6 posts tagged “research”

View other items tagged “research:” Photos, Audio, Videos, Books or Links. View “research” on Vox.

Grand Challenges in Vis

What are the Grand Challenges in (Info)Vis? was the topic for a panel earlier this week, feat. Georges Grinstein, Tamara Munzer, and Daniel Keim.


I won't resummarize here, but Tamara made a call for a vision of "total political transparency" as a problem that is infovis-complete, but many of us agreed we're far far away from that. Sure, we have the Sunlight Foundation and other efforts, but we're still drowning in a sea of data, full of schema borne in different oceans.


From a Grand Challenge perspective, InfoVis panelists often cited scalability and data integration as large concerns, but didn't unpack scalability in any way at all. I think from my perspective, we're at a point now where it's not that the number of data dimensions (dimensionality) is growing without bound, but rather that the number and size of heterogenous data sets is growing faster than our ability to make sense of them.


Traditional InfoVis or even Vis assumes you have a data set that may grow and get larger, rather than envisioning a world in which you have an ecology of data producers and consumers, schemas and lack thereof, bad data and missing data and the like.


Alon Halevy made mention over the summer of OpenII, an Open Source Information Integration suite, and there are efforts such as Freebase which try to collabo/crowd organize the world's data, but I think even for the space of Vernacular Visualizations, of spatial mapping and timelines, of health / finance and politics, we're far from a world where we can explore the world's heterogenous data space to make sense of it, and communicate subset views of the space to others.


Ah well, time to get to work :)

Post a comment Tags: research, vis

InfoVis 2008

InfoVis 2008 wrapped up yesterday, here in Columbus OH, and I have to say that the conference slightly restored my faith in InfoVis, the field and more specifically the research community that breathes the field alive.


But what is InfoVis anyway? In Stu Card, Jock Mackinlay, and Shneiderman's Readings in Information Visualization, Card et al. define visualization as


``the use of computer-supported, interactive, visual representations of data to amplify cognition'' (p. 7, Table 1.1) 


There are at least three things to note in this definition. First, that if the term "visualization" doesn't quite make sense, we can substitute "visual representations" and, at least to me, we can start to ground our thinking more concretely or perhaps more analogically. Just as words are textual representations, images embedded with some meaning can be thought of as visual representations, which can be chosen appropriately, manipulated, or explored in some way.

I'm not going to drill down into the differences between data, information, and knowledge, but I think the second point to note is that this definition focuses on the potential for "well-designed visual representations of data" to amplify cognition. At first, this may bias us towards thinking only of how vis. reps., let's say simple scatterplots, can help us as individuals to understand relationships between different data dimensions, but from a perspective of distributed cognition (see "Distributed Cognition as a Theoretical Framework for Information Visualization" in this year's InfoVis 2008 proceedings), the "unit of analysis should not be a human individual, but a cognitive system composed of individuals and the artifacts they use to accomplish a task".

From this perspective, we can think of the potential not of simply aiding analysis, but in aiding communication (see references of Communication-minded Visualization), whether simply unidirectional presentation or as a talking point which spurs discussion (and "social interaction").

Finally, I'm going to drill down on "data", noting that Card et al. define InfoVis (as opposed to general Visualization) as specifically "interactive, visual representations of abstract data". Card et al. divy up Vis into Scientific Vis (SciVis), vis. reps. of scientific data, typically physically based as opposed to Information Vis (InfoVis), vis. reps. of abstract data, nonphysically based data.

In fact, the InfoVis 2008 page suggests that the key challenge of InfoVis is in designing a "cognitively useful spatial mapping of a dataset that is not inherently spatial and accompanying the mapping by interaction techniques that allow people to intuitively explore the dataset". What I find problematic about this division is that it's hard to place everyday maps into this dichotomy. Maps are, of course, inherently spatial, and yet they are not always intended for scientific-minded analysis.

More generally, perhaps what made me lose faith in the field of InfoVis was the way in which it seemed the field "progressed" in the last ten years without having a clear impact on the world at large. I think part of that perception was assuaged when I heard Ben Shneiderman talk about treemaps and the decades it takes for ideas to be accepted (or not), especially ones that are directly user-facing (new machine learning algorithms go from Andrew Ng's lab to industry like 0-day exploits, they say). But beyond that, I'm mollified by more recent work on what some have called vernacular visualizations.
Vernacular vis was a discussion topic at last year's Vis -- this year, not so much.

