This this a proof of concept that works much like microsoft's surface display. The display is a little over seven feet tall according to the article and consists of two panels. One panel is a standard multi-touch LCD, and beside this is the holographic glass panel. While the LCD is used by customers for selecting and displaying video advertisements, the glass panel is used by customers for browsing special store promotions, or displaying a map guiding customers to special store sections.

The display is able to recognize the gender of the user and customize their experience with the device. For example, "customers may use the multi-touch holographic screen to explore merchandise, find out about promotions, submit feedback on products, read customer reviews, and view past purchasing histories." The device is also integrated with mobile phones. The display is also able to collect anonymous analytical data

An overview of research that has taken place related to large displays. The paper introduces the rise of new technologies that are enabling the increased consumer uptake of large and multi-display configurations. The authours then continue to highlight the benefits of using LSIDs such as increased productivity, recognition memory, peripheral awareness, and even virtual world navigation. The paper ends with a description of some of the usability issues that are involved with LSIDs including: cursor tracking, accessing information such as icons and menus, window management issues, task management issues, configuration issues, and methods with which to utilize the periphery. This overview is an excellent gateway into the research that has been conducted on LSIDs.

Key Literature

Large Display Research Overview

2006

Toward Characterizing the Productivity Benefits of Very Large Displays

2003

Mary Czerwinski
Greg Smith
Tim Regan
Brian Meyers
George Robertson
Gary Starkweather

This is a preliminary user-study designed to investigate whether or not there are any benefits to using a LSID for office work. The researchers had participants carry out a memory-loaded task that replicated common office tasks on a 15” display and a 42” display surface called DSharp. 15 participants were chosen from Microsoft’s research pool, 7 of which were female. Despite using a standard windows GUI that was not optimized for LSIDs, they found a statistically significant decrease in task-times for users on the 42” screen when compared to users on the 15” displays. In both cases, the researchers observed usability issues. For the 15” screen condition, they noticed users were spending lots of time working with switching windows around, resizing things, and found that users would lose track of what was opened or closed. For the 42” screen condition, they observed that many users wanted to sit further away from the screen, losing track of the cursor, static menu elements were too far away, and users forgetting that they need to click on a window to focus on it before working with it. The researchers concluded that productivity increases with larger displays, the GUI must be redesigned to work with the unique environment that the LSID poses in order to take full advantage of the displays.

A Comparison of Techniques for Multi-Display Reaching

2005

Miguel A. Nacenta
Dzmitry Aliakseyeu
Sriram Subramanian
Carl Gutwin

The researchers in this study set out to compare different methods of moving objects from a small tablet-sized display to a large tabletop screen. Both screens could be activated with a digital pen. These screens were then placed in two different configurations. For the first, both the tablet and the tabletop display were within arms reach. For the second, the tablet was within arms reach but the tabletop was not.

6 different techniques for moving on-screen items from a tablet to a tabletop display were investigated including: Pick-and-Drop, Corresponding-Gestures, Slingshot, Pantograph, Press-and-Flick, and Radar View. The participants were required to move items from one screen to the other using the techniques mentioned. In both conditions (within arms-reach and beyond arms-reach), participants were fastest with the Radar View technique. User reports also mirror this finding and 14 of 18 participants preferred the technique over the others.

Radar View is described as follows:

"The Radar technique uses a miniature representation (a map) of the surrounding environment (i.e., the tabletop display and the tablet PC). When the pen touches the object the map appears. The map is placed so that the position of the pen is the same in both representations (see Figure 6). The map provided is similar to that in Drag-and-Pick [1] but allows continuous positioning without distorting the shared space. When the user starts to move the pen (without lifting it from the tablet) a small line connects the starting point of the stroke with the actual position of the pen’s tip. The place on the map where the pen is released determines the final position of the object (see Figure 5)."

The researchers also suggest that:

1) In a spatially-aware environment, designers should consider Radar Views for multi-display reaching

2) If the environment cannot be made spatially aware, then Pick-and-Drop is probably the best option, as long as targets are within hand’s reach

3) Providing feedback locally (as in the Radar views or Pick-and-Drop) appears to be more effective than providing distant feedback (as in Pantograph, Slingshot and Press-and-Flick)

4) A control to display ratio of 1 (as in Radar Views and Pick-and-Drop) appears to be better than providing acceleration by decreasing the ratio (as in Pantograph and Slingshot)

5) Spatial manipulations should be used over pressure manipulation

6) Discontinuities in feedback, even if these are at non- critical locations, should be avoided.

