Tom Chi  

Interactive Visualization: the awkward years

July 6th, 2004 by Tom Chi :: see related comic

I, like many people, love information and data visualization. Sadly, it sucks. The examples of successful visualization are so few and far between that Tufte scours several centuries of material for a handful of good ones, and need only open the daily newspaper to get 50 bad ones. Now one need only add an interactive component to the equation and we go from a handful to essentially none.What’s keeping us from being able to do this well? Well to start, the field is extremely interdisciplinary. Proper data visualization, which allows many rich layers to be presented and interpretted simultaneously, requires a deep knowledge of visual design and a subtlety in execution that few have accomplished. Furthermore, such work work tends to require extreme technical skills to pull off non-standard representations and interactions. Lastly, a great interaction designer is needed to figure out the feel of the application. Do users feel lost in their own data? Do they understand the degree to which they are viewing vs. changing their data? How about the timing and feedback?

Even though HCI is an interdisciplinary field to start, interactive information visualization is particularly challenging because it doesn’t even have a good tool set to start with. In typical development situations one has buttons and dropdowns and dialogs, etc to build from, but when creating interactive visualization one typically starts with nothing. Worse yet, the subtle layers of color and form which are characteristic of the best data visualizations are not well preserved when those same colors and values are viewed on different monitors… (not to mention that they are inaccessible by design).

Despite these complexities, there are people working to make sucessful interaction visualizations possible. The prefuse project aims to create a powerful interactive visualization toolkit. Looking at the screenshots on the site, you’ll see something akin to the interactive trees and graphs that Xerox PARC touted with their work on hyperbolic trees. Sadly, they all have the same drawback of presenting text in an inherently unscannable and unreadable format (both unordered and dispersed in space).

The situations where I see interactive visualization making the most headway are not so much the visualizations of filesystems or email, but rather the visualization of things which are inherently physical to start with. Groups that work on creating visualization of morphologies whether they be topological or biological, are making significant headway because data dimensions map naturally to their spatial equivalents — so much of the data interpretation learning curve is automatically surmounted.

Sucessful visualizations also allow users to do things which simply could not be done before. Viewing a human brain in 3D with composite CT, MRI and PET data allows a doctor to non-invasively diagnose potential problems. Compare this to a system that lets you fly through your files… since there already exists plenty of ways to get to your files, designers have a high bar to overcome.

The last cluster of visualization tools that I see making some headway are the analysis tools which present primarily 2D data in realtime. Regular old charts and graphs can more than do the trick if you have the right data collection infrastructure. Sadly, for most organizations the right data is either hard to collect or is simply not being collected. One important exception to this is financial markets where stockbrokers have put together interesting proprietary tools to track thousands of indicies moving in time. If any of our readers work with these programs it would be cool to see some screenshots… :)

I think there are several reasons that interactive information visualization has not made its way into the mainstream of our daily work. First off, most of us don’t work on wall street or for NOAA. Much of our day to day data is verbal and symbolic and not inherently spatial. Thus visualizations which create spatial mappings actually introduce additional barriers to data understanding. Secondly there are not well established interaction metaphors for for interactive visualization (e.g. In some systems moving the mouse up makes the camera look down, in others the camera looks up). Thus most systems require significant training time/cost. Lastly the big guys are moving away from these approaches generally. The advent of search as a navigation metaphor has renewed our reliance on the keyboard and taken us a step away from visualization based approaches.

7 Responses to “Interactive Visualization: the awkward years”
iwolf wrote:

nice that you talk about interactive information visualization (because it’s exactly the field i’m working in ;-)

but what i have to add here is a little clarification on scientific visualization vs. information visualization. you talked about successful visualization as for example CT. in my understanding this type of visualization belongs to the field of scientific visualization where data that has a physical form is being visualized.

the characteristic of information visualization is its task to visualize non-physical information or data that has no clear mapping to the physical world like stock prices or projekt tasks. that’s why i think that your argument “but rather the visualization of things which are inherently physical to start with.” is not really valid here.

