Intelligent User Interfaces: Modelling the User

Current user interfaces are very complex with a large number of functions. Intelligent user interfaces try to help the users performing their tasks. We describe how (Horvitz et al. 1998) construct and created a user model to meet the different challenges which exist. Building on this, we first look how (Hui & Boutilier 2006) model the user’s features explicitly to suggest help based on the user’s state. Secondly, we explore how (Shen et al. 2006) predict the user’s current task. Introduction Current user interfaces (UIs) are often “bloated” (McGrenere, Baecker, & Booth 2002) and user tasks are getting more complex and more complicated. In these cases artificial intelligence can help to reduce the “bloat” in user interfaces and assist the user. However, several challenges arise when adapting the user interface or providing help: Interpreting user’s activities and predicting goals correctly is very difficult because it is many faceted and the system has only a limited amount of observations. These consist normally of the user’s interaction with the interface. Additionally the decisions have to be made online in a timely manner. For these predictions a model has to be created in most cases. Here it is difficult to determine which variables depend on others and in which way they influence each other. Once a model is created, the model’s parameters have to be learned and in some cases missing data to complete the model is a further difficulty. Structure of Intelligent UIs In general intelligent user interfaces are structured as shown in figure 1. We have a model of arbitrary complexity which observers the user interfaces, i.e. the user interaction with the computer. Based on prior knowledge (included in the model) and previous and current observations it decides how to adapt or affect the user interface for the user. A user model can consist of a large number of facets. We will discuss in this paper two of the most import ones in detail. On the one hand there are the users features and needs and on the other hand the user’s goals and tasks, i.e. what the user is doing and wants to achieve. First we will look on a more general approach on how to build user models. Then we will focus first on features and needs and then on goals and tasks. Artificial Intelligence 1 Uwe Schmidt, Marcus Rohrbach 2006-11-20 Intelligent User Interfaces: Modelling the User 6 Intelligent User Interfaces Features, Needs Goals Tasks User Interface Observations affects User Model