EDWARDISHAKPHD
UI Architect and Interaction Designer

Content-Aware Interaction in User Interfaces

PHD CANDIDATE (COLUMBIA UNIVERSITY)

My Ph.D. thesis made the following contributions:
Development and evaluation of content-aware transparency. Content-aware transparency (CAT), allows a user to interact with otherwise hidden content by varying the levels of transparency within different regions of a window. In our implementation, we render important regions opaque and unimportant regions transparent, with a smooth opaque-to-transparent gradient in between. Based on properties of the overlapping material, various image-processing filters are applied to obstructed content to help disambiguate the overlapping material. We designed, implemented, and evaluated a user interface that employs CAT. Our user study showed that participants were more effective with the use of CAT and also preferred user interfaces that employed CAT over ones that did not. We also developed a set of CAT interaction techniques that allow users to unambiguously interact with objects rendered with CAT: pop-through, focus filter, and mouseover pie menu. The pop-through technique allows a user to directly manipulate an obstructed object. The focus filter technique allows a user to temporarily restore obstructed image-processed content to its unfiltered form. The mouse-over pie menu technique allows a user to select an object to interact with from a pie menu of all objects that are currently under the mouse cursorÕs position.
Development and evaluation of content-aware scrolling. Content-aware scrolling (CAS) allows a user to scroll along a document path defined by the user or system, varying the direction, speed, and zoom of scrolling depending on the documentÕs content and the task at hand. We designed, implemented, and evaluated a user interface that employs CAS. Our CAS Document Viewer automatically extracts the reading path and search paths within text PDF documents, as well as the faces path within photographs containing peopleÕs faces, and allows one to traverse these paths using traditional scrolling gestures (e.g., using the mouse scroll wheel). Our user study showed that participants greatly prefer using CAS to peruse unfamiliar documents. CAS also significantly outperformed both traditional and vector (i.e., free) scrolling in short distance navigation tasks.
Development of content-aware layout. Content-aware layout (CAL) takes into consideration the contents of windows on a userÕs desktop to determine if and where they should be placed on the screen by applying constraints to content within the windows, rather than to the windowsÕ bounds. We developed a testbed application to demonstrate how CAL could be useful when perusing text documents. When a user selects text within a window, CAL rearranges other windows containing that text, horizontally aligning the search results of those windows with the selected text. Similarly, a user can perform a search across all open windows, such that search results are horizontally aligned in the center of the screen. Portions of windows not containing search results are used as available screen space, such that windows can overlap without obstructing any search result in a neighboring window.
Development and informal evaluation of a content-aware user interface combining CAT, CAS, and CAL. CASTLE (Content-Aware Scrolling, Transparency, and Layout Environment) incorporates coordinated implementations of CAL, CAT, and CAS to help users visualize and interact with related information across multiple windows. Using CAL, similar content across multiple windows is horizontally aligned on the screen. Content that would otherwise be obscured can be seen through unimportant regions of an overlapping window using CAT. Users can scroll through search results within and between neighboring windows by simply using the mouse scroll wheel. Transitions within windows are performed using CAS, while transitions between windows are performed with CAL. In an informal study, physicians used CASTLE to peruse patient status notes and reported that they are able to make otherwise important inferences much more easily and quickly than with their current system.