Workshop on Data Visualization

 

Sponsored by the National Science Foundation Information and Data Management Program, the Office of Naval Research, & the Navy Personnel Research and Development Center

 

Final Report

 

Jonathan I. Maletic

Department of Mathematical Sciences Division of Computer Science

University of Memphis

 

Priti Shah

Department of Psychology

University of Michigan

 

 

 

 

National Science Foundation, Information and Data Management Program

(Grant IIS-9907299)

 

Office of Naval Research

(Grant N00014-99-1-0522)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

*This workshop was jointly supported by the National Science Foundation Information and Data Management Program (under the grant IIS-9907299) and Office of Naval Research (under the grant N00014-99-1-0522).  All opinions, findings, conclusions and recommendations in any material resulting from this workshop are those of the workshop participants and do not necessarily reflect the views of the National Science Foundation or the Office of Naval Research.


Table of Contents

 

Introduction. 3

Workshop Structure. 3

Conclusions and Research Directions. 4

Goals for Data Visualization. 4

Principles of Data Visualization. 5

The State-of-the-Art in Data Visualization Tools. 6

General Research Guidelines. 6

References. 7

Appendix A: Workshop Participants. 8

Appendix B: Workshop Schedule. 9

Appendix C: Abstracts of Presentations. 10

John Behrens: Contextual issues in data visualization with novices and experts. 10

Stuart Card & Jock Mackinlay: Using Vision to Think. 10

Melody Carswell: Using Spontaneous Interpretations to Measure Graphical Efficacy. 10

Stephen Eick: ADVIZOR - A Visual Spreadsheet 11

Nahum Gershon: Visualization - Promises & Reality. What's Wrong?. 11

Jason Leigh: Collaborative Visualization in Immersive Environments. 11

Miron Livny: Metavisualization via Live Properties. 12

Tamara Munzner: Interactive Navigation of Large Graphs and Networks. 12

Murray Rowe, Gary Ropp, Charles Tatum: Navy Personal Need for Visual Information. 12

Diane Schiano: Perceiving can be Deceiving. 13

Priti Shah: Graph Comprehension: Cognitive Processes and Practical Implications. 13

John Stasko: Isn't My New Visualization Cool? Reflections on Progress in Information Visualization. 14

Barbara Tversky: Cognitive Origins of Graphic Productions. 14

Bill Wright: Business Visualization Applications. 14

Appendix D: Executive Summary. 15

Appendix E: Brief Summary. 17

 


Introduction

 

In today's Information Age, the existence of vast quantities of data and the availability of fast and abundant computing power are creating a need for systems that support the retrieval, analysis, and interpretation of information.  Fortunately, new technological developments such as animation and interactive graphics have the potential for altering the possible ways in which data can be visually presented (Cleveland, 1993; Wickens, Merwin, & Lin, 1994).  At the same time, research in the cognitive sciences on the comprehension of visually presented information has begun to provide some guidelines for the design of visual displays (Kosslyn, 1994).

 

Despite the recent strides in both the creation of information visualization tools as well as on our understanding of the cognitive processes involved in comprehending and interpreting visually presented information, there is relatively little interaction amongst the two groups.  Much of the research in cognitive psychology has focused on the comprehension process on relatively simple tasks and data sets.  And many of the technological innovations in computer science have been made with relatively little empirical research or theoretical models of human comprehension, with the exception of some basic human-computer interaction principles.

 

This relatively small-scaled workshop that brought together a group of leading cognitive psychologists who study the comprehension of visuospatial displays of quantitative information, and leading computer scientists who develop tools for information visualization.  The goal of this workshop is to provide a forum for exchange on the state-of-the-art research in each discipline, and to develop ideas for new research directions based on needs that arise at the symposium.

 

 

Workshop Structure

 

The workshop was organized over a two-day period; participants described their recent research, gave demonstrations of software tools, and discussed problems they are encountering.  In particular, the cognitive scientists provided information about the cognitive basis for using visual displays, empirical and theoretical principles for display design, and visual illusions and other possible pitfalls in display design.  Computer scientists described new technology for data visualization, including the use of virtual reality, displays for very complex tasks, and a variety of applications.

 

In addition to presenting their own research, participants focused on some specific data visualization problems presented by some researchers from the Naval Personnel Research and Development Center (NPRDC).  In the closing session, workshop participants developed a taxonomy of data visualization tasks, outlined principles for data display that are currently known (based on cognitive research), discussed the latest in tools for data visualization, and listed goals for future research.  The workshop was highly successful in defining the state of the art in data visualization. 

