aid the reader by having an explanatory document available by clicking each link and node. The merits of the explanation can be argued through a DocReview until it is well described. The theory represented by the EM needs to be associated with explanatory documents, with inference links, in order to justify propositions with a foundation of qualitative statements (Rao and Turoff 1990, 348). Harré refers to this set of qualitative statements as a sentential model (Harré 1970, 36).

One of the principal functions of the EM is to provide the basis for hypothesis formation. The hypothesis statement can hypertextually refer to any link or node explanatory document. When an experiment is designed to investigate the hypothesis it too can refer to the explanatory documents, object models, process models and link and node explanatory documents. In addition to those references, each term in the experimental protocol can be discussed in a document that argues the operationalization of the term. Such references constitute an auxiliary model (Blalock 1968).

The EM will be modeled in general system terms. Since the issue domain does not, and cannot, include all of reality, the EM of the issue domain will necessarily be an open system model (Sterman 1991). The establishment of the boundaries of the issue domain was discussed above (see §3.1.1). If the modelers identify and describe the interfaces with the world external to the issue domain, the embedding system to Brinberg and McGrath (Brinberg and McGrath 1985, 32), then the EM will be extensible in future research. We should keep in mind that the principal functions of the EM are to express theory and provide the basis for a simulation model. Theory need not be expressed mathematically, and indeed should not be if the mathematics purporting to describe the operation of the issue domain should constrict or misdirect thinking about the issue domain (v. Bertalanffy 1968, 24).

3.3.3 The Simulation Model

A simulation model (SM) has five major components:

The simulation model (SM) has several potential uses within the RW. It is a powerful means of validating work in the methodological domain. Behavioral experiments may be designed with the simulation model. In a well-developed SM the boundaries of the issue domain may be explored through sensitivity analysis.

3.3.3.1 Validation in the Simulation Model

The proposed simulation model is a disaggregated, discrete event model that may be run with an infinite number of scenarios, each a different combination of environment and script. There will be stochastic variables in the model in order to express uncertainty, and perhaps alternative processes that depend on the situation or state of the model at any time. Running the model many, perhaps thousands of times, using the same scenario will demonstrate robustness if the outcomes remain consistent.

The behavior of the model is profoundly affected by the processes defined in the explanatory model. These processes (algorithms) may be altered in order to establish the sensitivity of outcomes to variations in the processes. These variations represent new hypotheses and may suggest changes to the explanatory model. There may be hundreds of variables in a mature simulation model. Each variable may be examined to see its effect on the simulation. When the effects begin to depart from expected behavior, a limit of applicability of that variable is reached. Understanding the limits of the model allow the team to examine the more carefully in order to revise the explanatory model to increase the robustness of the simulation model.

3.3.3.2 Hypothesis Investigation in the Simulation Model

Hypotheses generated to extend the explanatory model must be examined by experiment in order to prove their validity. The hypothesis may be examined in the simulation model prior to design of the experiment. If the hypothesis is shown to produce expected or reasonable behavior, then the very expensive experiment may proceed with confidence.

Agre, Phil. 1998. Designing genres for new media: Social, economic and political contexts. In CyberSociety 2.0: Revisiting CMC and Community. Editor Steve Jones. Sage.

Anon. 2000. CrossRef Initiative Enhances ACM's Expanding Digital Library. MemberNet, A Supplement to Communications of the ACM vol. June, 2000: 1-2+.

Aronson, J. L., Rom Harré, and E. C. Way. 1995. Realism Rescued: How Scientific Progress is Possible. Chicago: Open Court.

Bazerman, Charles. 1988. Shaping Written Knowledge: The Genre and Activity of the Experimental Article in Science. Madison, WI: University of Wisconsin Press.

Benwell, George Lesmond. 1991. Casting Petri Nets into the System Development Life Cycle in the Context of Spatial Information Systems. Unpublished Doctoral thesis: University of Melbourne.

Bertalanffy, Ludwig v. 1968. General System Theory: Foundations, Development, Applications. New York: Braziller.