Earlier this year, F Viegas and M Wattenberg followed up with an Interactions article titled "Tag Clouds and the Case for Vernacular Visualization". From what I recall, they didn't try to unpack "vernacular", but simply seemed to chose the term as a nod to the design world dichotomies of low-brow and high-brow. But if we drill down further and look at how Vernacular refers to "the native language of a country or a locality" (so says Wikipedia, some semblance of the Vernacular Web), then we quickly get into the concept of diglossia, of high and low languages, of high prestige / the elite / the learned and of the everyday / quotidian.

In our Vispedia work this year, we primarily supported three visualizations: maps, timelines, and scatterplots, which seemed in that order used (commonly) and in that order compelling and, dare I say, vernacular.

So what makes a visual representation vernacular? Both the design world high/low-brow division and the diglossia sociolinguistic view suggest that not only do representations / media shift, hustle and flow back and forth into the vernacular over the decades (jazz, comics, and video games let's say) but that these forms are specifically situated in some socioeconomic context according to some frequency / pop.-populist distribution.

While I think those dimensions are worth further analysis, seeing for example how NYTimes InfoVis design affects vis. literacy in the digi / technorati and beyond, I think if we go back to looking at what is being represented (and secondarily, for what purpose do we chose to represent?), we achieve some deeper insight into what may typically fall into ze vernacular.

I'll posit here that the more closely a representation maps onto our everyday dimensions of space and time, the more likely that visual representation falls into the realm of The Vernacular. Secondarily, the more a vis. rep. is well designed for highly-probable / quotidian STUs -- situations, tasks, and users (see Dr. Olsen's "Evaluating User Interface Systems Research" PDF), again the more useful a vernacular vis. might be.

Along these dimensions, academic InfoVis (with the exception of M Wattenberg's work, for example this year's Word Tree work or the beautiful geometries & aesthetics of Stacked Graphs, which both play with our common notions of textual and stacked-graph spacetime) tends to veer largely away from the vernacular, by dealing primarily with abstract, high-dimensional data. On the other hand, academic SciVis (and VAST, in a different way) veers away from more everyday STUs towards a service lens of scientists, engineers, and analysts.

Where does that leave us? Earlier this quarter I had planned out a schedule for a Vernacular Visualization workshop at Stanford co-hosted by Adobe and the Stanford Graphics / Vis labs, but peopletime constraints had us push back the date to some indeterminate time in early 2009. But I think Vernacular Vis has a bright future! from academics studying InfoVis for the People from a Comparative Media Studies (MIT) perspective to IBM ManyEyes researchers M Wattenberg and F Viegas, from a long history of digital and analog cartography to recent developments in making geospatial mapping technologies (at least on the consumer side, not quite the design side just yet) accessible to large fractions of the population.

I'm going to write a bit more later on how Vernacular Vis fits into the "Grand Challenge" view of the InfoVis world but for now I'll pause here and summarize. The history of academic Vis research has tended to split into two factions (now three, including analysis), one of scientific visualization (i.e., how do we see airflow over surfaces? MRI scan data?), and one of abstract information visualization (how do we see protein-protein interaction or social networks? how can we make sense of financial market data or large document collections?). 

But just as hypertext researchers moved on right before the Web grew up and actually caught on, perhaps we're at a point where a more visual and interactive Web is growing up and the larger young/educated population is (?) more visually literate, and yet it's not clear where the last ten years of SciVis and InfoVis has to say towards having impact on that population. Still, with a more recent push towards more accessible / Vernacular Vis, towards communication-minded Vis and Vis as social spurs & catalysts, I'd like to say with some mild hope that we're on a decent, forward track :)

2 comments Tags: research, vis, infovis

re:mix, re: Vispedia

Long time no blog. Have been training as an artisan (bread and occasional bagel) baker and playing some hereandthere high-level Ultimate.

Here's part of an email I just sent to the mouseHole scripters list.

Alas, I haven't had time to work on the mouseHole of late -- I spent some time instead exploring the possibilities of the curious Firefox extension POW, which I got to half-work as a pseudo proxy.

