2.1 Classifications and Taxonomy

Generally, user models are grouped into two categories, empirical quantitative models and analytical cognitive models [Bra87]. Empirical quantitative models are based on an abstract formalization of general classes of users. These models contain only surface knowledge about the user, and no internal reasoning takes place. The knowledge about the user is usually taken into consideration explicitly only during the design of the system, and is then hardwired into the system. Most conventional help systems follow this approach.

Analytical cognitive models, on the other hand, try to simulate the cognitive user processes that are taking place during permanent interaction with the system. These models incorporate an explicit representation of the user knowledge. The integration of a knowledge base that stores user modeling information allows for the consideration of specific traits of various users. In this chapter we will only discuss analytical cognitive user models.

Rich [Ric83] introduces a taxonomy where she classifies (analytical cognitive) user models along three dimensions. In the first dimension, Rich distinguishes between canonical and individual user models. In a canonical model, there is one single, typical user, while an individual user model has to be able to tailor its behavior to a heterogeneous variety of users. In the second dimension, Rich separates between explicit and implicit user models. An explicit model is built explicitly by the user, while an implicit model is built by the system through monitoring user behavior and acquiring unobtrusively other user information. Rich's third dimension categorizes long-term and short-term modeling, where the short-term modeling systems focus on building up a user model during a single session, while the former models concentrate on information that changes more slowly over time, i.e., over a whole series of sessions.

The capabilities of today's user modeling systems are still very limited [Jon89]. There are many interesting academic systems, but they are applied to very limited problem domains; they contain many simplifying assumptions, and their underlying knowledge bases have been hand-coded for the specific problem domain.

To give a better impression of what user modeling is for, the next section will discuss a concrete application of user modeling in an expert system, the UNIX Consultant.