Blalock, Hubert M. 1968. The Measurement Problem: A Gap between the Languages of Theory and Research. In Methodology in Social Research. H. M. Blalock A. B. Blalock. 5-27. New York: McGraw-Hill.

Brinberg, David, and Joseph E. McGrath. 1985. Validity and the Research Process. Beverly Hills: Sage.

Chen, H. 1994. Collaborative Systems: Solving the Vocabulary Problem. IEEE Computer vol. 5: 58-66.

Cronin, Blaise, H. W. Snyder, H. Rosenbaum, A. Martinson, and E. Callahan. 1998. Invoked on the Web. Journal of the American Society for Information Science vol. 49, no. 14: 1319-28.

Cronin, Blaise, and Geoffrey McKim. 1996. Science and Scholarship on the World Wide Web: A North American Perspective. Journal of Documentation vol. 52, no. 2: 163-71.

Crowston, Kevin, and Marie Williams. 2000. Reproduced and emergent genres of communication on the World Wide Web. The Information Society vol. 16, no. 3: 201-15.

Dillon, Andrew, and Barbara A. Gushrowski. 2000. Genres and the web: Is the personal home page the first uniquely digital genre? Journal of the American Society for Information Science vol. 51, no. 2: 202-5.

Dunn, William N. 1982. Reforms as Arguments. Knowledge: Creation, Diffusion, Utilization vol. 3, no. 3: 293-326.

Eisenhart, Margaret, and Hilda Borko. 1993. Designing Classroom Research: Themes, Issues, and Struggles. Boston: Allyn and Bacon.

Funtowicz, Silvio O., and Jerome R. Ravetz. 1993. Science for the Post-Normal Age. Futures vol. September: 739-55.

Furuta, Richard, and Catherine C. Marshall. 1996. Genre as reflection of technology in the World-Wide Web. In Hypermedia Design. Proceedings of the International Workshop on Hypermedia Design (IWHD 95). 182-95. Springer-Verlag.

Gaines, Brian R. 1996. Dimensions of Electronic Journals. In Computer networking and scholarly communication in the twenty-first-century.. Eds. Teresa M. Harrison, and Timothy Stephen. 315-34. Albany NY: SUNY Press.

Girgensohn, Andreas, A. Lee, and K. Schlueter. 1996. Experiences in Developing Collaborative Applications Using the World Wide Web "Shell". in Hypertext '96. 246-55.

Gould, Stephen Jay. 2000. Linnaeus's Luck? Natural History vol. 109, no. 7: 18 et. seq.

Gross, Alan G. 1990. The Rhetoric of Science. Cambridge, MA: Harvard University Press.

Harré, Rom. 1970. The Principles of Scientific Thinking. Chicago: University of Chicago press.

Hendricksen, C. S. 1998d. An Exercise in Participatory Design Using Asynchronous Collaborative Methods.

Hendricksen, Charles S. 1989. Augmented State-Transition Diagrams for Reactive Software. Software Engineering Notes vol. 14, no. 6: 61-67.

Koku, Emmanuel, Nancy Nazer, and Barry Wellman. 2000. Netting Scholars: Online and Offline. American Behavioral Scientist vol. 43, no.

Lave, Jean, and Etienne Wenger. 1991. Situated Learning: Legitimate Peripheral Participation. New York: Cambridge University Press.

Liebow, Edward B., Judith A. Bradbury, Kristi M. Branch, Judith Heerwagen, R. Steven Konkel, and Jenniffer Leyson. 1998. The Landscape of Reason: A Scheme for Representing Arguments Concerning Environmental, Health, and Safety Effects of Chemical Weapons Disposal in the U.S. High Plains Applied Anthropologist vol. 18, no. 2: 115-38.

Meyer, Marc H., and Michael H. Zack. 1996. The Design and Development of Information Products. Sloan Management Review vol. 3: 43-59.