 

A schedule of the workshop is given at the end of this report.  Each session was organized to address a particular topic within the field.  The first session included some of the practical problems of information visualization encountered by the Navy in the domain of personnel management.  Stuart Card and Jock Mackinlay, both from Xerox Parc, gave a very good overview of information visualization research and issues.  Barbara Tversky and John Behrens discussed some of the cognitive issues of how natural visualizing information is for human from two very different perspectives.  Dr. Tversky’s perspective was historic (i.e., people have been trying to visualize information from the beginning of recorded time; what might be the cognitive basis for the benefit of using these visualizations) while Dr. Behrens’ was from an educational point of view.  The “Possible Pitfalls” session ended the first day with some enlightening (and sometimes very entertaining) examples of problems with “too much” visualization.

 

The second day started with talks and demonstrations from the developers of two visualizations tools.  Stephen Eick from Lucent Technologies and Bill Wright from Visible Decisions discussed the types of problems their customers encounter and a number of the features they have built into their tools to address these problems.  Priti Shah and Melody Carswell both talked about the cognitive processes involved in understanding visual displays of information.  The last session included a description of the CAVE virtual reality environment for visualizing information by Jason Leigh, a number of tools and methods for visualizing large graphs and networks by Tamara Munzner, and a environment for building visual interfaces in to data sets by Miron Livny.  A complete list of the presenters with their affiliations can be found at the end of this report along with short abstracts of each talk.

 

A web site for the workshop (http://www.msci.memphis.edu/~maleticj/visual/) contains much of the information in the report along with a number of the participant’s talks (in MS Power Point).  Links to participants web pages are also provided.

 

 

Conclusions and Research Directions

 

At the end of the workshop, an open discussion took place to contemplate issues addressed during the talks and attempted to draw any conclusions and future directions for research within the field of data and information visualization.  Some of the major conclusions and the suggested directions for future research, drawn from both the final session and individual talks, are outlined below.

 

 

Goals for Data Visualization

 

The first fundamental issue of discussion was the importance of characterizing the different goals, user tasks, and types of data sets for data visualization.  The discussion focused on the following major features:

 

·        What are dimensions that characterize data visualization situations?  Workshop participants discussed the need for a taxonomy to characterize data visualization situations.  A tentative taxonomy was proposed by workshop participants in which three major dimensions were identified.  The first dimension is the kind of data presented: structural, quantitative, and “mixed”.  The second dimension is the type of data depicted (e.g., geographical, web sites, sales, etc.).  The third dimension is the kinds of user tasks.  Tasks include data, data cleansing/quality analysis, understanding conceptual information such as relationships, and displaying data for communication purposes (e.g., presenting military personnel data to Congress, or displaying sales data to a Board of Directors).  A major point of discussion is that the task of simply presenting or even filtering data is not enough.  A goal of data visualization is comprehension of relationships and concepts, including the ability to predict other information. 

 

·        Can there be general principles for data visualization, or is the display so dependent on the kind of task that it is not possible to draw general conclusions?  The second issue of discussion involved how general the principles for designing data visualization tools can be.  Cognitive research suggests that tasks, data sets, and the type of data depicted all have an impact on the kinds of displays that are best for presenting information.  Although there are some very general principles for data display (e.g., avoiding certain visual illusions), and some highly specific principles for data display in specific contexts, the middle ground (general principles that focus on classes of situations) was considered an important direction for more cognitive science/ human-computer interaction research.

 

·        Can we make better progress with some standard problems or data sets?  A third conclusion about data visualization tasks is that the field (and interaction between Cognitive Science and Computer Science) would greatly benefit from shared data sets and problems for which researchers might propose and empirically evaluate different solutions.  It was suggested that a web site be set up with pointers to such data sets and problems.

 

 

Principles of Data Visualization

 

A second fundamental issue identified by the workshop participants was the need for empirical or theoretical guidelines for data.  On the positive side, there were many design principles, based on cognitive research, that were identified and discussed during the workshops final section.  Less encouraging was the fact that there had been less communication about these principles to computer scientists that were creating large-scale data visualization software (especially in industry).  In addition, it was clear that much research was still needed; much cognitive research focused on relatively simple data sets and tasks, and could not always be directly scaled-up for the class of problems and tasks in real-life contexts, such as military and business applications.  Below, some of the major design principles and important issues for future research are listed. 

