Visualization of Analytical Processes

There is currently a major discrepancy between the dramatic improvements in hardware for sensing, communication, and storage of raw data and the capacity of humans to analyze and act on this data in a meaningful way. There is every reason to believe that this development will continue in the near future, given the revolutionary changes to hardware and software in the World Wide Web, the Sensor Web, the network of hand-held and mobile devices, and the Smart Grid. The PIs argue that improvements in the science and technology for integrating and combining analytical processing and human-computer interaction are urgently needed, so that human decision making is not overwhelmed by the flood of raw data. With that goal in mind, in this project they will develop novel mathematical, computational, and visualization methods, such that analytical processing is done partly by the computer and partly by the human, with visualization playing a central role in the communication and collaboration between the two parties. Specifically, the PIs will explore ways to integrate mosaic-based visualization and analytical processing by means of graphical models that contain thousands or millions of random variables, so as to both improve analytical learning and monitoring by human decision makers. The PIs will create feature transformation and data synthesis techniques based on probabilistic graphical models including Bayesian networks, and emphasizing multi-objective abstraction and refinement. The domain-independent techniques to be developed will be applied to electrical power systems as a test bed.

Broader Impacts: It is difficult for humans to reason under uncertainty, all the more so when they are under time pressure or stress and when there are massive amounts of data to cope with. Visualization of uncertainty has been shown to improve human performance, and there has been substantial progress in the areas of learning and reasoning using probabilistic graphical models, including Bayesian networks and Markov networks. By integrating uncertainty visualization and uncertainty processing, this research will dramatically compress the time it takes humans to make decisions under uncertainty in complex domains, will improve the quality of the decisions made, and will reduce the effort or cost associated with these analytical processes. The project's focus on electrical power system is due to the domain's importance in aerospace vehicles as well as to the urgency of creating a smart electrical grid in response to the climate change and energy crises.