M. Eduard Tudoreanu

Designing effective program visualization tools for reducing user's cognitive effort

Program visualization holds great potential for conveying information about the state and behavior of a running program. However, barriers exist to the realization of this potential, and the limited knowledge about the factors that affect program visualization makes the identification of these barriers difficult. We present arguments that the economy of information and tasks related to the visualization environment has a significant impact on the users performance in solving algorithmic problems. We apply this knowledge to develop an approach for creating application-specific visualizations solely through interactions with program visualizations and textual views of the computation, thus promoting economy of interaction. The approach consists of a multiple technical contributions that are surveyed in the paper.

Tour generation for exploration of 3D virtual environments

Navigation in complex and large-scale 3D virtual environments has been shown to be a difficult task, imposing a high cognitive load on the user. In this paper, we present a comprehensive method for assisting users in exploring and understanding such 3D worlds. The method consists of two distinct phases: an off-line computation step deriving a grand tour using the world geometry and any semantic target information as input, and an on-line interactive navigation step providing guided exploration and improved spatial perception for the user. The former phase is based on a voxelized version of the geometrical dataset that is used to compute a connectivity graph for use in a TSP-like formulation of the problem. The latter phase takes the output tour from the off-line step as input for guiding 3D navigation through the environment.

The impact of immersive virtual reality displays on the understanding of data visualization

This paper presents evidence that situational awareness in a visualization of data benefits from immersive, virtual reality display technology because such displays appear to support better understanding of the visual information. Our study was designed to de-emphasize perceptual and interaction characteristics of the dis-plays and found that the task of counting targets is strongly influenced by the type of system used to render the visualization. Immersive-type displays outperformed traditional monitors. The target objects in the study have distinguishing features that cannotmbe identified from a distance to alleviate the effect of perceptual differences among displays. Counting was chosen because it entails basic understanding of the relationship among the data values in order to recognize previously counted items. The display choices consisted of a traditional monitor and three configurations of an immersive, projection environment, obtained by selectively turning off one or two projectors of a three-wall CAVE.

Balanced cognitive load significantly improves the effectiveness of algorithm animation as a problem-solving tool

The question of whether animations depicting running algorithms are beneficial in helping users has received mixed results. This paper provides evidence that the effectiveness of animations as an aid in performing a task is affected by several factors that have no influence on text-only aids. We make our case by comparing two separate studies designed to measure whether the addition of animations to textual descriptions of a running computation promotes understanding of distributed algorithms. The text-based conditions in both experiments resulted in statistically similar results, while the animation-added conditions resulted in different outcomes. The best visualization condition outperforms all other conditions by a significant margin. The factors that were different between the two experiments fall under the concept of cognitive economy. Therefore, the degree of cognitive economy is important both for empirical studies that involve animation and for practical applications in which people rely on dynamic visualization to solve an algorithmic problem.

Evaluating the effectiveness of occlusion reduction techniques for 3D virtual environments

We present an empirical usability experiment studying the relative strengths and weaknesses of three different occlusion reduction techniques for discovering and accessing objects in information-rich 3D virtual environments. More specifically, the study compares standard 3D navigation, generalized fisheye techniques using object scaling and transparency, and the BalloonProbe interactive 3D space distortion technique. Subjects are asked to complete a number of different tasks, including counting, pattern recognition, and object relation, in different kinds of environments with various properties. The environments include a free-space abstract 3D environment and a virtual 3D walkthrough application for a simple building floor. The study involved 16 subjects and was conducted in a three-sided CAVE environment. Our results confirm the general guideline that each task calls for a specialized interaction--no single technique performed best across all tasks and worlds. The results also indicate a clear trade-off between speed and accuracy; simple navigation was the fastest but also most error-prone technique, whereas spherical BalloonProbe proved the most accurate but required longer completion time, making it suitable for applications where mistakes incur a high cost.