 

·        What are the design principles that can be specified based on research that has already been conducted?  Some of the major principles that were presented include information about the potential benefit for animation for particular situations (i.e.,  the use of animation for showing transitions from one state to another), the benefits and potential pitfalls of plotting data three-dimensionally, the influence of prior knowledge/expectations, the different kinds of interpretations given to different graphic formats, possible misperceptions of static and animated displays (i.e., visual illusions, distortions, etc.)

 

·        What are important questions for further research that will help guide data visualization development?  Despite the fact that there was a consensus of some design principles, it was clear that future research needs to focus on 1) more complex data visualization problems/tasks, 2) the new and possibly beneficial tools available such as animation and interactivity, and 3) principles that focus on classes of tasks and data sets (rather than overly specific or overly general principles).

 

 

The State-of-the-Art in Data Visualization Tools

 

A third major focus of the workshop was the presentations by computer science researchers who have developed the state-of-the-art tools in data visualization.  Researchers presented several tools and new kinds of displays that might help viewers deal with highly complex data, such as virtual reality based systems, drill-down tools, hyperbolic displays, and “personal dashboards.”  Several issues of specific concern were highlighted during the workshop, and include the following questions.

 

·        What are the different issues that arise when developing software for specific applications and those that are designed for general use (and those that are general but can be customized for individual use)?  The process of creating software for specific applications was shown to be somewhat different than the creation of more general-purpose software. 

 

·        Which of the new tools and capabilities are useful, and for what purposes?  Although many of the new features have intuitive appeal, it was agreed that usability studies, and evaluation based on cognitive principles, need to be performed.  One issue that might be considered is the possible trade-off between aesthetically appealing displays, and displays that might lead to the best performance on certain information visualization tasks.  A second issue regarding the usefulness of displays is that the evaluations need to occur in the context of actual users, as opposed to domain novices.

 

·        What other advances are required in the development of visualization tools?  Software for collaborative data visualization was considered an important direction for development.

 

·        What about training people to use these tools?  Cognitive research has shown us that people have trouble comprehending even simple displays and understanding relatively simple data.  Given the relative cost (much cheaper than in the past) of sophisticated visualization software and the ability for relative novices to access the software, how can we teach them to better utilize such tools? 

 

 

General Research Guidelines

 

The final fundamental issue of the workshop included discussions of the most important research directions and general guidelines for future research.  In general:

 

·        There is a need for more interaction between cognition and perception researchers and computer scientists who are creating tools for information visualization.

 

·        Principles for the visualization of information need to be better outlined so that tool developers can easily apply them.

 

 

·        Empirical research (both basic cognitive research and more applied usability studies) must go beyond the comprehension processes involved in relatively simple data interpretation tasks and the use of domain-novice populations.

 

·        There is a great need to make large real-world-style data sets and problems widely (and centrally) available to researchers in data visualization.  A repository of problems should help focus research on some grand changes and also server as benchmarks for new methods and techniques.

 

 

References

 

Cleveland, W. (1993).  Visualizing Data.  Murray Hill, NJ: AT&T Bell Laboratories.

 

Wickens, C. D., Merwin, D. H., & Lin, E. L. (1994).  Implications of graphics enhancements for the visualization of scientific data: dimensional integrality, stereopsis, motion, and mesh, Human Factors, 36, pp. 44-61.

 

Kosslyn, S. (1994).  Elements of Graph Design, New York: WH Freeman.

 


Appendix A: Workshop Participants

 

