On designing effective user interfaces

Graphical user interfaces, e.g., menus, have been designed to provide easy access to the functionalities, also called commands, of software. However, due to the complexity of the software (or the nature of the application domain of the software) and the expansion of users to the general public, it is still a challenge for an average user (sometimes expert users) to access certain commands of the software in an effective manner. The objective of this dissertation is to design and study more effective graphical user interfaces. Observing that a user normally only uses a very small portion of all the commands of a software, our first proposal is to equip a graphical user interface with cache. The cache area will hold the commands a user used before, aiming to provide easy access when they are accessed later by the user. Given that a user might not know what command to use to achieve her intended task or might not be able to locate the command from the interface even when she knows what command to use, our second proposal is to augment a graphical user interface with search. The search facility takes a user intent described in natural language as input and outputs a set of directly accessible commands hopefully helping achieve the users intent. Our experiments showed that both proposals offer more efficient ways for users to access the functionalities of a software. Specifically, for interfaces with cache, our study on well known command usage data sets indicates that there is a high chance (72%) for a user to find a command in the cache. We then evaluate the impact of caching to users’ performance and satisfaction in a real setting. The results showed that the users have a better performance when the chance to find a command in the cache is higher. We further propose a few different ways to organize the cache: one cache for the whole interface, one cache for each first level sub-interface of an interface, and a cache for every sub-interface of an interface. Our experiments showed that these different organizations offer different performance improvement. In the study of interfaces with search, it is natural to use the accuracy of the search, i.e., how likely the search returns the right command for the user’s intent, as a measurement of the effectiveness of search. Our experiments showed that the higher accuracy implies the better performance. We then proposed search algorithms, aiming to achieve higher ac- curacy than base line algorithms. Currently there is no data sets in the domain of human computer interaction community to evaluate these algorithms. We design experiments to collect data sets on commands (for simplicity, we take the intent the same as the command to achieve it) and the English description a user may use to search such commands. Our experiment on the data sets showed that the proposed search algorithms and the baseline algorithm produced significant differences in accuracy. Finally, we evaluated user performance and satisfaction using Microsoft Word 2013. Results showed that Word augmented with search offers better performance and user satisfaction than the original software (without search).