I'm not sure how much POW has progressed since I last looked at it, but I think in the long term, building on something akin to POW might be the way forward, even if Ruby isn't yet directly supported in most browsers. I described some of these concepts here in "Remixing the Web":

   http://graphics.stanford.edu/~lwu2/ (third paper link)

essentially trying to replicate a mouseHole-like programmable proxy within Firefox itself, driven by jQuery and JavaScript, which at least reduces the programming burden if you've read jResig's JS book and feel comfortable enough in that tongue with jQuery at your side.

Since that work, I have also heard about Jaxer but not read much about it. It might be what some of us are half looking for?

On the one hand, there's something _why elegant about the tight combination of mouseHole, Hpricot and Mongrel, but on the flipside, Firefox's core can support SSL, is HTML/DOM compatible with itself, and can make use of extensions such as POW, GreaseMonkey, and the like.

To reiterate, it's a common reductionist pattern to think of the alternatives as being either the browser as the point of modification (extensions) xor programmable proxies as the point of remix (for example, see Bolin's MIT thesis where he makes this exact bifurcation), but we're at a point now where we can have our programmable proxy cake and eat the browser kiddies too.

Does that help? Googling for "jaxer programmable-proxy" doesn't seem to show any hits right now, and while I've seen discussion of doing mouseHole-like programmable-proxy ish things in POW, I don't know if anyone else has followed up on that vector of attack, beyond the proof of concept we built for "re:mix", but to me it seems like the Right Enough way forward as we try to democratize these tools of production and ICT.

~L

In other news, the Stanford University visualization research group now has a new web presence at graphics.stanford.edu/vis, and my most recent collabo project, called Vispedia, also has a new project-page-home on the web at graphics.stanford.edu/projects/vispedia.

Finally, in putting together a brief list of my recent work as a CS PhD student here at TheFarm, I realized that four of my recent projects and my current "Sniff and Scratch" work all have the theme I briefly mentioned in my mouseHole mail, that of "democratizing the tools of production and ICT".

And while there is great work in the larger space of using technology in the journey towards social justice, and in reducing digital divides, I am sort of more interested in looking at technology with a social justice and democratizing lens, and beyond the analytics that are common in the history of science and technology, I think it's important to try to have time-flow-positive impact in the coming years and decades, in a constructive here-it-is way, in contrast to the retrospective here's how it was fifty 2 a hundred years ago.

Kay, in the spirit of my flexitarian with a vegan offset foodie philosophy of late, I'm going to make some food for lunch. Chow, ciao.

Post a comment Tags: web, research, vis

on proto types

I heard about Richard's K-Sketch from Mentalguy, who seemed to think it was as yet unreleased, but a CURIS undergrad and I were both able to get this kinetic sketch pad up and running on both Windows XP and Vista, with VS 2005 Professional and C# Express respectively.

I had thought of sending out the link to why's HacketyHack mailing list -- I guess I'll do that now -- telling folks to hack together K-Sketch with HH, to help motivate programming through storytelling.

srk has us video prototyping these days, as a way of quickly sketching user experience, and here's what I tried out.

Iteration zero wasn't a video at all, but rather a UX comic put together with a combination of my old Samsung 1.3M pixel cell phone, Flickr, and plasq's Comic Life (which comes with recent MacBooks).

I quite enjoyed iteration zero, as I was able to wander around PA snapping shots with my cell, sketch overlays with a sharpie, and then Photoshop/Seashore them together.

a UX comic life
a UX comic life

For prototype One, I mainly combined a PNG version of my UX comic with Keynote transitions, in addition to using online demos of Opera Mini, the 3GP video recorder on my old Samung cell, and the freely available CamStudio to tape a portion of a Photoshop screen, manipulated off(cam)screen with some PSD layer legerdemain.

CamStudio worked well enough on Windows, and the mobile phone video had the added benefit of quick turnaround (Bluetooth transfer) and a low-to-medium fidelity feeling to its digital film grain. I couldn't find a OS X screen (video) capture utility that I could quite trust, and ended up using Windows Movie Maker on Vista to splice things together.

Prototype Two was done mostly in Keynote, which seems to offer a nice suite of transitions and build-in / build-outs, but it didn't seem to offer the kind of motion paths one gets with K-Sketch. In concert with a Flickr (Creative Commons) search, this video prototype got built a lot faster. Unfortunately, exporting QuickTime from Keynote seems to look pretty jerky (on my 2GHz 2GB RAM MacBook Pro), and SWF export seemed to screw up the Comic Life speech bubbles which I had copy+pasted into Keynote. Sigh.