Name

Institution

e-mail

John Behrens

Arizona State University Division of Psychology in Education

mailto:behrens@asu.edu

Stuard Card

Xerox PARC

card@parc.xerox.com

Melody Carswell

University of Kentucky Department of Psychology

cmcars00@pop.uky.edu

Stephen Eick

Lucent Technologies Visual Insights

eick@research.bell-labs.com

Nahum Gershon

MITRE

gershon@mitre.org

Jason Leigh

University of Illinois Chicago Electronic Visualization Lab

spiff@evl.uic.edu

Miron Livny

University of Wisconsin Madison Computer Science Department

miron@cs.wisc.edu

Jock Mackinlay

Xerox PARC

mackinlay@parc.xerox.com

Jonathan Maletic

The University of Memphis Department of Math Sciences Division of Comp Sci

jmaletic@memphis.edu

Tamara Munzner

Stanford University Computer Science Department

munzner@cs.stanford.edu

Murray Rowe

NPRDC

rowe@nprdc.navy.mil

Gary Ropp

NPRDC

ropp@nprdc.navy.mil

Diane Schiano

Interval Research Corporation Stanford Research Institute

schiano@interval.com

Priti Shah

The University of Michigan Department of Psychology

priti@umich.edu

John Stasko

Georgia Tech College of Computing

john.stasko@cc.gatech.edu

Charlie Tatum

NPRDC

tatum@nprdc.navy.mil

Barbara Tversky

Stanford University Department of Psychology

bt@psych.stanford.edu

William Wright

Visible Decisions Inc

billw@vdi.com

 

 

Some of the other individuals who attended the workshop included:

Tim Brogdon - NPRST

Jan Dickieson - ONR

Helen Gigley - ONR

Art Graessere – The University of Memphis

Max Garzon – The University of Memphis

Xiangen Hu – The University of Memphis

Andi Marcus – The University of Memphis


Appendix B: Workshop Schedule

 

The University of Memphis

Fogelman Executive Center

June 18th & 19th, 1999

 

 

Time (CST)

Presenter & Title

Arrival

Thursday June 17th

 Hotel Check in: Fogelman Executive Center

Day 1

Friday June 18th

Room 308 Fogelman Executive Center

Introductions

9:00am - 9:30

 

Session 1

Practical Problems

9:30 - 11:00

Gary Ropp, Charles Tatum, Murray Rowe

Navy Personal Need for Visual Information

Break

15 min.

 

Session 2

Thinking Visually

11:15 - 12:45

Stuart Card & Jock Mackinlay: Using Vision to Think

Lunch

12:45 - 1:45

Fogelman Executive Center Dinning Room (3rd Floor)

Session 3

How Natural are Visualizations?

2:00 - 3:30

Barbara Tversky: Cognitive Origins of Graphic Productions

John Behrens: Contextual issues in data visualization with novices and experts

Break

15 min.

 

Session 4

Possible Pitfalls

3:45 - 5:15pm

John Stasko: Isn't My New Visualization Cool? Reflections on Progress in Information Visualization

Diane Schiano: Perceiving can be Deceiving

Nahum Gershon: Visualization - Promises & Reality. What's Wrong?

 

Dinner

7:00pm

Local Restaurant: Tsunami in Midtown (Cooper-Young)

 

 

 

Day 2

Saturday June 19th

Room 308 Fogelman Executive Center

Session 5

Visualization Systems

9:00am - 10:30

Stephen Eick: ADVIZOR - A Visual Spreadsheet

Bill Wright: Business Visualization Applications

Break

15 min.

 

Session 6

Cognitive Processes

10:45 - 12:15

Priti Shah: Graph Comprehension: Cognitive Processes and Practical Implications

Melody Carswell: Using Spontaneous Interpretations to Measure Graphical Efficacy

Lunch

12:15 - 1:30

Fogleman Executive Center Dinning Room (3rd Floor)

Session 7

Visualization Environments

1:30 - 3:00

Jason Leigh: Collaborative Visualization in Immersive Environments

Tamara Munzner: Interactive Navigation of Large Graphs and Networks

Miron Livny: Meta-visualization via Live Properties 

Break

15 min.

 

Capstone

3:00 - 4:30

Where Are We and Where Do We Go?


Appendix C: Abstracts of Presentations

 

 

John Behrens: Contextual issues in data visualization with novices and experts

 

While data visualization is sometimes recommended because it supports application of one's knowledge about the natural world to aid data interpretation, most visualization tasks occur in abstract symbol systems whose meanings and goals must be learned.  Accordingly, the social, semiotic, and cognitive setting of the visualization activity is important to consider.  In the first part of this paper I discuss instructional problems and solutions that have emerged from teaching graphical data analysis to novice educators. Instructional recommendations are made with regard to symbol-system understanding, goal specificity, pattern recognition development, and social modeling.  The second part of the paper describes initial efforts to understand the variety of roles and activities undertaken by experts in the practice of data visualization.  The goal here is to determine unique and common aspects of "visualization cultures" across a variety of visualization activities including artistic expression, stereoscopic modeling, and statistical graphics.

