Proceedings of the workshop "Adaptive Systems and User Modeling on the World Wide Web",
Sixth International Conference on User Modeling, Chia Laguna, Sardinia, 2-5 June 1997

User Modeling for Electronic Catalogs and Shopping Malls in the World Wide Web

Tanja Joerding
Dresden University of Technology
Multimedia Technology Group
Duererstr. 24, 01062 Dresden, Germany
tel. +49 351 463 8117 fax +49 351 463 8518
tj4@irz.inf.tu-dresden.de


Abstract: The development of multimedia presentations for the World Wide Web increasingly attracts attention not only in large software enterprises but also in small and middle companies. With the growth of the Internet and the expressiveness of multimedia there are new possibilities for advertising and electronic shopping. To improve the acceptance of customers and realize 'new shopping fun' it seems to be helpful to add individual components like personalized advice. This paper focuses on different forms of adaptation in electronic catalogs and shopping malls and their demands on user modeling considering different kinds of user data, the privacy and confidence of the user, as well as the need of simple integration and reuse of user modeling and adaptation components.

1. Introduction

If people go shopping there are normally sales assistants who help them to find the desired products. Observing the feedback the assistants try to give the customers enough and expected information without boring them. If the customers come in this shop frequently, the sales assistants get to know them, their special interests, and the way they want to be advised. Experienced sales assistants are able to help the customers better and more efficiently.

In the area of 'electronic shopping' today the user can browse through a vast amount of information in more or less organized catalogs, try to walk in virtual shopping malls, and has sometimes search mechanisms to find the desired products. The acceptance of such presentations is low [Diller, 96] [Mertens & Schuhmann, 96]. The time for downloading is too long, the catalogs are often less attractive and convenient, there is no 'shopping fun', and people miss individual conversations.

To overcome these problems there must be more personalized advice and the advantages of the Internet must be better exploited. A first approach can be seen in the 'Garden Escape' catalog [Garden Escape, 97] where users can design their individual gardens. But there are much more new exciting possibilities. It is unneccessary to feel embarrased when ordering clothes in the size of a tent. Adapting the appearance of the customer individual fashion shows can be realized . There is no need to be shy when asking a auto mechanics what a spark plug is. Sales assistants can be implemented who adapt to the knowledge of the user and who answer every stupid question without grinning. Customers are in virtual shopping worlds. They can view hundreds of cars, virtually go inside, and ask a lot of questions without feeling pity with the sales assistant.

To realize such 'shopping fun' a user modeling component is an important precondition. It gives the possibilitiy to transfer parts of the described personal relationship between customer and sales assistant to multimedia presentations in the WWW. Based on such a component the second task is the realization of different adaptation techniques symbolized with an electronic sales assistant. Important aspects of the user modeling component and the sales assistant and therefore a third task are the different possibilities of interaction between the customer and the electronic catalog or shopping mall.

In this paper we want to focus on the demands of an appropiate user modeling component for this special application. Considering the required information about the user section 2 of this paper shows shortly the different possibilities of adaptation in electronic catalogs or shopping malls in the WWW, section 3 discusses the problem of privacy and confidence of the user, and section 4 considers the difficulties to develop and integrate complex user modeling and adaptation techniques. Finally, section 5 discusses our future work on this topic.

2. Different Forms of Adaptation

This section gives an overview of different possibilities of adaptation in electronic catalogs and shopping malls as well as their benefits for the customer. Furthermore it outlines the neccessary information that must be saved in user models.

2.1 Adapting the structure of multimedia presentations

Advertising and electronic shopping is not only directed at people with access to powerful workstations but also at users with personal computers at home. A lot of the possible customer use modem dialup connections with 14.4 or 28.8 kbps, some have ISDN connections with 64 or 128 kbps, and very few can use T1 connections with 1.544 Mbps or better. They have different processor types, memory size, some have multimedia capabilites, like graphic card, sound card, microphone, etc. Because of such different preconditions there are very different possibilities of using multimedia presentation in the WWW.

