Mixed-Initiative Support for Customizing Graphical User Interfaces

Graphical user interfaces (GUIs) are becoming increasingly complex, motivating research into ways of providing users with interfaces that are customized or personalized to their individual needs. Two opposing approaches to interface customization that have received the most attention to date are adaptable and adaptive approaches. An adaptable approach places the user in charge of customizing the interface, whereas with an adaptive approach, the system performs the customization automatically. Since both the adaptive and adaptable approaches have unique advantages and disadvantages, this thesis investigates a mixed-initiative solution to interface customization that aims to maximize each of their advantages, while minimizing their disadvantages. As our first step, we conducted an exploratory experiment with simulated users. Using GOMS analysis, we evaluated the benefits of an appropriately customized interface. We also identified ways in which adaptive support could help users customize more efficiently, and identified user and task factors that impact effective customization. Based on the results of our simulation experiment, we designed and implemented the MICA (Mixed-Initiative Customization Assistance) system. MICA provides users with a facility to customize their interfaces according to their needs, but also provides them with system-controlled adaptive support to help them customize effectively. MICA’s adaptive support relies on a novel application of GOMS analysis to reason about the potential performance implications of different customization decisions. Using this formal reasoning, MICA generates customization recommendations aimed at maximizing the user’s performance with the interface. MICA also communicates predicted time savings to the user in its rationale component, which describes why and how MICA makes recommendations. We evaluated our mixed-initiative approach through two user studies. Study One examined the general benefits of MICA’s approach in comparison to a purely adaptable alternative. The results indicate that users prefer MICA’s