 

 

Stuart Card & Jock Mackinlay: Using Vision to Think

 

Information visualization is best understood as a way to use external cognition to aid performance.  As Bertin says, "Graphics is the visual means of resolving logical problems".  We will combine our two presentations to make several points about the relationships between tasks, interactive computer graphics visualizations, and cognitive performance:  We will suggest that many tasks of interest to the navy and others are of a certain form, which we will call knowledge crystalliization.  We will give a simple taxonomy of the different means by which information visualization an act to improve performance on these tasks and also mention four different levels at which information visualization can be applied.  We will then proceed to discuss the design space of information visualizations in terms of a reference model and the types of possible mappings that derive from the structural limitations of data and visual representations.  An interesting class of mappings are focus + context displays that attempt to partition human bandwidth between local task and situation awareness.  Finally, we will look at some attempts to get insight into user behavior with these systems.  Information foraging theory analyzes the information access portion of knowledge crystallization in terms of information gain per unit cost.  It leads to a notion of "information scent" or degree-of-interest.  Analyzing the information ecology of the worldwide web leads to how information scent can be spread propagated by spreading activation, then visualized.  The relationship of information scent to visual attention is a current concern of ours and we will briefly discuss some current work.

 

 

Melody Carswell: Using Spontaneous Interpretations to Measure Graphical Efficacy

 

Much of the research that has been conducted on the usability of graphs has focused on the speed and accuracy with which research participants can answer predetermined questions.  Several studies in our lab have explored the utility of an additional approach to the measurement of display design success -- the analysis of participants? spontaneous interpretations. When participants perform spontaneous interpretations, they rather than the researchers determine which questions will be asked of the displayed data.  The technique allows an analysis of both qualitative and quantitative aspects of graph-reading performance.  Analysis of spontaneous interpretations has revealed differences in sensitivity to data features as a function of display format (line graphs vs. bar graphs vs. tables).  The technique has also revealed differences in processing strategies used by participants with different levels of preference for graphical rather than tabular displays.  In addition, the technique has been used to study hypothesis generation among scientists as a function of format characteristics in statistical maps.  Analysis of spontaneous interpretations is likely to be particularly well suited to research on display support for tasks that involve both question finding and question answering.

 

 

Stephen Eick: ADVIZOR - A Visual Spreadsheet

 

ADVIZOR, analogous to a spreadsheet, is a platform for creating visual applications.  The paradigm is that authors write application templates for viewers.  The templates customize ADVIZOR to answer specific business questions, enabling faster more informed decisions.  The significant technical contributions in ADVIZOR include:

1) a wide selection of visual analysis and reporting components

2) dynamic linking of all visual components

3) integrated support for visual querying

4) rich color encoding

5) unique drill-down, filtering, and analysis capability

6) ubiquitous data access

7) visual scripting and session management

8) ease of programming, using standard languages such as VB, C++, or J++

 

 

Nahum Gershon: Visualization - Promises & Reality. What's Wrong?

 

No abstract available.

 

 

Jason Leigh: Collaborative Visualization in Immersive Environments

 

This talk outlines the Electronic Visualization Laboratory's past, present and future work in Tele- Immersion: the union of collaborative virtual reality and image based rendering in the context of significant computation and data-mining.  When participants are tele-immersed, they are able to see and interact with each other and objects in a shared virtual environment.  Their presence will be depicted by life-like representations of themselves (avatars) that are generated by real-time, image capture, and modeling techniques.  The environment will persist even when all the participants have left it.  The environment may autonomously control supercomputing computations, query databases and gather the results for visualization when the participants return. Participants may even leave messages for their colleagues who can then replay them as a full audio, video and gestural stream.  The issues that will be highlighted are the broad set of challenges in human-factors, visualization, and networking that must be closely coordinated in order for Tele-Immersion to be successful.

 

 

Miron Livny: Metavisualization via Live Properties

 

When confronted with a visualization, new and experienced users alike ask the same questions: What do I see?  How can I control it?  How do I change it?  These questions suggest that active manipulation of visual *context* is as important as active manipulation of visual *content*.  Nevertheless, current data visualization software systems isolate design, layout, and interaction stages.  Effective visualizations emerge only after a time-consuming iterative process performed by an expert, often the visualization researcher.  An open challenge for data visualization software research is to unify context and content manipulation by merging design, layout, and interaction in visualization applications.  Our response is Live Properties, an object-oriented user interface architecture which combines reusable interface components built around properties with a multicast communications mechanism similar to active values.  By binding active values to component properties, we have unified direct manipulation with tight coupling.  Moreover, a Live Property visualization can be self-describing through a family of techniques we call metavisualization.