Millert-Raffort, Françoise. 1995. Some Cognitive and Technical Aspects of Hypermaps. in Cognitive Aspects of Human-Computer Interaction for Geographic Information Systems. eds. T. L. et. al. Nyerges. 197-211. Boston: Kluwer.

Ngwenyama, Ojelanki K. 1991. The Critical Social Theory Approach to Information Systems: Problems and Challenges. In Information Systems Research: Contemporary Approaches and Emergent Traditions. Eds. H. E. Nissen, H. K. Klien, and R. Hirscheim. 267-81. Elsivier.

Orlikowski, Wanda J., and JoAnne Yates. 1994. Genre Repertoire: The Structuring of Communicative Practices in Organizations. Administrative Science Quarterly vol. 39, no. 541-74.

Orlikowski, Wanda, and JoAnne Yates. 1998. Genre Systems: Structuring Interaction through Communicative Norms. Center for Coordination Science Working Paper #205, Sloan Working Paper #4030. http://ccs.mit.edu/papers/CCSWP205/.

Pickering, Jeanne M., and John Leslie King. 1992. Hardwiring Weak Ties: Individual and Institutional Issues in Computer Mediated Communication. in CSCW '92 Proceedings., eds. Jon Turner, and Robert Kraut 356-61.New York: ACM.

Poole, Marshall Scott. 1990. Do We Have Any Theories of Group Communication? Communication Studies vol. 41, no. 3: 237-47.

Poole, Marshall Scott. 1994. Breaking the Isolation of Small Group Communication Studies. Communication Studies vol. 45, no. 1: 20-28.

Pröpper, Igno M. A. M. 1993. Sound Arguments and Power in Evaluation Research and Policy-Making: A Measuring Instrument and Its Application. Knowledge: Creation, Diffusion Utilization vol. 15, no. 1: 78-105.

Rachel, Janet, and Steve Woolgar. 1995. The discursive structure of the social-technical divide: the example of information systems development. Sociological Review vol. 43, no. 2: 251-73.

Rao, U., and M. Turoff. 1990. Hypertext Functionality: A Theoretical Framework. International Journal of Human-Computer Interaction vol. 2, no. 4: 333-57.

Roberts, Gregory F. 1998. The Home Page as Genre: A Narrative Approach. Proceedings, 31st Annual Hawaii International Conference on SystemSciences.

Sayer, Andrew. 1992. Method in Social Science: A Realist Approach. 2nd ed. New York: Routledge.

Schön, Donald A. 1983. The Reflective Practitioner : How professionals Think in Action. New York: Basic Books.

Sterman, John D. 1991. A Skeptic's Guide to Computer Models. in Managing a Nation: The Microcomputer Software Catalog. eds. G. O. et. al. Barney. 209-29. Boulder: Westview Press.

Thagard, Paul. 1993. Societies of Minds: Science as Distributed Computing. Studies in History and Philosophy of Science vol. 24, no. 1: 49-67.

Thagard, Paul. 1997. Collaborative Knowledge. unpublished manuscript, Philosophy Department, University of Waterloo. http://cogsci.uwaterloo.ca/Articles/Pages/Collab.html.

Toulmin, Stephen, R. Rieke, and A. Janik. 1979. An Introduction to Reasoning. New York: McMillan.

Wadsworth, Yoland. 1998. What is Participatory Action Research? Action Research International vol. November: online-unpaged.

Wood, Donna J., and Barbara Gray. 1991. Toward a Comprehensive Theory of Collaboration. The Journal of Applied Behavioral Science vol. 27, no. 2: 139-62.

Zeigler, Bernard P. 1990. Object-Oriented Simulation with Hierarchical Modular Models:Intelligent Agents and Endomorphic Systems. New York: Academic Press.

