In researching an essay on information visualization, I interviewed Marti Hearst. A professor in the School of Information Management Systems at UC Berkeley, Marti's background includes working at Xerox PARC. Learn more about her here: http://www.sims.berkeley.edu/~hearst/ While I'll try to quote as directly as possible, much of what follows is paraphrased (I can only type so quickly!) But I'm confident I'm getting the basic gist across. So, What I Learned About Information Visualization from Marti Hearst My first question to Marti was about the success of information visualization as a field. Because, well, you don't see many visualizations out there. Marti agreed, and commented that in her latest class on visualization, she encouraged students to emphasize any usability studies if any had been conducted. Unfortunately, for the most part, there has been little or no validation of information visualization applications. There are some success stories, most notably Smart Money's Map of the Market. The original idea of tree maps, developed by Ben Shneiderman, had a number of problems -- poor aspect ratios, no interactivity, it was ugly, and it solved a problem no cared about. The people at SmartMoney were able to take the fundamental core of this 2-d space-filling idea, and improve it by: correcting the aspect ratios (the elements are more squarelike, not long thin sticks); made fixed locations for the boxes, so that industries are always in the same area on the screen; not try to provide too much detail in the overview--just the green or red to show, generally, whether the market was up or down; relating the size of the box to a meaningful piece of data, a company's market share (the original tree map related the size of a box to the size of a file on a hard drive... who cares?); applying it to a very targeted task; choosing appealing colors; and making the interacting with it quite intuitive, through mouse-over and drilling down. Marti forecasts a significant change in how visualizations are approached. In the past, they've been treated as standalone applications, "Look at this thing! And how beautiful it is!" Where as the key for the future will be incorporating it as a small part in a larger system, integrating it with the rest of the interface. In doing so, this will require visualizations to seriously take the problem that users want to solve into account, a motivation currently lacking from many visualizations. A visualization that particularly impressed Marti, and which hasn't received much notice, is the spiral visualization used to represent serial and periodic data. The essay which describes it(PDF) explains how it's been used to track food consumption by chimpanzees, for which there is a remarkably complex data set. Though perhaps quite obtuse to the layperson, Marti believes that for the scientists studying this phenomenon it's an innovative solution, using data tailored to the domain. An intriguing application of a spiral visualization can be played with at Rhizome, the net.art community, where you an use a spiral to navigate their article database. One of the most common applications of information visualization is searching and browsing textual data (as I discussed yesterday). Marti, who has focused a lot of energy on this specific problem, is coming to the conclusion that very little works. She has a class lecture devoted to the topic (PowerPoint), a section of which is called "Why Text is Tough", and which points out that "abstract concepts are difficult to visualize," that "language only hints at meaning," and that categorization is insufficient in understand the meaning of documents. She pointed out you could use visualizations for data mining applications on text, but that's a very different task. You're not trying to understand relationships so much as analyze a concordance. When discussing how visualizations might be used by the intelligence community (particularly in light of the Current Situation), she commented that in her initial conversations with folks in the IC, it's clear that have a lot of other issues (mostly organizational) to tackle before visualizations become at all useful. As an aside, she mentioned that the IC folks are researching software used by casinos to track individuals. It seems that casinos have some of the best intelligence software out there. Biotechnology is considered a prime market for visualizations, and while Marti believes it is likely to be so, she spoke of a discussion she had with someone at Genentech, who have "the fanciest tool ever," but no one uses it because it's too complicated. In no uncertain terms, Marti believes that the future of visualizations does NOT lie with 3-D. (An interesting counterpoint to another researcher, Mary Czerwinski, who states in an interview on InfoVis, "I am predicting--I would even bet money on it--that you'll see more 3D environments.") Instead, Marti believes that we'll see animation used more, taking advantage of humans' natural understanding and reaction to motion. I'm sure there's a lot more Marti could have said, but we had to stop at some point, and this was it.
11 comments so far. Add a comment.
Previous entry: "Communication, not Content." Next entry: "Interview with social network researcher Valdis Krebs."