Tom Chi wrote:

Ah, cool. So according to that definition, “information visualization” is a term to be used only with data that has no inherent physical analog? I’m not a specialist in the field so I didn’t know there was any taxonomical difference between scientific visualization and information visualization. In my essay I use “information visualization” as an umbrella term when perhaps the more appropriate umbrella would be “data visualization”.

Anyhow, regardless of definitions, a lot of the same technologies and interaction challenges are broached by both. Both would certainly benefit from a flexible development toolkit and the establishment of solid, widely understood interaction metaphors.

reed wrote:

So where is the line between “imaging” and “visualization”? I think if we are trying to find ways to visualize abstract data well, we can take a lot of hints from the “imaging” side. For example I am a huge fan of the NASA Earth Observatory for it’s absolutely beautiful images. A good direction to go in if we want to develop visualization is to start with imaging and cartography — which map to real spaces — then move slowly from there into the land of the abstract by literally building the abstract symbols on top of the concrete maps.

Reed Hedges wrote:

Infovis Turbulence

From OK/Cancel (by way of UV): Comic Strip Tom Chi muses…

iwolf wrote:

yes tom, there is a difference between scientific visualization and information visualization (at least as it is used in literature).

just and add-on from the book preview of “The Craft of Information Visuaization, Readings and Reflections” by B. Bederson and B. Shneiderman:

“In contrast to scientific visualization, which is primarily concerned with three-dimensional physical objects and processes such as blood flow, tornado formation, and protein structuresm, information visualization adresses abstract phenomena-stock market movements, social relationships, gene expression levels, manufacturing production monitoring, survey data from political polls, supermarket purchases-which are not always observeable as natural physical realites. While both kinds of data sets com from the physical world, instead of dealing with three-dimensional aspects, the users of information visualization tools are interested in finding relationships among variables, discovering similar items, and identifying patterns such as clusters, outliers, and gaps.
The interactive nature of information visualization stems from the use of powerful widgets that enable users to explore patterns, test hypothesis, discover exceptions, and explain what they find to others. Interacting with the data set gives users the chance to rapidly gain an overview, explore subsets, or probe for extreme values. Information visualization tools become telescopes and microscopes that allow users to see phenomena previously hidden.”

another definition by C. Chen in the editorial of the Information Visualization Journal (2002): “For many of us, information visualization can be broadly defined as a computer-aided process that aims to reveal insights into an abstract phenomenon by transforming abstract data into visual-spatial forms. … Information visualization traditionally focuses on finding meaningful and intuitive ways to represent non-spatial and non-numerical information to people.”

and yes, you’re right it is surely not the most important issue if the one belongs to the other or if they are separate entities. i’m not sure, if the term “data visualization” refers to the “mother” of both because this is mostly used to describe “tufte-like” static plots and diagrams.

an interesting article/presentation can can be found here:
(Karl Fast: “Information Visualization: Failed Experiment or Future Revolution”)

Dan wrote:

The problem with information visualization is that tools are missing that allow the practioner/data expert to easily pick the right algorithms to do the data analysis and visualize the results. Instead people often need to hammer together a home made solution, which doesn’t leave much room for consideration of HCI principles and good user design. Furthermore, the need to use preexisting tools does not leave much room for interactive experimentation, as more effort needs to be placed in building the system. Take a look at the Information Visualization CyberInfrastructure (IVC) at Indiana U. which tries to develop a framework that facilitates easier development/use of algorithms for InfoViz.

Reed Hedges wrote:

Infovis Turbulence

From OK/Cancel (by way of UV): Comic Strip Kevin Cheng says it ROCKS No Way says Tom Chi Kevin visualizes the future a bit Tom muses further about its suckage Also, is four years old now….

Leave a Reply

OK/Cancel is a comic strip collaboration co-written and co-illustrated by Kevin Cheng and Tom Chi. Our subject matter focuses on interfaces, good and bad and the people behind the industry of building interfaces - usability specialists, interaction designers, human-computer interaction (HCI) experts, industrial designers, etc. (Who Links Here) ?