Definition:User models are a tool used in the construction of information systems. This system design tool is used to adapt the structure of the system to the specific needs of the user. The use of this tool is based on the principle that interaction between people and machines must take into account characteristics of the people themselves in order to be effective and efficient. (Rich, 1999) User models often utilize stereotyping techniques in order to develop the various user model categories. (Zhang and Han, 2004) There are two major varieties of user model systems. The first category is group models. This variety of user model system is the more likely of the two to use stereotyping in its construction. The group model simply places the individual user within a broad pre-constructed group definition, then proceedes to operate the system interaction within the parameters of model for that specific group. In contrast are individual models. Individual models seek to mold the system to the individual user, not to a group within which the user happens to fall. (Rich, 1999)

Three dimensions of user model space:

1. Lone model of a single "typical" user vs. a variety of models of differentiated users.
2. Models articulated in advance by system designers or users vs. systems which adapt in real time based on the user's interaction.
3. Reference to long-term characteristics such as career or hobbies vs. reference to short-term characteristics such as the particular problem at hand.

(Rich, 1999)

Examples of Information Systems and their User Models:

• An electronically based information system in a museum may classify its users under the categories expert, tourist, and student of art. (Zhang & Han, 2004, 4)
• A system was proposed for classifying UNIX users under the categories intermediate-user, novice-user, editor-user, programmer-user, network-user, and math-user. (Zhang and Han, 1004, 4)
• A reference information system in a university library could be set up to address the needs of undergraduates, graduate students, professors, and professional university librarians, as the research of each may entail vastly different needs.
• Similarly, a librarian working at a reference desk may classify the users they interact with in a similar way, or by reference source required according to the type of information they are seeking in a model paralleling the Mudge method.

Problems and Counter-Arguments:

• Use of Stereotypes: The use of stereotypes, aside from its negative connotations, can easily misrepresent the user. Since stereotypes are based on visual images, and not on the user's knowledge, user models themselves may come to be based on subjective judgments. Such subjective judgments are inherently prone to error or misrepresentation, and therefore may complicate rather than facilitate user interaction with the system. (Zhang & Han, 2004, 2-3)
• Logistics of Construction: Systems that effectively utilize user models are difficult to design and construct. To design a model that individualizes the user rather than imposing stereotypes grows exponentially more complex. Some models for individualizing the user even attempt to do so by expanding on stereotypes, not by eliminating them. (Rich, 1999) In the attempt to individualize the user, it has been suggested that the system must be able to perform two functions:

1. It must be able to attach a system of ratings and justifications to stereotypes it has already assumed or learned, which will allow it to predict other stereotypes to be utilized.
2. It must regard the user as unfixed, having unique and changing needs, and as a source of feedback resulting in its own growth and development.

(Rich, 1999, 338)

Such requirements on a technological system suggests the idea of artificial intelligence, which as a technology is, as of this writing, still a dream, and the complexity of such a system may not be achieved for quite some time.

• What is left to user control?: Information systems must make decisions about which functions to automate, and which to leave up to the user. A function should not necessarily be automated simply because it is possible. Too little user control leaves the user at the mercy of a system that may put them in the wrong place. Too much user control defeats the purpose of an information system as a research tool. This dilemma is made more difficult by the fact that some users may desire more control than others. (Borgman & Plute, 1992, 189-190)

Significance for Reference Service:User models as discussed above hold implications for any area in which information systems are involved. The last two examples provided of user modeling however point to the significance user models hold for reference work. User models can be useful in determining how to help information-seeking library patrons. Whether the user is attempting to access information on their own using the library's information system, or using the system through the help of a reference librarian, the use of user models narrows what might otherwise be unwieldy and unfocused information searches. Thus, the user models can be of use to the reference librarian seeking to answer a question as well as the user. The reference librarian may perform additional mental user modeling which works, at least in the case of a well-trained and knowledgeable individual, in conjunction with the system's user models to provide a double screen through which to filter the needs of the user. The mental user models developed and used by the reference librarian may in some cases even serve to balance the rigid group modeling of the pre-planned system with a measure of individual modeling as only personal and real-time interaction with an individual may provide.

Bibliography:

• Borgman, C. & Plute, Y. (1992). User models for information systems: prospects and problems. In Lanchaster, F. & Smith L. (Eds.), Artificial intelligence and expert systems: will they change the library?. University of Illinois at Urbaba-Champaign. Graduate School of Library and Information Science.
• Rich, E. (1999). Users are individuals: individualizing user models. International Journal of human-computer studies, 51, 323-338.
• Zhang, X. & Han, H. (2004). An Empirical testing of user stereotypes of information retrieval systems. Information processing and management, 41, 651-664

Justin D Weddle