 

 

Tamara Munzner: Interactive Navigation of Large Graphs and Networks

 

There are many large real-world datasets that can be characterized as node-link graphs.  Scalability is a major problem, since many of these datasets are much larger than can be handled by most existing graph drawing systems.  I'll discuss a few interactive graph/network visualization systems that I've built and the differences in their visual metaphors and interaction techniques due to the quite different tasks that each was designed to support.  Much of the successful information transfer from cognitive psychology to information visualization has dealt with the kind of static 2D encoding of visual information that can be accomplished with a pen and paper.  Fluid interactivity is one of the most important differences between computers and static paper, but it remains ill-characterized.  We might be able to use insights from fields like environmental cognition in order to better exploit human spatial experience in the real 3D world.  However, understanding real-world experience is only part of the battle when answering questions like "For which tasks are distortion viewing techniques appropriate?".  In the real world navigation can only be accomplished by rigid motion across fixed terrain.  An information visualization system might use a very different visual metaphor where "navigation" consists of warping structures or even the space itself in response to user actions.  A clearer understanding of the tradeoffs of these differing approaches could help in the design of more effective tools.

 

 

Murray Rowe, Gary Ropp, Charles Tatum: Navy Personal Need for Visual Information

 

Navy personnel management is a large, complex enterprise that encompasses recruiting, classification, training, assignment, retention management, career progression, and separations for a force of about 385,000 people.  Personnel managers are faced with the challenging task of attaining and maintaining the right number of people with a wide variety of required skills and experience in the right assignments to assure the readiness of the Navy force to respond to any threat to national security.  This challenge has become increasingly difficult in recent times as increased fiscal constraints and force reductions have decreased the margin for error allowed to the personnel managers.  In addition, the Information Age has brought the expectation that personnel managers should be able to more easily track and report significant information about the state of the force, leading to reductions in management personnel.  In fact, although much more data are available than ever, little progress has been made in making the data easy to assimilate, and much time and effort are wasted sifting through data.  As a result, managers who are overburdened trying to accomplish their routine daily tasks have little opportunity to benefit from the tremendous volume of incomprehensible numbers at their disposal.  There are many questions that could be answered if the available data were easily interpreted.  For example: What ships are going to be understaffed in some skills for an upcoming deployment?  What skills are losing people at a faster rate than they can be replaced?  Can enough people be put into training classes to make up an impending shortage in a critical skill?  In what areas should additional recruiting resources be allocated?  What significant changes are occurring in the composition or behavior of the force?  There is a complex chain of interconnected components that lead to force readiness.  Visual representations of data for each of the components and of the interactions between the components would greatly improve the ability of personnel managers to understand the impacts of their policies, and to direct the force to desired readiness.

 

 

Diane Schiano: Perceiving can be Deceiving

 

The visual representation of data can promote insight but also illusion.  In this talk I'll describe some work demonstrating that even the simplest and most straightforward data graphics can produce systematic distortions.  I'll then briefly discuss some principles for the design of more effective visual displays. 

 

 

Priti Shah: Graph Comprehension: Cognitive Processes and Practical Implications

 

Graphs and other "visualizations" are used extensively to facilitate the communication and comprehension of quantitative information, perhaps because they seem to exploit natural properties of our visual system such as the ability to process large amounts of information in parallel.  In this talk I will present research that suggests that rather than a holistic pattern recognition process, graph comprehension is a complex, interactive process akin to text comprehension.  Viewers form a mental model of the quantitative information displayed in the graph through serial, iterative cycles of identifying and relating the graphic patterns to associated variables.  The subsequent interpretation formed by graph viewers is constrained by a number of factors.  First, the graphic format influences what information is perceptually salient, and thus, the kinds of inferences viewers can easily make.  In addition, the semantic content has an influence on viewers interpretations: viewers' are likely to describe what they expect, not necessarily exactly what they see.  Finally, a viewers' graph reading skill has an influence on their interpretations; novices are more likely to provide surface-level descriptions of data and thus be constrained by graphic format, whereas experts are more likely to be able to go beyond these surface-level descriptions and make some inferences. 