Spotlight: Directing Users’ Attention on Large Displays

2005

Azam Khan
Justin Matejka
George Fitzmaurice
Gordon Kurtenbach

The researchers investigate a method with which to draw the attention of users towards a specific area of a LSID. The technique itself is exactly as it sounds, it dims all areas of the screen surrounding the area of importance. In order to test the efficacy of such a technique, the researchers hypothesized that users would be able to locate the cursor quicker in a spotlight assisted scenario when compared to an unassisted scenario.

The experiment they designed had the user standing in the center of a U-shaped wall-display. By varying the field of view that targets were located in, they were able to discern that as the targets were located in a wider and wider field of view, the spotlight outperformed the cursor more and more. With a small field of view requirement, the spotlight outperformed the cursor by a factor of 1.5 while with a large field of view requirement, the spotlight outperformed the cursor by a factor of 3.4.

Subjective preference also heavily favored the spotlight techniques over the traditional cursor technique.

The Perceptual Scalability of Visualization

2006

Beth Yost
Chris North

This research paper compares large displays and standard sized displays with two different types of visualizations: space-centric, and attribute-centric. In the space-centric scenario, the researchers had one large map of the U.S.A with all relevant attributes superimposed on top. In the attribute-centric version of the visualization, the researchers had multiple maps each with one attribute superimposed on top.

The researchers measured the task-completion times of participants while they answered a series of questions pertaining to the information presented in each visualization. The participants were randomly assigned to either a 2MP display or a 32MP display.

Yost and North found that accuracy was not affected by either the 2MP display or the 32MP display meaning that the visualizations tested are perceptually scalable. With regards to task-completion times, the 2MP display was actually faster than the 32MP display. However, the 32MP display had much more information displayed. When normalized for the amount of data displayed on each screen, the participants on the 32MP display were faster than the participants in the MP condition. A 20X increase in the amount of data presented to the users on the 32MP display resulted in a 3X slower completion time. When users were polled regarding their preference for each display, Yost and North found that users weighted graphical encoding more importantly on the smaller display and spatial encoding more importantly on the large display.

VisionWand^?: Interaction Techniques for Large Displays using a Passive Wand Tracked in 3D

2003

Xiang Cao
Ravin Balakrishnan

The authors detail a method for using a motion-tracked physical wand to control a LSID. While the most popular types of control for a LSID usually involve single-finger or pen-based system, they require the user to be right up at the display which is not always ideal when working with a LSID. Furthermore, they only allow users to interact with the display in a mouse-like fashion.

The paper details the benefits and limitations of current systems such as SmartBoards^?, Liveboard, and Stanford Interactive Mural. Furthermore, the system implementation, design principles, interaction techniques, interface widgets, are discussed.

Informal user feedback from graduate students at the University of Toronto indicates that the system is fairly easy to learn and use.

Sweep and Point & Shoot: Phonecam-Based Interactions for Large Public Displays

2005

Rafael Ballagas
Michael Rohs
Jennifer G. Sheridan

This paper details two methods with which we can use camera enabled mobile phones to control LSIDs in public areas. The two methods, point-and-shoot and Sweep, allow users to use their mobile phones like a mouse to select objects and move a cursor. With the point-and-shoot method, a grid system is established through visual codes on the display and when the camera is pointed at a given coordinate, the screen on the phone displays a crosshair and the corresponding segment of the large display. The second method, Sweep, tracks movement with the camera based on any continuous image. With the Sweep method, the user may point the phone at anything since movement is tracked and a cursor on the LSID is moved relative to the amount of motion detected by the camera.

The researchers found that technology available at the time of writing was insufficient to provide a satisfactory user experience. In the Sweep method, the mobile phones used were not powerful enough to process the data with low enough latency; this resulted in long task completion times. With the Point-and-Shoot method, task completion times were as good as using a standard joystick. However, the cameras in the phones were difficult to use and resulted in a high error rate.