Although downloading time is presumably no problem in the future, it is still a big problem today. People with a 28.8 kbps modem have to wait nearly 2.5 min for downloading 350 kB; a 30 seconds video file needs 2 or 3 MB (depending on several parameters).The solution to 'use only few multimedia data' is unsatisfactory. Multimedia bears great possibilities for expressiveness and many people like coloured, animated, loud and beautiful web pages.

So before thinking about adapting the content of a multimedia presentation, it is important to adapt the structure of the presentation to the hard- and software as well as the preferences of the individual user. People with no sound card need a substitute for sound data, people with a slow connection may prefer images and audio instead of video data. The quality of video and audio (size of the window, stereo/mono, resolution, compression rate, frames per second, etc.) can be adapted to the individual demands. A first step in this direction is taken in the AVANTI project [Fink et al., 97] where information that requires high bandwidth like videos and high-resolution pictures for users with low-bandwidth network access is replaced by appropriate equivalents.

2.2 Adapting communication and interaction techniques

Electronic shopping is addressed to people with very different background knowledge and capabilites. There are some very experienced people who are often bored by the normal forms of interaction and who would like to get a faster way of searching products. On the other hand there are also a lot of unskilled user who find navigation and normal interactions much to difficult. Especially in virtual shopping malls that are realized with VRML [VRML, 96] or VRT of Superscape [Superscape, 97] this user need special help. To facilitate the use of such catalogs or malls to all user, interaction models should be adapted to the different skills of the user e.g. its familarity with the mouse or with moving in a three dimensional virtual worlds and their individual preferences.

2.3 Adapting the presentation of products

When the customer is interested in a special product he wants to get a description with the expected information. If he is familar with the kind of product, maybe a computer, he can be provided with detailed and deep information whereas a novice on this area wants to get easy to unterstand explanations. For textual descriptions different methods of adaptive hypertext presentation can be applied like storing several explanation variants or preparing additional explanatation that can be shown or hidden [Brusilovsky, 96]. For multimedia elements different variants can be prepared that are adapted to user groups [Popp & Lödel, 96]. Another approach is suggested in the PPP project [André et al., 96], where multimedia presentations are planned from basic elements with respect to the requirements of the user.

2.4 Adapting the navigation throught the catalog

To reduce the information overflow it would be comfortable to get individual recommendations depending on the preferences and the chosen products like suggestions for nice shirts that go together with the chosen trousers. To realize this, known navigation support methods form the area of adaptive hypertext [Brusilovsky, 96] can be used like the adaptive annotation of links to give advice for fitting shirts. For more complex advice there is a special advising conversation necessary where an electronic sales assistant asks for the shopping aim of the customer and considers the given facts and potential constraints. [Popp & Lödel, 96] realized such a sales assistant that evaluates and suggests products with the help of fuzzy multiple criteria analysis.

Summing up, we need a user modeling component that is able to save two kinds of information. To realize the described forms of adaptation there must be firstly a general description of the user like his hard- and software, his age, his outer appearance or special characteristics like modern, conservative, prices most important etc. This information is nearly stable in a shopping session and can be used in different advising conversations. Secondly, some features must be modelled that are special for each kind of product. There is the shopping aim, i.e. a special product the customer wants to buy, the background knowledge and some special characteristics and individual preferences of the customer concerning these products (e.g. if the user wants to buy flowers, bushes and trees for his garden, the sales assistant needs to know the size of the garden, whether the user is allergical, what are his favourite colour, if he has cats or dogs, etc.). Therefore the user model for computerized selling has to consist of a general description of the customer and a lot of temporal models concerning the different kinds of products the customer wants to buy.

