Figure 1: Dominant Communication Models of WIS
Christian
Bauer, Electronic
Commerce Network, Curtin
Business School, Curtin
University of Technology, PO Box 6469, East Perth 6982, Western
Australia.
BauerC@cbs.curtin.edu.au
Arno Scharl, Department
of Management Information Systems,
Vienna University of Economics and Business Administration,
Augasse 2-6, A-1090 Vienna, Austria (currently employed as visiting
research fellow at the Electronic
Commerce Network, Curtin
Business School, Curtin
University of Technology, PO Box 6469, East Perth 6982, Western
Australia).
Scharl@wu-wien.ac.at
In this paper future trends regarding Web-based interaction patterns are identified and put into the context of currently dominating communication models. The resulting evolution model is employed to illustrate potential approaches to developing Web Information Systems (WIS) matching the specific requirements and characteristics of each evolutionary stage. Cybernetic feedback loops are introduced to explain suitable WIS development approaches with an emphasis on communication and are applied to derive and evaluate the attributes of such approaches.
The commercialisation of the Internet and the exponential growth of the World Wide Web (WWW) introduce communication channels to millions of consumers and computer users that previously were not utilised. The success of the WWW is not only grounded on the graphical user interfaces provided by browsers rendering HTML (Hypertext Markup Language) documents, but also in the simple but effective communication architecture of HTTP (Hypertext Transfer Protocol). In the following a framework based on distinct communication patterns is employed to classify Web Information Systems (WIS). This classification builds the foundation for the subsequent outlook into further trends of WIS architectures and development.
The classification of currently dominant WIS according to the applied communication model is based on the distinction between implicit and explicit acquisition of user/client information. Table 1 lists sources as well as potential acquisition methods and illustrates available information for both types. Implicit sources of information are exploited without the assistance of WIS users, in fact many users might not even be aware of the information being transmitted. Such information can be found in server log files, persistent client state HTTP cookies, hidden CGI data, or components of Microsoft's Active Server Platform. Information that is consciously provided by the users of WIS is summarised as explicit information. This information can either be submitted interactively (e.g., through on-line forms) or gathered from past records and off-line user surveys. Naturally, on-line information acquisition is the preferred mode of data gathering.
|
Source |
Acquisition Methods |
Information |
|
|
Implicit [Client/ |
Network |
Environment Variables, |
e.g., Remote Host (Name), Browser, etc. |
|
Browser |
Cookies, Java-Applets, |
Visited Pages, |
|
|
Explicit |
Interactive |
On-line Forms, etc. |
Questionnaires, etc. |
|
Records |
Customer Database |
All of the Above |
|
The respective application of implicit and explicit information gathering constitutes the two dimensions of the classification framework for WIS. The resulting four communication models and their information flows are presented in Figure 1. Some implicit information is always transmitted with every HTTP request. However, developers of WIS may choose not to exploit this source of information due to various reasons, which range from lack of competencies to incurred cost and privacy concerns. Therefore, the utilisation of implicit information is defined as gathering such data, analysing it and reacting accordingly.
Figure 1: Dominant Communication Models of WIS
Traditional WIS consisting of static hypermedia documents represent the simplest type of communication. This does not imply that such Web sites cannot be sophisticated in the way they present information. On the contrary, HTML, client-side plug-ins and JavaScript offer quite complex options for structuring content and user interface. However, no analysis of user information is taken into account for the Web site design.
Obviously, such Web sites do not exploit the full potential of the HTTP communication architecture. Web developers overcame such one-way communication, which is normally a characteristic of broadcast-oriented mass media. They do so by looking at two types of user response - i.e., feedback in a cybernetics system approach as introduced by Norbert Wiener in "Cybernetics or Control and Communication in the Animal and the Machine" (Wiener, 1961), the term being derived from the ancient Greek word kybernêtês (steersman): On-line forms and early CGI-based programs are used to gather explicit customer information while log-file analysis, visualizations of user clickstreams, and persistent client state HTTP cookies provide implicit information. Additionally, other more sophisticated technologies became available over time. The limitations of traditional WIS communication are reached with the combination and utilisation of all previously mentioned technologies. The next section will describe how existing communication patterns can be replaced, how the corresponding WIS models can be extended, and how this evolution will affect WIS development approaches.