 

In summary, our data suggest that graph comprehension, like text comprehension, is a constructive, integrative process in which viewers form a mental model of the quantitative information in graphs based on a number of display, content, and individual differences factors, rather than a holistic pattern recognition process.  This constructive, integrative model of graph comprehension has numerous implications both for data display, as well as for the teaching of graphical literacy skills.  In the final portion of this talk, I will discuss some of the main implications.

 

 

John Stasko: Isn't My New Visualization Cool? Reflections on Progress in Information Visualization

 

Information visualization is an emerging and exciting new research area.  As with any new discipline, the area is likely to encounter some growing pains as it matures.  In this short talk, I'll identify some important problems and issues that I feel researchers must not overlook in their studies.  In particular, I'll pick on the "Gee whiz, isn't it cool" research methodology and attempt to articulate how the field can mature and grow even farther.  I'll also briefly discuss some of my group's recent work on empirical studies of information visualizations and on visualizing dynamic information sources.

 

 

Barbara Tversky: Cognitive Origins of Graphic Productions

 

Graphics, such as maps, are an ancient invention.  Cross-cultural and developmental studies suggest that graphic schematizations reflect natural correspondences between mental elements and relations to external space, preserving varying levels of information.  Our current projects comparing route directions and maps and comparing bar and line graphs reveal common cognitive schematizations underlying both.  Design implication: use space meaningfully.

 

 

Bill Wright: Business Visualization Applications

 

Four case studies will be reviewed on the use of interactive 3D visualization for business analysis and reporting.  The reasons these companies are using visualization, and the value they are achieving, will be discussed.  Screen shots will be used to identify the layouts and representations that are effective.  Appropriate use of interactive 3D will be discussed as a sidebar.  If time permits, a ten-minute demo of 3D business visualization will be provided.


Appendix D: Executive Summary

 

Workshop on Data Visualization

Sponsored by the National Science Foundation Information and Data Management, the Office of Naval Research, & the Navy Personnel Research and Development Center

 

Executive Summary

 

Introduction.  In today's Information Age, the existence of vast quantities of data and the availability of fast and abundant computing power are creating a need for systems that support the retrieval, analysis, and interpretation of information.  Fortunately, new technological developments such as animation and interactive graphics have the potential for altering the possible ways in which data can be visually presented.  At the same time, research in the cognitive sciences on the comprehension of visually presented information has begun to provide some guidelines for the design of visual displays.

 

This relatively small-scaled workshop brought together a group of leading cognitive psychologists who study the comprehension of visuospatial displays of quantitative information, and leading computer scientists who develop tools for information visualization.  The goal of the workshop was to provide a forum for exchange on the state-of-the-art research in each discipline, and to develop ideas for new research directions based on needs that arouse at the symposium.

 

The workshop was divided into eight sessions.  During the first session, NPRDC presenters discussed a set of needs for data visualization.  The following sessions included talks of recent research by the workshop participants.  Finally, there was a capstone session during which the major conclusions of the workshop and directions for future research were identified.  Below, are outlined some of the major conclusions and the suggested directions for future research.

 

Major Conclusions.  The workshop was highly successful in defining the state of the art in data visualization.  Cognitive scientists provided information about the cognitive basis for using visual displays, empirical and theoretical principles for display design, and visual illusions and other possible pitfalls in display design.  Computer scientists described new technology for data visualization, including the use of virtual reality, displays for very complex tasks, and a variety of applications.

 

Goals for Data Visualization.  The first fundamental issue of discussion was the importance of characterizing the different goals, user tasks, and types of data sets for data visualization.  The discussion focused on two major questions.  What are dimensions that characterize data visualization situations? Can there be general principles for data visualization, or is the display so dependent on the kind of task that it is not possible to draw general conclusions?  Tentative answers to these questions were proposed, based on workshop presentations, but it was agreed that to address these questions adequately, we need significant future research,  In particular, three fruitful approaches were identified.   First, although we have very general descriptions of users tasks, Card & MacKinlay emphasized the need for more detailed evaluations of users in fairly complex problem-solving contexts (e.g., their studies people who write reports distilling large amounts of information).  Research that focuses on classes of problems (e.g., the relative benefit for animation for different classes of tasks) is an important direction for applied cognitive psychology. Finally, it was agreed that the field would benefit from shared data sets or problems for which researchers might propose and empirically evaluate different solutions.