3. Privacy and the problem of confidence

When modeling the user and his system a lot of private data has to be saved like age, individual interests, information about the family, his hard- and software konfiguration, etc. The better the system wants to get to know the user, the more private data have to be managed. Because the World Wide Web can be accessed really 'world wide' with more or less security mechanisms, a lot of user are distrustful to give private data especially in the area of electronic commerce. Nobody wants to get an amount of advertisments by email and maybe they are afraid of being controlled (e.g. the problem of unregistered software). If adaptation takes place automatically, the user looses control and gets even more distrustful.

An additional problem in the area of 'electronic commerce' is the conflict between interests of the companies and interests of the customers. The companies want to know as much as possible of the user (e.g. customer earns a lot of money, can buy expensive products, etc) and they want to use the data for selling more products. On the other side, customers want to have their privacy and want to control all information the system gets. Facing these difficulties there must be some effort to build up the confidence of the users.

As a basic precondition the company must guarantee careful using of private data, this means that the data will not be shared with anyone and that they use only "save and proved" Internet technologies and maybe additional security mechanism like encryption and authentication methods.

The input of personal data should be voluntary. There must be the normal catalog with product descriptions. As an additional service, an adapted electronic sales assistant can take the customer through the catalog and present the products in a personalized way. The user must have control similar to a real shopping situation, where he can look by himself or ask the sales assistant. For that it could be advantageous to visualize the possibillity of adaptation e.g. as an information desk or an animated figure as it is realized in [André et al., 96] or [Jörding & Wachsmuth, 96].
The company and his intentions should be more visibly. One succesful approach can be seen in [Bruscha, 96] . In that catalog they follow up the central idea of a transparent logistics, where the customer can check the status of his orders online and where he get a picture of the warehouse and the last packed parcel. The openness of the company can support the process of building up the confidence of the user.

4. Simplicity and reuse of user modeling components

As described in the introduction there are not only the large software companies, who want to have multimedia presentations in the WWW, there are a lot of small and middle companies that want to advertise and sell there products in the Internet. To realize personalized advising the integration of user modeling and adaptation components in normal electronic catalogs must be quite simple. If it should become practical the companies need moduls that are easy to adapt to their own demands and products.

At present there are only single projects that realized their ideas in a special application domain [Lenz, 94] [Mertens & Schuhmann, 96]. In the area of user modelling there are some efforts of generalisation by developing flexible user modeling shells [Brajnik & Tasso, 94], [Kay et al., 94], [Kobsa & Pohl, 95]. Considering a lot of fundamental problems, like non-monotonicity, uncertainty and the accessibility of the user model, these shells could be used for very different applications. For a practical use in the area of electronic shopping they are too extensive and not simple enough. But following this approach components could be developed that are specialized to the area of shopping in the WWW and that are general enough to be applied to different kinds of companies and their products. Another approach is suggested in [Rosewitz & Timm, 96] where the authors want to build an special editor for the development of personalized advising components.

5. Future Work

In cooperation with Multimedia Software GmbH we are building a prototyp of a virtual multimedial shopping mall. Because the mall should be usable for customers with different hard- and software, downloading of this shopping mall should be adaptable to these different preconditions. In the electronic mall we want to realize with agent-based techniques an animated figure that symbolizes a sales assistant. This electronic assistant should help the customer walking through the mall, searching the desired products and presenting these products in a adapted way. To give the customer control over that adaptation, the sales assistant should only be active if the user asks him for help. Main focus of our work is the development of a specialized but reusable user modeling component as a basic for the different adaptation techniques described in section 2. At present we are concentrating on the following questions:

Customer modelling. What features of the customer have to be modeled? How are the dependencies to the products? What kind of communication can be realized between general and temporal models? How can a simple but realistic user modeling be realized in this special application? How can the user modeling component be specialized to be simple? And how can it be possible to reuse the user modeling component for other multimedia presentations in the WWW?

Adaptable downloading. How can downloading of the virtual world be adapted to the hard- and software as well as the individual preferences of the user? What are alternative components? How can the shopping mall be saved, such that different components can be linked together dynamically?