In this section emerging paradigms for WIS design, adaptive Web Information Systems, and complex agent negotiation are explored. WIS are put into an evolutionary context and underlying communication models of the proposed stages are compared. Figure 2 visualises observable changes in the graphical representation of the communication models for each stage regarding design (D), analysis (A) and negotiations (N). Implications of the corresponding communication structure on WIS design are derived from these constructs.
The matrix in Figure 1 covers dominant communication models of existing WIS, but does not provide any insights into future extensions and advanced communication requirements of WIS. In Figure 2, however, the classification into stages introduces an evolutionary aspect to the analysis of communication in WIS (compare Scharl, 1997; Scharl and Brandtweiner, 1998). The first two stages are direct reflections from the communication models in Figure 1, while the third and fourth stage are additions and represent new communication paradigms. The third stage introduces adaptive components and is described in more detail below. The fourth stage, complex agent negotiation, is illustrated thereafter.
Figure 2: Evolution of WIS communication models and technologies
Communication in static WIS mainly flows unidirectional from the server to the client, since user feedback is not analysed in this early stage of WIS. Nevertheless, the development of such Web sites can be quite substantial in scale and requires a planned, organised and structured approach. Such methods have been suggested in the research of Nanard and Nanard (1995), Isakowitz et al. (1995), Bichler and Nusser (1996), or Scharl (1998). A classification of academic and commercial approaches can be found in Bauer (1998). These and similar publishing approaches integrate a conceptual data and navigational model and are usually equally applicable for structured as well as unstructured information. Meta-data object type definitions and interactive, graphical tools provide designers with a framework and graphical notation during the development process.
Figure 3. WIS Design and Implementation
The development process can be broken down into a very simplistic model with only two phases at this early stage. Figure 3 provides a graphical representation of this model with a planned WIS design process, supported by a WIS design method (e.g., eW3DT) and a suitable tool (e.g., WebDesigner) leading to a static implementation on a Web server. Since user feedback is disregarded in the first stage, any redevelopment starts again with the design process, but no (formal) links between the implementation and the redesign are established. It becomes obvious that not many commercial WIS developers will remain at this stage of WIS evolution, but the adoption of a more integrated approach as suggested in stage two correlates with a significant raise in expenditure.
The analysis of WIS user feedback is introduced into the underlying communication models in stage two. This added functionality is represented by the feedback arrows from the analysis (A) to the design (D) in the communication model for this stage in Figure 2. Depending on the sub-model (2a-2c), the analysis covers implicit and/or explicit feedback (compare Figure 1). The data from explicit user feedback is obtained relatively easily (although the design of WIS becomes critical to ensure user participation) and well researched based on traditional disciplines as for example marketing research. Gathering, reporting, and visualisation of implicit feedback usually requires more sophisticated approaches (Scharl and Bauer, 1998) and is still underutilised in many commercial WIS despite the availability of an array of suitable software solutions (Malchow and Thomson, 1998).
Figure 4: Integrated feedback cycle for developing dynamic WIS
The more sophisticated communication model of stage two needs to be supported by new development processes. In Figure 4 a closer integration of WIS development is achieved by extending the design and implementation phases with usage and analysis and combining all four phases to an on-going feedback cycle. The user feedback, represented by WIS usage behaviour and patterns (e.g., clickstreams) can be combined with additional explicit information that is available about particular users or user groups and fed into tool-based WIS analysis methods. Minimum requirements for such tools are data mining capabilities for maintenance of large databases with user records, statistical analysis for aggregation of available information and visualisation for human-computer interfacing and establishing causal relationships. The integration of analysis with design methods is still missing in almost all development methodologies and tools. WebMapper, a Java-based prototype for visualising individual and aggregated access patterns, represents an effort to overcome this shortcoming by interpreting log-file data of HTTP servers and matching the results to the Web site design represented in the eW3DT methodology (extended World Wide Web Design Technique; Bauer and Scharl, 1999).
While the user feedback has an (more or less) important impact on the Web site design in the second stage of the WIS evolution, this impact does not become apparent immediately. The WIS implementation can be called unvolatile and pre-defined in so far, as the hypermedia structure and content does not react to implicit and/or explicit feedback in real time. Adaptive WIS add an automatic element to WIS implementation by introducing immediate responsiveness. User feedback is causing instantaneous regeneration of content, presentation, and navigational structure. The move from manually designed to automatically generated WIS content will help to overcome currently dominant communication patterns and sets stage three apart from the previous stages in the WIS evolution. In this stage integrated solutions for designing and analysing WIS become a necessity ("A+D" in Figure 2), processing explicit and implicit customer feedback simultaneously in order to update the underlying user model.