Prototype Three, the most recent one, was done half in Photoshop CS2 (thank goodness for Academic site licensing) and a licensed copy of Techsmith's Camtasia Studio. I highly recommend both Photoshop and Camtasia for video prototyping, in spite of their high (for a student) price points. CamStudio suffices for simple video screen capture, but Camtasia goes a step further and offers the ability to quickly edit the captured footage: cutting out unnecessary footage, inserting captions and focus points, zoom+pan and the like.

I don't have first-hand experience enough to report on any FOSS video editors or compare them against iMovie / Windows Movie Maker or even Final Cut Express, but I'm guessing non-gratis, non-libre software wins there again.

Is there any academic work on video prototyping? I'm not sure, I have yet to look, but I feel as if there ought to be a useful point to explore in that space, beyond bpb's DEMAIS and Richard's K-Sketch, somewhere in between Expression Blend, Keynote, Photoshop, and Camtasia too.

K-Sketch might not be a bad place to start, although it depends on MSFT's Ink SDK, and it's hard to say how painful it would be to port K-S's C# to Mono.

A friend of mine working the graveyard shift in Beijing recommended me a book, titled China, Inc. -- oh the wonders of the Internets and IM -- and in looking at Bunnie's reports on Chumby things Made in China, I'm glad I don't have to directly compete with that large sliver of the economic world.

But back to K-S and HCI. How much would K-S benefit from layers and/or the ability to import PSDs/PNGs? I don't know, but a K-S-enabled HH would be pretty sweet.

I've been reading Bonnie Nardi's Small Matter of Programming once more and I highly recommend it. I have also been reading Cypher (ed) 's Watch What I Do: Programming by Demonstration again, and getting up to speed on theories of all sorts.

I'll end on some recommended links for you to chew on...

...and some papers too.

  1. Harel, "Statecharts: A Visual Formalism for Complex Systems" (1987).
  2. P. Baudisch, "Summary Thumbnails" (2005)
  3. S. Baluja's recent work at GOOG on the Mobile experience
  4. My recent faves: Mary Czerwinski, Bonnie Nardi, and Eszter Hargittai

 

Discbee #10
Discbee #10

Post a comment Tags: research, hci

Nok-i-a

My adviser gave a talk at Nokia this week, and a few of the slides mentioned my recent work on EdgeRank.

(Not yet published, but I might as well mention the term quasi-publicly.)

I've been working more on this mashup research project this quarter, but I took some time to compare EdgeRank (as defined and implemented by me) to good ol' PageRank (as implemented by JUNG), for a specific data set, to see how they match up.

Let us ask, shall we, what are the relevant links on the Nokia entry on Wikipedia?

EdgeRank says:
    United States, Wireless, Apple Computer, GSM, multimedia, bureaucracy, VHF

PageRank says:
    United States, Germany, Finland, World War I, 1000000000 (number), Nokia, Tokyo

L.Wu One, L.Page Zero.

:p

(Okay, I'm biased and I got to make up the game... I think I'll try to chat with Terry or Hector about this particular lil experimental result tomorrow if they're around.)

Post a comment Tags: research

on learning

As a first quarter PhD student, they don't really tell us what to do. It's not a course-based program, as I tell the other PhD students that I meet, here at TheFarm and outside of campus.

Instead, we're supposed to do research. But what does this really mean?

I suppose I have been a little lost, even though I have written research papers and am used to getting things done on my own or in mildly organized program teams in industry.

I said hi to srk, and he mentioned once again how it's a good idea to work with someone else for the first quarter or so, just to ease into things and get to be part of the culture of the lab. That's what ph said too, but I guess I'm picky about ideas somehow.

srk lent me a short book titled "Situated learning: legitimate peripheral participation" by Jean Lave and Etienne Wenger, and suggested (as Rajeev did) not to worry about comps or requirements or all of that.

I guess it's an easy thing to worry about, as that's what students are used to doing, studying and learning enough to pass exams, but that's not why we're here, and it's hard sometimes to remember this.

Post a comment Tags: research
lwu

About Me

lwu
United States
View my profile
mellow and yellow

Neighborhood

Explore friends, family, friends & family, or entire neighborhood.

View my neighbors

Tags

View my tags

Archives