 

Principles of data visualization.  A second issue identified by the workshop participants was the need for empirical or theoretical guidelines for data.  On the positive side, there were many design principles, based on cognitive research that were identified and discussed during the workshops final section.  Less encouraging was the fact that there had been less communication about these principles to computer scientists that were creating large-scale data visualization software (especially in industry).  In addition, it was clear that much research was still needed; much cognitive research focused on relatively simple data sets and tasks, and could not always be directly scaled-up for the class of problems and tasks in real-life contexts, such as military and business applications.

 

The state-of-the-art in data visualization tools.  A third major focus of the workshop was the presentations by computer science researchers who have developed the state-of-the-art tools in data visualization.  Researchers presented several tools and new kinds of displays that might help viewers deal with highly complex data, such as virtual reality, drill-down tools, hyperbolic displays, and “personal dashboards.”  Some major issues for future research include differences in the design of software for specific purposes, empirical evaluations of new software tools/displays with real users, the development of collaborative visualization tools, and the training of users to use such tools.

 

General Research Guidelines.  The final focus of the workshop included discussion of the most important research directions and general guidelines for future research.  They include:

·        There is a need for more interaction between cognition and perception researchers and computer scientists who are creating tools for information visualization.

·        Principles for the visualization of information need to be better outlined so that tool developers can easily apply them.

·        Empirical research (both basic cognitive research and more applied usability studies) must go beyond the comprehension processes involved in relatively simple data interpretation tasks and the use of domain-novice populations.

 

Workshop organizers:

Priti Shah                         priti@umich.edu                        Dept. of Psychology, University of Michigan

Jonathan I. Maletic                         jmaletic@memphis.edu                         Computer Science, University of Memphis

 

For more information about the workshop, a list of participants, conclusions, and downloadable versions of a number of the presentations, see:

http://www.msci.memphis.edu/~maleticj/visual/

 

 

*This workshop was jointly supported by the National Science Foundation Information and Data Management Program (under the grant IIS-9907299) and Office of Naval Research (under the grant N00014-99-1-0522).  All opinions, findings, conclusions and recommendations in any material resulting from this workshop are those of the workshop participants and do not necessarily reflect the views of the National Science Foundation or the Office of Naval Research.


Appendix E: Brief Summary

 

Workshop on Data Visualization

Sponsored by the National Science Foundation Information and Data Management, the Office of Naval Research, & the Navy Personnel Research and Development Center

 

The University of Memphis

June 18th and 19th, 1999

 

In June of 1999, an intimate workshop was held at the University of Memphis focusing on the topic of Data and Information Visualization.  A diverse set of approximately twenty participants from the Cognitive Science and Computer Science communities were in attendance.  Participants included those who study basic cognitive principles of information comprehension, industry and academic based computer scientists who are developing sophisticated information-visualization tools, and industry and military personnel and statisticians who have specific information visualization needs.  The goal of the workshop was to provide a forum for exchange on the state-of-the-art research in each discipline, facilitate collaboration between researchers in these disciplines, and to develop ideas for new research directions based on needs that arise at the symposium.

 

During the workshop, participants described their recent research.  In particular, the cognitive scientists provided information about the cognitive basis for using visual displays, empirical and theoretical principles for display design, and visual illusions and other possible pitfalls in display design.  Computer scientists described new technology for data visualization, including the use of virtual reality, displays for very complex tasks, and a variety of applications.  In addition to presenting their own research, participants focused on some specific data visualization problems presented by some researchers from the Naval Personnel Research and Development Center (NPRDC).  In the closing session, workshop participants developed a taxonomy of data visualization tasks, outlined principles for data display that are currently known (based on cognitive research), discussed the latest in tools for data visualization, and listed goals for future research. 

 

Workshop organizers:

 

Priti Shah                         priti@umich.edu                        Dept. of Psychology, University of Michigan

Jonathan I. Maletic                         jmaletic@memphis.edu                         Computer Science, University of Memphis

 

For more information about the workshop, a list of participants, conclusions, and downloadable versions of a number of the presentations, see:

 

http://www.msci.memphis.edu/~maleticj/visual/

 

*This workshop was jointly supported by the National Science Foundation Information and Data Management Program (under the grant IIS-9907299) and Office of Naval Research (under the grant N00014-99-1-0522).  All opinions, findings, conclusions and recommendations in any material resulting from this workshop are those of the workshop participants and do not necessarily reflect the views of the National Science Foundation or the Office of Naval Research.