Figure 5. Adaptive sub-processes responsible for document generation and presentation
The same tools are responsible for generating the documents as well as the customised link structure on-the-fly during the design phase of the adaptive sub-process(es) as depicted in Figure 5. "implementation" in this context refers to document presentation; i.e., the rendering of HTML or XML (Extended Markup Language) documents. Integrated architectures will have to support standardized description models for user profiles like the Open Profiling Standard (OPS) or the Platform for Privacy Preferences (P3P). The World Wide Web Consortium's P3P provides a framework for disclosing customer information during online interactions. By allowing users to specify what kind of personal information they are willing to divulge to WIS, P3P applications support tailored relationships with specific WIS. Users can delegate decisions to their software agent. It is also a way for users to know what the corporate privacy policies are and to reconcile the preferences of the users with the policies of the company (Lee and Speyer, 1998). While OPS and P3P are somewhat similar, the specific focus of each technology is different. While OPS deals with secure transport and control of user data, P3P concentrates on enabling the expression of privacy practices and preferences. The commercially available product Cupcakes [HREF1], for example, is one of the first products built upon the OPS foundation.
The software components mentioned in the previous paragraph heavily rely on adaptive technologies like neural networks, genetic algorithms, natural language generation (Milosavljevic, 1998), case based reasoning (Finnie and Wittig, 1998), or related soft computing approaches. Applied research on these methods represents an established field and will start to influence electronic commerce substantially in stage three and increase the functionality of deployed applications. Keeping track of the user's interactions and reasoning about his/her intentions (Milosavljevic, 1998), dynamic solutions avoid repetition, facilitate navigation, and increase the overall perceived value of the provided information or services. With reduced barriers between productive data processing (transactions) and dispositive data processing like market analysis, Web-tracking, or data warehouses, the wide-spread consideration of dynamic user models for customizing Web-based information systems will become a necessity for commercial projects.
The adaptive WIS of stage three cannot be developed without a clear understanding of the appropriate parameters on which the adaptive behavior has to rely on. The definition of these parameters requires a detailed economic analysis and marketing research as well as an assessment whether it will be technically feasible to gather and incorporate the required information into a user model. The immediate reaction of adaptive WIS to user response and profiles leads to increased requirements for WIS planning and preparation. WIS developers will have to rely on the support of visual design methods and tools for the creation of commercial applications. Such methods may be derived from existing WIS modeling methods by extending their syntactic structure to dealing with adaptive systems. Such tools will be required to incorporate new additions to the repository of Web technologies, such as Artificial Intelligence, and meta-data information models, to make these technologies available for commercial applications.
While the feedback loop for the development cycle of adaptive WIS does not change so much in structure but in the level of sophistication as outlined above, the processes inside the implementation and usage phase undergo a dramatic change. Based on a central user model with information about that particular user (e.g., stereotypes or overlay domain models), the WIS is automatically designing (generating) and implementing (presenting) adaptive documents. These sub-processes do not require human intervention since they are providing users with an instant response, but are planned by the WIS developers in the design and implementation phases. In the analysis phase the information obtained from the WIS usage is complemented with previously recorded data and matched with the underlying, pre-defined user model by Artificial Intelligence-derived (e.g., case-based reasoning, neural networks) or statistical (e.g., cluster and regression analysis) methods. It goes without mentioning that these automated feedback loops and the resulting adaptive hypertext documents occur at a much higher frequency than the underlying "manual" WIS development cycles at stage three.
The enhanced functionality of autonomous agents negotiating with each other over a computer network requires rethinking, reinventing and rebuilding World Wide Web communication. The adaptive WIS in stage three of the WIS evolution of Figure 2 are still based on the same, slightly advanced network information infrastructure, but improve their communication with dynamic responses. Agents as proactive, intentional systems promise to further increase flexibility and will radically change inherent characteristics of WIS. Principal-agent relations will replace the traditional client/server approach with agent software acting as clients and servers at the same time (Hansen and Tesar, 1996), depending on their principal's preferences, the requirements of their tasks, and predefined coordination mechanisms. The predominant "request-response" model of the Web and HTTP will be dissolved by the direct interaction of equal partners in an agent-driven communication network environment. This described development becomes most apparent in the automation of complex negotiation mechanisms and models for commercial applications (Bichler and Segev, 1998). The communication model dominant in stage four, therefore, does not primarily refer to design or analysis any more, but emphasizes the process of negotiation (N) as depicted in Figure 2.