Comments:
COMMENT #1 Boy genius Ben Fry has done a couple things that attempt to visualize text-based data. he's also worked a lot on visualizations of Genome data. See http://acg.media.mit.edu/people/fry/ for his stuff.
Posted by Andrew @ 06/26/2002 06:12 AM PST [link to this comment]
COMMENT #2 and here's another visualization, that i had wanted to include, but couldn't find before: TextArc More texty visualizations.
Posted by peterme @ 06/26/2002 08:23 AM PST [link to this comment]
COMMENT #3 Plumb Design Visual Thesaurus Imaging words, floating in space, your slightest click pulls to the front a concept and shows the links to other words.
Posted by Kenneth Hunt @ 06/27/2002 10:45 AM PST [link to this comment]
COMMENT #4 I agree that 3D is NOT the way to go. I;ve been saying that since Jaron Lanier first opened his mouth - years ago. Another assumption you can make is that ONE system or visual metaphor will probably NOT solve ALL issues, problems, navigational challenges, etc. So that means: - that it's gonna be some combination of visualization, visual metaphors, charts, what-have-you
- with good old fashioned directories, categories, lists, headlines, etc.
- with some new combinations of user experiences, interface controls, sequences of animations and/or video and (heaven forbid) something else that's just not clear yet!
- :-) There's only one thing that completely clear - it ain't gonna be in HTML! :-)
Posted by Marc Canter @ 07/01/2002 12:59 AM PST [link to this comment]
COMMENT #5 just thought i'd link bewitched :) also found this!
Posted by kenny @ 07/01/2002 08:33 PM PST [link to this comment]
COMMENT #6 A common thread here is the work of Martin Wattenberg, who not only deserves credit as the implementer of the beautiful Map of the Market (he's still working at Smart Money, I believe), but also did the Rhizome.org visualization mentioned by Peter. In addition, he did the site bewitched.com cited in the comment above. Martin is a prolific producer of first-rate data visualization. Another work of his worth mentioning here is the "Shape of Song" code seen at http://www.turbulence.org/Works/song/gallery/gallery.html. PS: From a self-promotional point of view, I've had some good fun lately visualizing networks of open-source cooperation and inheritance resulting from a programming contest. See http://www.mathworks.com/contest/molecule.cgi/final.html and http://www.mathworks.com/matlabcentral/fileexchange/Files.jsp?fileId=736 for more info.
Posted by Ned Gulley @ 07/01/2002 08:59 PM PST [link to this comment]
COMMENT #7 I'm a big fan of Martin, and have exchanged email with him over the years. It turns out he's now working at an IBM lab in Cambridge on this stuff... I guess he's left SmartMoney to pursue some purer research!
Posted by peterme @ 07/03/2002 08:47 AM PST [link to this comment]
COMMENT #8 Martin Dodge, Researcher Centre for Advanced Spatial Analysis (CASA), University College London (United Kingdom) made a presentation at last years Content Summit. He is the author of the cyberatlas and a data visualization heavy weight. Martins presentation report is at http://www.contentsummit.com/conference/eventdetails01_day3.php#1370 . The presentation with lots of illustration can be downloaded at the same place. Tim Bray is a co-inventor of XML and is producing visual data mining applications. His map of the internet has been around for a while and can be accessed at http://www.antarti.ca
Posted by Norbert Specker @ 07/11/2002 02:03 PM PST [link to this comment]
COMMENT #9 From a self-promotional point of view, I've had some good fun lately visualizing networks of open-source cooperation and inheritance resulting from a programming contest.
Posted by zip code @ 02/12/2003 02:26 PM PST [link to this comment]
COMMENT #10 From a self-promotional point of view, I've had some good fun lately visualizing networks of open-source cooperation and inheritance resulting from a programming contest.
Posted by zip code @ 02/12/2003 02:27 PM PST [link to this comment]
COMMENT #11 From a self-promotional point of view, I've had some good fun lately visualizing networks of open-source cooperation and inheritance resulting from a programming contest.
Posted by zip code @ 02/12/2003 02:28 PM PST [link to this comment]
Add A New Comment:
|