The researchers conclude that while the two systems did not result in ideal results, they are a good proof of concept for future ideas. Perhaps using mobile phones to control LSIDs might be better now as the paper was presented at CHI 2005, mobile phone technology should have progressed quite a bit since the study was carried out.

Works Cited

Bezerianos, A., Balakrishnan, R. (2005). View and space management on large displays. Computer Graphics and Applications, IEEE, 25(4), 34-43.

Cao, X., Balakrishnan, R. (2003). VisionWand: Interaction Techniques for Large Displays using a Passive Wand Tracked in 3D. Proceedings of UIST 2003 – the ACM Symposium on User Interface Software and Technology, 173-182.

Czerwinski, M., Robbins, D., Robertson, G., Meyers, B., Smith, G., Robbins, D., et. al. (2006). Large Display Research Overview. Proceedings of CHI 2006 – the ACM Conference on Human Factors in Computing Systems.

Malik, S., Ranjan, A., and Balakrishnan, R. (2005). Interacting with Large Displays from a Distance with Vision-Tracked Multi-Finger Gestural Input. Proceedings of UIST 2005 – the ACM Symposium on User Interface Software and Technology, 43-52.

Yost, B., North, C., (2006). The Perceptual Scalability of Visualization. IEEE Transactions of Vizualisation and Computer Graphics, 12 (5), 837-844.

Research Literature Review

Our bibliography is being maintained at RefWorks at the University of Alberta - this needs a login to get to.

Literature Review Protocol

• Titles should be capitalized as in the work
• In the notes there should information about who added the reference, whether it has been read and summarized, and whether we have a copy. Eg. "GR added, skimmed, PDF saved"
• Where the reference is to a visualization project a image typical of the project should be uploaded.

-- GeoffreyRockwell - 29 May 2009

The BigSee - Environmental Scan and Literature Review - First Draft

Large displays are becoming more and more popular (Czerwinski, 2006). However, researchers have currently have not fully investigated how a big display affects the design of visualizations. Current research has either used a large display as simply a place to display things for many people to see or concentrated on the interface and control aspects.

While the latter is similar, the questions we are asking are slightly different. We will be investigating the following: How do large format information displays (LFID) affect visualizations of text? For example, it is possible that using a large display will enable us to gain new insights into written works. Also, what can we do visually on a large display that we cannot on a regular one? And finally, are large format displays used in a different manner than regular displays?

To provide several practical examples of our questions for example:

How would answering and sorting our e-mail look like when we are working with a LFID? If we had a whole wall, how would you use that screen real estate to handle all of your email.  If we take a novel, Frankenstein, how can the big screen help us understand the whole novel quickly? Since we have a whole wall,  how can we use it to allow multiple people to interpret one piece?

These questions are becoming important to answer as LFIDs are becoming more common.

Literature

(This section will be improved as there are about 30 papers I have yet to read that all seem to be relevant to our interests. They were all referenced in the Czerwinski paper.)

The research in this field is limited. Most of what is readily available deals with questions of user interface and human-computer interactions with a large display. For example, there are many papers that discuss different strategies for window management and task management or physical interactions on a LFID (Bezerianos, 2005; Cao, 2003; Malik, 2005).

One project specifically discussed the scalability of information visualizations. Yost and North (2003), found that increasing display size from 2MP to 32MP did not yield any increase in normalized task performance time; also, the scaling of the visualizations did not lower accuracy. Furthermore, they also noted a change in user preference. While on the 2MP screen, participants felt that an attribute-centric design was preferable to a space-centric one and on the 32MP screen the reverse was true. Yost and North (2003) concluded that while using a large display, it is spatial organization that takes precedence over the attributes presented on screen that matter more when attempting a specific task.

Projects

The following projects listed is selected from our environmental scan. All of the following projects are relevant to LFIDs. It may be that some are scalable to a large display, while others have certain control affordances that would be applicable on a LFID.