Software agents may be categorized according to their functionality and architectural attributes into information, co-operation and transaction agents (Schubert, Zarnekow and Brenner, 1998). Information agents are used for individualizing Internet-based communication and present each user with a personalized, intuitive interface that hides the more complex system architecture. They typically store a user profile together with the learning algorithm at the client side (for general Internet browsing) or server-side (for specific WIS). For the customisability of WIS they are most useful as a "gateway" to information offerings. Co-operation agent systems are based upon the subjective evaluations of persons (social information) about the delivered content. Collaborative filtering systems correlate one customer's ratings with those of others in order to identify patterns of preferences resembling those of that particular customer. The interconnected system combines the ratings of these like-minded individuals to recommend items of interest (Balabanovic and Yoav, 1997). In the same context of adaptive WIS, co-operation agents are employed to synchronize, share and communicate the subjective preferences of user groups using analogies, correlations, and other statistical methods. Negotiation or transaction agents are highly specialized programs that are used to perform the negotiation and settlement phases of electronic commerce (Schmid and Lindemann, 1998). Negotiation processes between business entities are characterised with high degrees of unpredictability, complexity and importance to organisations (Beam and Segev, 1997). Negotiating agents, therefore, will extend and overcome existing frameworks for WIS and electronic markets, although they will have to provide backward-compatibility and -interoperability to enable seamless integration (Lindemann and Runge, 1998). The highly sophisticated algorithms and communication architectures enable customers and suppliers to participate in electronic marketplaces by setting strategies and endorsing transaction agents. An early example of an agent-mediated transaction system is Market Maker (formerly known as Kasbah), an ongoing research project at the Software Agents Group of the MIT Media Laboratory [HREF2]. In the Market Maker system, negotiation is straightforward, bilateral, and competitive. It provides buyers with three different negotiation strategies: anxious, cool-headed, and frugal (depending on the selected price raise functionm which can be specified as being linear, quadratic, or exponential).
Figure 6. Dual feedback loops connected via standardised transaction environments
The transformation of the client/server-infrastructure into an agent environment with bi- respectively multi-lateral interaction between independent participants in electronic markets requires the extension of the modelling approaches to include the WIS development at both ends. Figure 6 incorporates the feedback loops for the development of two (commercial) negotiation systems interacting in an electronic market. The developers on both ends will complete the same cycle of design, implementation, usage, and analysis when improving their agent. The design of a negotiation agent deals with the same challenges as the WIS from the earlier stages, but additionally has to address increased infrastructural requirements of transaction agents and, most importantly, to fulfil the principal's strategies. The reasons for the complexity of the implementation phase of agents have already been described above. In the usage phase agents are less passive compared to earlier stages of WIS evolution, searching actively for negotiation partners in electronic market environments. It is the usage phase were the agents and their underlying development loops are tangent to each other, with their interaction facilitated by an electronic market or WIS infrastructure. After the usage principals will evaluate results and review negotiation processes. The outcomes and findings of this analysis phase will then be fed back into the design of the next agent.
At this moment, no prevalent design and/or analysis approaches for agent negotiation can be identified, mainly because of the complexity of such a task and the insecurity in the nature of this upcoming information infrastructure. Negotiation processes between business entities are characterized by high degrees of unpredictability, complexity, and strategic importance to organizations (Beam and Segev, 1997) and it can be assumed that a number of design and analysis methods for commercial utilization will be made available with the progress of this technology.
Changing communication models raise the need for corresponding modifications, extensions and innovations of WIS development paradigms, techniques and tools. Each stage of the WIS evolution described in this paper shifts the focus and the timing of design and user feedback analysis into new dimensions. A cybernetic feedback approach has been presented that supports both the evolutionary nature of the World Wide Web and the increasingly important role of user modeling for generating customized views on published information.
Balabanovic, M., & Yoav, S. (1997). Fab: Content-Based, Collaborative Recommendation. In Communications of the ACM, 40(3), 66-72.
Bauer, C. (1998). Using Reference Models to Develop WWW-Based Applications. In Edmundson, B. & Wilson, D. (Eds.), Proceedings of the Ninth Annual Australasian Conference on Information Systems, Sydney (Australia), 14-25.