Action research -- Research into social processes that involves not only researchers, but also "the researched" and "the researched for", the last in two senses: those with the problem, and those responsible for caring for those with the problem. A cyclical research process deciding on actions to take to solve the problem, implementing the actions, observing the effects, evaluating the effectiveness of the actions, and finally suggest new actions. --- (Charlie Hendricksen)

Auxiliary model -- A model designed to test a theoretical model. The assumptions of the auxiliary model operationalize the theoretical model. The incommensurability of the two models are bridged by explanation of the assumptions made in operationalization. --- (Hendricksen - from Blalock 1968 The Measurement Problem: A Gap between the Languages of Theory and Research )

Boundary object -- Objects that are plastic enough to adapt to local needs and constraints of the several parties employing them, yet robust enough to maintain a common identity across sites. Four types are: repositories, ideal types, terrain with coincident boundaries, and forms. --- (S.L. Star, 1989. The Structure of Ill-Structured Solutions: Boundary Objects and Heterogeneous Distributed Problem Solving)

Canonical document -- An authoritative document. In the context of Research Webs, the canonical document is an essay that incorporates or references the accumulated knowledge about a topic as interpreted or synthesized by the research team. --- (C. Hendricksen)

Convenor -- A founding member of the Research Web team. Generally convenors are all principal investigators. --- (Charlie Hendricksen)

Critical apparatus -- The means by which scholars criticize the works of others. In the context of Research Webs, DocReview provides the critical apparatus. --- (C. Hendricksen)

Criticism -- The action, process, or result of passing judgement, evaluating or analysing documents. In the Research Web, criticism is taken to be constructive and is considered new scholarship, itself subject to criticism. --- (Charlie Hendricksen)

Facilitator -- In a Research Web, a person who is charged with the duties of creating and maintaining the RW's web site, training team members in the operation of software tools, and actively searching for ways to take unproductive cognitive load from the team members. --- (Charlie Hendricksen)

General Systems Theory -- ... general systems theory investigates systems inductively, looking at structure, behavior involving energy transfer, boundaries, the environment, the state of the system, and characteristic parameters. --- ( Laura Laurencio)

Image map -- Image maps allow authors to specify regions of an image or object and assign a specific action to each region (e.g., retrieve a document, run a program, etc.) When the region is activated by the user, the action is executed. --- (http://www.w3.org/TR/html401/struct/objects.html#h-13.6.1)

Issue domain -- The topic for a research effort. The issue domain is comprehensive and defines the scope of the research effort. --- (unattributed)

Methodological domain -- contains the methods, designs, and research strategies used to examine concepts and phenomena. --- (Brinberg and McGrath, Validity and the Research Process)

Model -- Real or imagined representations and analogues of naturally occurring entities, structures and processes. --- (Aronson, Harré and Way, 1995, Realism Rescued, p3.)

Operationalization -- The process of creating procedures to measure real properties based on abstractions from a theoretical model. For example, how does one measure argumentativeness? --- (Hendricksen - See Blalock 1968 The Measurement Problem: A Gap between the Languages of Theory and Research )

Scientific Realism -- ... the common sense (or common science) conception that, subject to a recognition that scientific methods are fallible and that most scientific knowledge is approximate, we are justified in accepting the most secure findings of scientists "at face value." --- (Stanford Encyclopedia of Philosophy)

Scientific coordinator -- A researcher who is delegated to oversee the management of the content of the Research Web. Those duties include responsibility for defining the boundaries of the RW's issue domain, so as to maintain interdependence of the research carried out by the authoring teams. --- (Hendricksen)

Simulation model -- A computer program, usually very complex, that when given a set of initial conditions and a script of actions (scenario), will produce an outcome that a real system would produce given the same scenario. --- (Hendricksen)

Substantive Domain -- contains the phenomena, processes, or focal problems of interest. --- (Brinberg and McGrath. Validity and the Research Process)

URL -- Uniform Resource Locator -- the address of a page on the World Wide Web. --- (C. Hendricksen)

VNS -- An acronym for the Validity Network Schema. See Brinberg, David and McGrath, Joseph E. 1985. Validity and the Research Process. --- (C. Hendricksen)

Table I The Genre System of the Research Web

References

Glossary of Terms