Oakland Crime Spotting

Oakland Crime Spotting is a visualization that allows users to visualize the latest criminal occurrences within the greater oakland area. The visualization provides a clock and calendar glyph in order to control the time span displayed by the visualization. This is quite unique and an excellent way to control time based visualizations. To be more specific, the clock and calendar metaphor for representing time is a well understood metaphor regardless of culture or language. It is also very easy to manipulate accurately due to the shortcuts provided for times such as “swing shift”, ”day”, ”night”, or “nightlife.”

The visualization may also be simplified as it can get quite cluttered. Specific types of crime may be omitted from the list according to the choices of the user.

Elastic Lists

Elastic lists is rather more of a principle rather than a specific project. However, it does exemplify many features that would work well on a larger display. At it’s core, it is a system used to filter out unwanted results. However, the controls also provide the user with a lot of information. For example, the size, brightness, and color of each selection correlate with the weights of metadata corresponding to the specific values. Furthermore, spark-lines may be added to each field within the list and information can be extracted at-a-glance. It is a wonderful way of incorporating a large amount of data at one locus; which i believe is important for a large display to prevent unnecessary physical movement of the user from one spot (visually or physically).

Notice that as each column is changed, the rest of the columns will update themselves to reflect the filtered data. In the images, when “United States” is selected, there are 80 male nobel laureates and 6 female; conversely, when “United States” is not selected, there are 175 male and 7 female nobel laureates. The spark-lines roughly show when the specific prizes were given out. This is quite a lot of information provided to the user without even looking at the results of the filtering.

The controls of our visualization must be intuitive to our users and yet very informative at the same time. The two example visualizations here highlight both aspects.

Similar Diversity

A text visualization of different holy books from religions including: Christianity, Islam, Hinduism, Buddhism and Judaism. All five books were visualized together rather than separately. The 41 most frequently mentioned characters in each book are listed alphabetically with the size of each arc above their names representing the frequency with which they are mentioned. Below each name are the most frequent verbs associated with each character. The arcs that connect each character are based upon the relatedness of their actions. For example, a darker, thicker arc will represent a higher level of similarity between two characters.

This visualization is filled with quite a bit of information and other information can be inferred from it. It also is necessarily large in order to display the smaller arcs visibly. The visualization succeeds because the scale is large enough so that it does not seem as cluttered as some of the other visualizations presented in this scan.

Trendalyzer

 Trendalyzer is developed by Hans Rosling and his organization, Gapminder. It has since been acquired by Google. This visualization is quite popular and has been featured in TED talks three times. At first glance, it appears to be nothing more than a simple scatterplot of data. However, the power of this application arises when data is tracked over time. The system can successively load multiple sets of data from different time periods and animating it in order to allow the user to see trends. The application takes complex data and simplifies it into an easily understandable visualization. The features of Gapminder are simple yet incredibly well made. As the system truly excels at showing trends, it must be seen in motion to be truly appreciated.

Trendalyzer represents the data that it is visualizing in a very simple manner. Each region is simply a dot. However, the creators of this visualization has mapped data onto every possible attribute of each dot. For example, the color represents the nation or continent to which the area belongs while the size represents the population. Both of these metaphors are very easy to grasp; for example, nations are often represented with flags of different color while size has always been related to amount. By choosing common metaphors, the creators of Trendalyzer ensure that the application is as universal as possible.

What succeeds about Trendalyzer is that it employs easily understood metaphors for data representation which allows it to be simple to understand yet powerful enough to be useful.

Newsmap

Newsmap is a news aggregator for Google News feeds. While it is again a simple visualization (each news story is represented by a colored block), it is quite powerful.

Each news story is given a specific color, while the entire feed is categorized by country. Also, the size of each news block represents the number of related articles with larger blocks for the stories with more related articles. The controls allow for very quick and easy comparison of current events in two or more countries. While it is possible to effectively use the visualization when looking at a single country or even comparing two countries, it is much too cluttered when attempting to compare more countries.

This visualization is particularly interesting to me due to the fact that it would work very well on a tiled wall display. It reminds me that we must choose a visual idea that compliments our hardware capacities in order to succeed.

Munterbund

This is an interesting attempt at visualizing multiple essays that are in a collection. While it is aesthetically pleasing to look at, it is much to complicated to use for detailed text analysis. I believe that the visualization would work much better on a larger display where the details of the glyphs can be appreciated.

