| Popis: |
During collaborative data analysis participants are reliant on a shared common ground to be able to understand others' locations and actions, breakdowns in which require remedial action before other activities can continue. This thesis investigates the frequency and severity of common ground breakdowns that occur during collaborative navigation, when a wall-sized display user (master) directs a desktop user (slave) within a Collaborative Virtual Environment (CVE). A series of experiments were conducted to investigate this form of collaborative navigation and evaluate the developed solutions. These experiments required participants using a desktop display to view a wall-sized display user's navigation to targets within a 3D landscape, before attempting to re-visit these targets themselves. Experiments 1 and 2 were conducted to establish the typical frequency and severity of common ground breakdowns when the desktop user is attempting to find singular and multiple targets. Participants exhibited non-trivial levels of common ground breakdown over different types of movement and input device used by the wall-sized display master. Although they frequently had sufficient common ground to reach the approximate area of a target, this was insufficient for them to be able to complete the task. The remainder of the research investigated two distinct classes of solution to these breakdown: additional views and path visualisations. For additional views, a large field of view (FOV) context view and local overview map were provided. Experiment 3 evaluated their effect and demonstrated that neither of which reduced the level of common ground breakdown exhibited by desktop users. Behavioural changes of participants using the context view still led to similar task failures. For path visualisation, two representations of paths between targets, string and heatmap were provided. Experiment 4 provides evidence that both representations significantly increased participant's success rate. Behavioural data showed that participants visited areas unrelated to their search less frequently and remained closer to the desired path. In conclusion, this research has four major contributions. First, a classification of the types, severity and frequency of common ground breakdowns that occur between desktop and wall-sized display users conducting master-slave navigation within a CVE. Second, evidence is provided that additional views are not beneficial in reducing the level of breakdown for the desktop users. Third, path visualisations are shown to be effective in reducing the level of breakdown experienced by the desktop users, and allowing more effective navigational behaviour. Finally, the successful application of heatmaps in aiding navigation when previous applications have been limited to analytical use. |