Bauer, C., & Scharl, A. (1999). Acquisition and Symbolic Visualization of Aggregated Customer Information for Analyzing Web Information Systems. In Proceedings of the 32nd Hawai'i International Conference on System Sciences (HICSS-32), Maui (USA), IEEE Computer Society Press.
Beam, C., & Segev, A. (1997). Automated Negotiations: A Survey of the State of the Art. In Wirtschaftsinformatik, 39(3), 263-268.
Bichler, M., & Nusser, S. (1996). W3DT - The Structured Way of Developing WWW-Sites. In Proceedings of the 4th European Conference on Information Systems, Lissabon (Portugal), 1093-1101.
Bichler, M., & Segev, A. (1998). Brokerage in E-Commerce: State-of-the-Art and Open Issues. In March, S. & Bubenko, J. (Eds.), Proceedings of the Eight Annual Workshop on Information Technology and Systems (WITS '98), Helsinki, Finland, 53-64.
Finnie, G.R., & Wittig, G.E. (1998). Intelligent Support for Internet Marketing with Case Based Reasoning. In Cooper, J & Burgess, L. (Eds.), Proceedings of the CollECTeR'98 Electronic Commerce Conference (pp. 6-14). The University of Wollongong [HREF3].
Guttman, R., Moukas, A., et al. (1998). Agent-mediated Electronic Commerce: A Survey, in: Knowledge Engineering Review, 13 (3). In print.
Hansen, H. R., & Tesar, M. (1996). Die Integration von Masseninformationssystemen in die betriebliche Informationsverarbeitung. In Fachtagung "Data Warehouse" at the Gerhard-Mercator-Universität GH Duisburg, Duisburg, Germany.
Isakowitz, T., Stohr, E.A. & Balasubramian, P. (1995). RMM: A Methodology for Structured Hypermedia Design. In Communications of the ACM, 38 (8), 34-44.
Lee, K., & G. Speyer (1998). Platform for Privacy Preferences Project (P3P) & Citibank, Citibank Advanced Development Group.
Lindemann, M. A., & Runge, A. (1998). Electronic Contracting within the Reference Model for Electronic Markets. In Baets, W. (Ed.), Proceedings of the Sixth European Conference on Information Systems, Aix-en-Provence (France), 44-59.
Malchow, R. & Thomsen, K. (1997). Web-Tracking. In Screen Multimedia, September 1997, 57-61.
Milosavljevic, M. & Paris, C. (1998). Electronic Commerce via Personalised Virtual Catalogues, In Cooper, J & Burgess, L. (Eds.), Proceedings of the CollECTeR'98 Electronic Commerce Conference (pp. 26-37). The University of Wollongong [HREF4].
Nanard, J., & Nanard, M. (1995). Hypertext Design Environments and the Hypertext Design Process. In Communications of the ACM, 38(8), 49-56.
Scharl, A. (1997). Referenzmodellierung kommerzieller Masseninformationssysteme: Idealtypische Gestaltung von Informationsangeboten im World Wide Web am Beispiel der Branche Informationstechnik. Frankfurt, Vienna: Peter Lang.
Scharl, A. (1998). Reference Modeling of Commercial Web Information Systems Using the Extended World Wide Web Design Technique (eW3DT). In Proceedings of The Thirty-First Hawaii International Conference on System Sciences (HICSS-31).
Scharl, A. & Bauer, C. (1998). Informational Requirements for Participating in Electronic Business Ecosystems. In Western Australian Workshop on Information Systems Research, Curtin University of Technology, Perth (Australia).
Scharl, A., & Brandtweiner, R. (1998). A Conceptual Research Framework for Analyzing the Evolution of Electronic Markets. In International Journal of Electronic Markets, 8(2), 39-42.
Schubert, C., Zarnekow, R. & W. Brenner (1998). A Methodology for Classifying Intelligent Software Agents. In Baets, W. (Ed.), Proceedings of the Sixth European Conference on Information Systems, 304-316.
Wiener, N. (1961). Cybernetics or Control and Communication in the Animal and the Machine (2nd edition). New York: John Wiley & Sons.
Christian Bauer and Arno Scharl, © 1999. The author assigns to Southern Cross University and other educational and non-profit institutions a non-exclusive licence to use this document for personal use and in courses of instruction provided that the article is used in full and this copyright statement is reproduced. The author also grants a non-exclusive licence to Southern Cross University to publish this document in full on the World Wide Web and on CD-ROM and in printed form with the conference papers and for the document to be published on mirrors on the World Wide Web.