The other issue with this visualization is the use of metaphors. Whereas the designers of Trendalyzer has picked excellent metaphors, the designers of Munterbund have not done as well. For example, in order to explain the visualization, the creators of the visualization put forth the following:

“...the essays are depicted as a pie slice that has been color-coded according to Johannes Itten’s color circle and whose width of the central angle derives from the number of coinciding words between the essay in question and that of the secondary author.”

This is a bad metaphor for the visualization as it requires a lengthy explanation. In order to succeed, I believe that our visualization must be simple enough to be understood with little or no explanation.

Software Studies Initiative

 The Software Studies Initiative (SSI) is a group out of UC San Diego. Led by Lev Manovich, they strive to investigate the ways in which software shapes the world. The work done by the SSI has heavily incorporated cultural analytics and large format visualization.

The techniques that they have used in order to transform qualitative data into quantitative data is quite interesting. The data measured and visualized in the arthistory.viz project is quite interesting and different from what most people would look for in a piece of art.

Going into depth and breadth at the same time, big screens and digitized collections allow us to do this.

Hardware

Allosphere

 The Allosphere is a multimodal display located at UC Santa Barbra. The display consists of two domes placed beside each other to form a sphere with a foot-bridge intersecting the center of the sphere. The sphere has been built for multimodal, multiperson presentation of data. While we cannot afford to build an allosphere for our project, the allosphere does remind us that representations of data do not necessarily need to remain only visual in nature. Audio can play a large part in the understanding of data, and does not interfere with a large format display in any way. Furthermore, the sphere is meant to be used with others. A visualization on a wall display stands in the unique position to be used collaboratively.

CAVE Systems

CAVEs (Cave Automatic Virtual Environment) are visualization systems that are made up of between 3 and 6 rear-projection screens arranged into a cube. The user, inside of the walls, wears a special pair of glasses that enable a 3-d view of the images projected onto the cube. The installation at the University of Alberta is a 3-walled system. With more advanced systems, the user is able to walk around the projected images.

The CAVE system would be quite unique to develop a visualization for. From the projects listed on the Advanced Man-Machine Interface Lab at the University of Alberta (http://www.cs.ualberta.ca/ammi/), most of the projects created there have been medical or engineering imaging projects.

There has been one project related to the humanities by Robyn Taylor, a performing artist that developed a stage performance utilizing voice-controlled graphics.

The addition of a third dimension also increases the challenge that we will face if we choose to operate on this platform. The

Principles:

A. Affordances

1. LSIDs are for coordinated views.

They can show many panels with different, but coordinated information. They make possible showing many views on the same phenomenon that are synchronized.

2. LSIDs are shared in time. 

They are in public spaces that are shared. They make possible the shared use by many people at the same time.

3. LSIDs show breadth and depth.

They can show all of the items in a dataset and all of each item. 

4. LSIDs are seen to be approached.

The way to approach is seen so that you can see at different distances through moving close and back.

5. LSIDs need not show control.

They aren't where you show the interface controls so display can be controlled, though you might show the selection choices.

B. Design Ideas

1. Design so that all relevant views are on the screen at the same time and coordinated. Design so that the arrangement and coordination can be altered to foreground different things.

2. Design for partial use and collaborative use by different people. Design so that different people might work at different distances and on parts of the screen.

3. Avoid pointers, labels, and shortcuts. Where possible show the entire dataset (a prospect on the whole) and entire records. 

4. Use design so that different aspects are seen at different distances. Design so that users see how to approach and how to back up.

Show information not affordances. Separate the control from the display. You shouldn't show menus, toolbars, and other controls on a LSID. 

Employ clear, logical, and universally understood metaphors when visualizing metadata.

Research Literature Review

Our bibliography is being maintained at RefWorks at the University of Alberta - this needs a login to get to.

Literature Review Protocol

• Titles should be capitalized as in the work
• In the notes there should information about who added the reference, whether it has been read and summarized, and whether we have a copy. Eg. "GR added, skimmed, PDF saved"
• Where the reference is to a visualization project a image typical of the project should be uploaded.

-- GeoffreyRockwell - 29 May 2009