Forms of interaction. What kinds of navigation can be realized? How can the system support unskilled and experienced user? What kind of interaction can be realized with normal navigation of the mouse, with the development of special pathes or with interactive navigation? How can the conflict between interests of customer and company (e.g. some customers want to go as fast as possible to special products, some companies wants to show a lot of advertisments on the way) taken into account?

6. References

[André et al., 96]
André, E., Müller, J., Rist, T. (1996), The PPP Persona: A Multipurpose Animated Presentation Agent, in: Advanced Visual Interfaces, ACM Press, pp. 245-247.
[Brajnik & Tasso, 94]
Brajnik, G. & Tasso, C. (1994), A shell for developing non-monotonic user modeling systems, in: International Journal Human-Computer Studies (1994) 40, pp. 31-62.
[Brusilovsky, 96]
Brusilovsky, P. (1996), Methods and techniques of adaptive hypermedia. in: User Modeling and User-Adapted Interaction, 1996, 6 (2-3), pp. 87-129.
[Bruscha, 96]
Bruscha, B.(1996), transtec im Web: Praktischer Einsatz von WWW-basierten Informationssystemen für Kunden und Mitarbeiter, transtec AG, http://www.transtec.de/
[Diller, 96]
Diller, H. (1996), Akzeptanzstudien zum elektronischen Homeshopping, in: 3.Workshop und Minimesse "Elektronische Verkäufer mit und ohne Netz, FORWISS Erlangen.
[Fink et al., 97]
Fink, J., Kobsa, A., Nill, A. (1997), Benutzerorientierte Adaptivität und Adaptierbarkeit im Projekt AVANTI, in: Proceedings of Software Ergonomie '97, Dresden, pp. 135-143.
[Garden Escape, 97]
Garden Escape, Inc, (1997), Garden Escape - Shopping in the Garden, http://www.garden.com/
[Jörding & Wachsmuth, 96]
Jörding, T., Wachsmuth, I. (1996), An Anthropomorphic Agent for the Use of Spatial Language, in: Proceedings ECAI-96 Workshop: Representation and Processing of Spatial Expressions, full paper: KI-NRW (Applications of Artificial Intelligence in North-Rhine Westphalia) Report 96-01.
[Kay et al., 94]
Kay, J. (1994), The um toolkit for cooperative user modelling, http://www.cs.su.oz.au:80/~judy/
[Kobsa & Pohl, 95]
Kobsa, A., Pohl, W. (1995), The User Modeling Shell System BGP-MS. In: User Modelling and User-Adapted Interaction 4(2), 1995, pp.59-106.
[Lenz, 94]
Lenz, M. (1994), Fallbasiertes Schließen für die Selektion von Reiseangeboten, in: Kunze, J. & Stoyan, H. (eds) KI-94 Workshops, pp.246-247.
[Mertens & Schuhmann, 96]
Mertens, P. & Schumann, P. (1996), Electronic Shopping - Überblick, Emtwicklungen und Strategie, in: Wirtschaftsinformatik 38 (1996) 5, pp.515-530.
[Popp & Lödel, 96]
Popp, H. & Lödel, D. (1996), Fuzzy Techniques and User Modeling in Sales Assistants, in: User Modeling and User-Adapted Interaction 5(3-4), pp. 349-370.
[Rosewitz & Timm, 96]
Rosewitz, M., Timm, U. (1996), Basismechanismen für den Einsatz von Produktberatungskomponenten in Verbindung mit Elektronischen Produktkatalogen, in: 3. Workshop und Minimesse "Elektronische Verkäufer mit und ohne Netz", FORWISS Erlangen.
[Superscape, 97]
Superscape Inc. USA, (1997) http://www.superscape.com/
[VRML, 96]
VRML, (1996), The Virtual Reality Modeling Language Specification, Version 2.0, August 4, 1996 http://www.igd.fhg.de/www/igd-a4/research/vrml2/final_spec/