![]() |
Experimental Dataset | |
| User-Oriented Resource-Adaptivity |
The picture below shows the screen that subjects worked with throughout
the experiment. They used a mouse to click on the buttons labeled with
digits and on the large OK button.


For more details on the experimental setting, subjects and results, the reader is referred to the publications listed below.
| Variable | Description |
|---|---|
| CaseID | unique case identifier |
| SubjectID | unique subject identifier |
| Number_of_Steps | total number of steps in the instruction sequence |
| Presentation_Mode | mode of presentation: bundled or stepwise |
| Secondary_Task | presence or absence of a distractional task (yes/no) |
| Number_of_Presses | total number of clicks to perform the whole task |
| Number_of_Flashes | total number of flashes of the secondary task |
| Error_in_Task | presence or absence of errors concerning the primary task (yes/no) | Execution_Time | execution time of the current task (ms) | Indiv_Ave_Execution_Time | a subject's average execution time for all tasks during the whole experiment (ms) | Inter_Task_Pause_Time | pause length bewteen two consecutive tasks (ms) | Indiv_Ave_Pause_Time | a subject's average inter-task pause length during the whole experiment (ms) | Error_in_Flash_Reactions | presence or absence of errors concerning the secondary task (yes/no) |
Abstract: How can an adaptive intelligent interface decide what particular action to perform in a given situation, as a function of perceived properties of the user and the situation? Ideally, such decisions should be made on the basis of an empirically derived causal model. In this paper we show how such a model can be constructed given an appropriately limited system and domain: On the basis of data from a controlled experiment, an influence diagram for making adaptation decisions is learned automatically. We then discuss why this method will often be infeasible in practice, and how parts of the method can nonetheless be used to create a more solid basis for adaptation decisions.
| Portable Document Format (PDF) | GZipped PostScript | Uncompressed PostScript |
Abstract: We present issues and initial results of our research into methods for learning Bayesian networks for user modeling on the basis of empirical data, focusing on issues that are especially important in the context of user modeling. These issues include the treatment of theoretically interpretable hidden variables, ways of learning partial networks and combining them into a single network, ways of taking into account the special properties of datasets acquired through psychological experiments, and ways of increasing the efficiency and effectiveness of the learning algorithms.
| Portable Document Format (PDF) | GZipped PostScript | Uncompressed PostScript |