450 likes | 474 Vues
ARIVA PI Conference, November 2005. ARIVA. SAGE. UNCLASSIFIED//FOR OFFICIAL USE ONLY. Team Roles. Manage project/sustain vision Intel and human factors expertise. HII Research Network data Image management. Cognitive Science research Build VizArch, simBorgs Build Slate.
E N D
ARIVA PI Conference, November 2005 ARIVA SAGE UNCLASSIFIED//FOR OFFICIAL USE ONLY
Team Roles • Manage project/sustain vision • Intel and human factors expertise • HII Research • Network data • Image management • Cognitive Science research • Build VizArch, simBorgs • Build Slate • Novel Gaming Interfaces • CMU/Dynamix ARGUS • Cognitive Systems Engineering • Nutech ABEM
(Not based on ppt source.) The Slate System: Four New Developments Selmer Bringsjord Andrew Shilliday, Joshua Taylor Konstantine Arkoudas Sangeet Khemlani, Eric Pratt, Gabe Mulley, Bettina Schimanski Rensselaer AI & Reasoning (RAIR) Laboratory Department of Cognitive Science Department of Computer Science Rensselaer Polytechnic Institute (RPI) Troy NY 12180 US 11.30.05 SAGE
Some Innovative Features • Reflects new theories of hypothesis generation • E.g., MMOI-based hypo gen • E.g., scenario generaion via MDF • Automatic (v1) report generation • Includes a proved-to-be sound system, S, for assembling proofs, arguments, composite arguments, meta-arguments — and these arguments can be automatically assessed • Facilitates not just deduction, but abduction, induction, and model-based reasoning • Seamless integration with the world’s best automated and interactive machine reasoning systems and model finders (Vampire, Paradox, Athena, etc.) • IKL-compliant (empirically confirmed)
All of CS #4 Translated into Common Logic So, we’re making progress, and will have this “mega” round trip working @ Nov/Dec 05 PI meeting SAGE
The Visualization “Hookup” Challenge We are swimming in an expanding sea of systems for exploring and visualizing data in the IA domain. We need interoperability here as well! SAGE
Challenge Reduced to Software Engineering: • The Vivid family, mathematically, up to the challenge of visual data integration and interoperability. SAGE
The Uncertainty Challenge can be met; we have the formalism; simple example SAGE
79% 82% 77% SAGE
Novel Methods of Human-Computer InteractionNetwork Data Visualization Catherine Plaisant, Benjamin B. Bederson Bongshin Lee, Hyunmo Kang Human-Computer Interaction Lab University of Maryland SAGE
Two applications forNetwork Data Visualization • TreePlus • Node-link graph visualization • Tree-centered solution • Improve incremental navigation • NetLens • Interactive coordinated overviews • Dynamic filtering • Bipartite graphs
TreePlus Project Goals • Develop a good tree-centered solution for iterative inspectionof large networks andin cooperation with the team of Wayne Gray • Determine for which tasks and data sets different solutions works better • Better understand how people use visualization tools to explore graphs ++ readability ++ Rensselaer
TreePlus • Extraction of spanning tree • “Plant a seed and watch it grow” • Dynamic root selection • Incremental Exploration + Integrated Search • Pan/Zoom + Animation • Interaction techniques • Highlight and preview of adjacent nodes • Hints of graph structure • Animated transitions
TreePlus Written in C# with Piccolo.NET Bars givea preview of how fruitful it would be to go down a path When cursor hovers over a node, a preview of all nodes connected to that node appear on the right
Controlled experiment TreePlus GraphPlus
Promising Results • 28 subjects • Compared TreePlus to GraphPlus • Controlled interface “density” Study in collaboration with Wayne Gray - RPI
Two applications forNetwork Data Visualization • TreePlus • Node-link graph visualization • Tree-centered solution • Improve incremental navigation • NetLens • Interactive coordinated overviews • Dynamic filtering • Interaction history • Bipartite graphs
NetLens – Papers (ACM DL) & Authors Papers on the leftAuthors on the right Overviews provided for all attributes (here for number of papersper year) Filtered to show only papers related to visualization, and the people who wrote those papers are shown on the right side, aggregated by institution type.
NetLens –Email (Enron) & People Emails on the leftPeople on the right Overviews provided for all attributes (here for emotional tone on emails side) Filtered to show only emails related to CAenergy crisis; and the people who sentthem are shown on the right side.
Complex Visualizations of Massive Data:The Impact of Design on Mind ARDA ARIVA Workshop Orlando, Florida 30 November 2005
Intelligently Designed Interfaces for Intelligence Analysts • NIMD will bring a superhighway of information technology to the Intelligence Analyst • But, by themselves the NIMD technologies will not bridge the crucial human-computer-information gap • All of these technologies REQUIRE a user interface for the IA
Interface = Visualization + Interaction Is this a good interface?
Interface = Visualization + Interaction How about this?
Design Performance Tasks Mind Understanding the Impact of Design on Mind • A good interface is one that optimizes performance on a task that you care about • Understanding what makes an interface good or bad requires understanding the constraints and affordances in the interactions between Mind, Design, and Tasks
Our Goals: Forecasting Dynamic Changes in Cognitive Workload • Create a new generation of tools inspired by cognitive science theory that will enable us to predict the changing demands on human cognition, perception, and action imposed by an interface during task performance • Apply these tools to aid the evaluation and design of the next generation of visualization and interaction techniques for software intended for use by Intelligence Analysts Cognitive Metric Profile-- showing dynamic changes in task demands on human cognitive, perceptual, and action resources
Our Approach • 1st – Play well with others (VIA) • Develop tools and techniques so that our data collection and modeling systems can interact with tools built by others • 2nd – Simulated Human Visualizers (simBorgs) • Develop generation of cognitive theory based models of human visualization and interaction • 3rd – Cognitive Metrics Profiling • Toolkit for accessing and predicting dynamic changes in demands that an interface makes on human cognition, perception, and action as the human user performs a typical task • Human data – eye fixations, scanpaths, mouseclicks, performance time • simBorg data – predictions of changes in internal use of memory, attention, and perception
CogWorks VIA: Visualization-Interaction Architecture Current research focuses on three nested technologies
VIA • VIA release 0.0 used at UMd in August/September to evaluate an innovative design for visual display of abstract data • VIA release 0.9 used to interact with a random application downloaded from the internet • required addition of two lines of code
Demos • VIA in action -- demo of ease of instrumenting C# applications to run with VIA • Playback of logfile created during UMd session • Logfiles analyzed for UMd dissertation • Eye data analyzed for UMd dissertation • simBorg written that uses VIA to perform tasks on UMd software • Information scent implemented in simBorgs in real-time use
Bridging the Human-Computer-Information Gap by . . . • Engineer interfaces for the next generation tools developed for the IA by creating a new generation of human performance models focused on cognitive issues in usability • To tailor the designed environment to best fit the ways in which the IA • thinks • perceives • and acts