A study of the performance of steering tasks under spatial transformation of input

Typical direct manipulation tasks often suffer from an inherent indirection between the virtual objects that form the computer interface and the input devices through which the user interacts to manipulate these objects. This paper studies the effect of spatial indirection on the performance of interaction. For continuous input devices, spatial transformation can be decomposed into translation, rotation, and scale. Translation alone simply shifts a movement from the device space to a different position in the virtual space, preserving the direction and size of that motion. Rotation changes the direction, while scale modifies the size. This study found evidence that rotation and scale are significant factors in interaction performance. We propose a model based on these factors that can be employed to predict the time required for a task of tracing and staying inside a non-linear shape. Contrary to our initial hypothesis, moderate translation changes did not register significant variations in the required time. The results of this study are also applicable to the placement and ergonomics of physical input devices.

Interactive legends

The benefits of information visualization may be increased by adding a visual representation of the data-to-graphics encoding employed in the visualization. This paper introduces interactive legends that provide both an economical format for conveying a mapping and a widget through which the mapping can be adjusted by users. The explicit representation of the visual encoding creates an environment in which the user can focus cognitive resources on understanding the displayed data rather than on making sense of how the visualization is organized. Legend keys also promote a continuous style of interaction that allows users to adjust the appearance of the observed data according to their understanding and interest. We show the flexibility of legend keys by using them to query the information based on the properties of interest and to focus the presentation on the objects and properties relevant to the current task.

Interface for Augmenting Spatial Orientation of Pilots via Low Cognitive Load, Peripheral Vision

Untrained pilots flying in reduced visibility are at a high risk of losing control of a small aircraft when using typical instrumentation. The reduced visibility makes it impossible for the pilot to see outside references, which increases the workload and cognitive stress on the pilot. An advanced visual interface that focuses on the use of non-central vision to allow pilots to feel rather than compute the basic spatial orientation of their small aircraft is presented in this paper. To promote our goal of reducing the pilots cognitive load, the interface is controlled by a single input button in addition to the routine controls of the aircraft. The visual display aims to provide an immersive experience to the user, while not relying on the pilots central focus of attention. Thus, the user can fully exploit and focus their attention on any available instrumentation and technology present in the cockpit, such as glass displays, synthetic or enhanced vision, GPS devices, with our interface offering an additional layer of awareness. The visualization provides both instantaneous and trend information to the user. A preliminary user study that revealed additional future improvements to the visual display is also briefly described.

Methods for extracting social network data from chatroom logs

Identifying social network (SN) links within computer-mediated communication platforms without explicit relations among users poses challenges to researchers. Our research aims to extract SN links in internet chat with multiple users engaging in synchronous overlapping conversations all displayed in a single stream. We approached this problem using three methods which build on previous research. Response-time analysis builds on temporal proximity of chat messages; word context usage builds on keywords analysis and direct addressing which infers links by identifying the intended message recipient from the screen name (nickname) referenced in the message [1]. Our analysis of word usage within the chat stream also provides contexts for the extracted SN links. To test the capability of our methods, we used publicly available data from Internet Relay Chat (IRC), a real-time computer-mediated communication (CMC) tool used by millions of people around the world. The extraction performances of individual methods and their hybrids were assessed relative to a ground truth (determined a priori via manual scoring).

Visualizing weighted networks: a performance comparison of adjacency matrices versus node-link diagrams

Ensuring the proper and effective ways to visualize network data is important for many areas of academia, applied sciences, the military, and the public. Fields such as social network analysis, genetics, biochemistry, intelligence, cybersecurity, neural network modeling, transit systems, communications, etc. often deal with large, complex network datasets that can be difficult to interact with, study, and use. There have been surprisingly few human factors performance studies on the relative effectiveness of different graph drawings or network diagram techniques to convey information to a viewer. This is particularly true for weighted networks which include the strength of connections between nodes, not just information about which nodes are linked to other nodes. We describe a human factors study in which participants performed four separate network analysis tasks (finding a direct link between given nodes, finding an interconnected node between given nodes, estimating link strengths, and estimating the most densely interconnected nodes) on two different network visualizations: an adjacency matrix with a heat-map versus a node-link diagram. The results should help shed light on effective methods of visualizing network data for some representative analysis tasks, with the ultimate goal of improving usability and performance for viewers of network data displays.