Adaptation and Personalization for Information Visualization

点击翻译按钮获取中文摘要

Research in Information Visualization (Infovis) has traditionally followed a one-size-fits-all approach that does not account for user differences. Despite increasing evidence that user-adaptive interaction, i.e. interaction adapted to suit each user’s specific needs and abilities, has the potential to improve users’ experience while interacting with visualizations, the effects of both user differences and different forms of adaptation in information visualization remain largely unexplored. In order to fill this gap, in the last 4 years the applicant has engaged in novel research to lay the foundations for adaptive visualization that can help tailor visual displays to specific user needs and abilities. Through several user studies, we showed that cognitive abilities such as perceptual speed, and visual and verbal working memory can impact user experience with different visualizations, and that their effect is mediated by task complexity. These studies also identified possible reasons for the change in performance for users with different measures of these cognitive abilities, based on the analysis of user’s gaze patterns collected via an eye-trackers. For instance, users with low perceptual speed were found to spend more effort in processing the legend in bar graph visualizations, which tends to slow down their performance. This result indicates that these users might benefit from support in processing legends. We also showed that gaze data can be leveraged to predict, in real-time, user and task characteristics relevant for adaptation. These results provided important evidence that user-adaptive visualizations are worth investigating, but they only scratch the surface of the know-how needed to deliver effective personalization for visualization processing. Thus, this proposal aims to conduct further research on user-adaptive visualizations that will allow closing the so called adaptive loop. In particular, we will investigate how to design and deliver personalization that is effective and unobtrusive. We will also research how to detect specific user transient states that indicate when the user is most in need of help during visualization processing, e.g. when the user is in a state of confusion. Finally, we will investigate practical applications of the proposed approaches, for instance providing adaptive support for reading complex multimodal documents (e.g., articles from the Economist) where text describes different aspects of the embedded graphs.