Wednesday, February 24, 2010

Q&A With Shawn Allen of Stamen Design

Shawn Allen is a partner at Stamen Design, a small San Francisco-based firm with big ideas. Stamen does lots of work with maps, data visualization, information trends, and collective think-tankery. Indirect Collaboration's Tim Lillis interviewed Shawn about Stamen's process and philosophy in February 2010.

Lillis: Stamen Design seems to occupy a space somewhere between a design firm and a think tank. Are your research projects driven by your commercial work, vice-versa, or other?

Allen: Probably more vice-versa. I think of most of our work as "research" in the sense that we often don't really know what we're building when we start. Through the process of investigating the data and visualizing it for ourselves, we come up with something that our clients can either (in the case of more commercial projects) provide to their users or (in the case of our more think-tanky work) use to learn something interesting about their own information. The really good projects provide us with an opportunity to play with some new technology or investigate an idea that's been rattling around in our heads for a while. If we don't find that opportunity in our commercial work, though, we usually end up just building it for ourselves—this is how Crimespotting came to be.

Lillis: Crimespotting was a big hit for you guys. Can you tell me more about its inception?

Allen: Oakland Crimespotting was originally a research project of Mike Migurski's. He was laid up in bed with a bad back and wanted to figure out a way that he could get useful information out of his city's official crime map. This thing is awful. You have to click through at least five forms until it even shows you a map, and when it does the data is often filtered down to too small of a subset to be useful. So he set about building an application that would basically submit every permutation of the application's forms, grab the images generated by each submission, find crime icons on them then put the locations and other metadata into his own database. In the process of creating interactive maps to look at the data he was collecting, he also built the first version of Modest Maps, a open source map interaction library which we now use for pretty much every single one of our map-related commercial projects.

Lillis: What are some of the public conversations that have opened up as a result of Crimespotting?

A lot of people see Crimespotting as a shining example of open data, and a way for citizens to inform themselves about their community. During the site's first year we got some great feedback from Oakland residents who were bringing spreadsheets generated by Crimespotting to their regular meetings with local police officers and asking them what they were doing about, for instance, the recent rash of auto thefts in their neighborhood. Crimespotting armed these citizens with information that they used to have to rely on the police to get.

It's been a while since we've heard any more of those stories, though. Most mentions of Crimespotting that we read on the internet these days simply marvel at the number of dots on the map. Crimespotting does much more than just show you points on a map, though. You can get RSS feeds for crimes in your beat. We've applied for a grant to invest some serious time in Crimespotting and turn it into a site that engages journalists and (hopefully) fosters more direct civic engagement.

Lillis: Many of your projects depend on information coming from somewhere else. What has your experience been with government sources vs. citizen, or crowdsourced data?

Allen: My experience has been that just getting the data in the first place is the most difficult part of the process, regardless of the source. Crimespotting's original Oakland manifestation was an exercise in freeing a source of data that had never before been made publicly available, whereas the San Francisco version was built in less than a week after some very nice municipal employees who'd been tasked with opening the city's data provided us with a KML feed. The devil is in the details, though, and we often spend the entire duration of a project working out the specifics of data formats, timeliness, and completeness with our clients. We still don't have homicides in our San Francisco crime feed, for instance.

Lillis: In your In The News project, it seems that at some point you had to shut it down because you had too much information. Did you think about adding filters to create a unique experience for each user? Are there other ways you considered dealing with this surplus?

Allen: I actually just started working with Mike and Eric after they'd finished In the News, so I'm not qualified to answer that first question. But generally speaking, yes: while our first inclination—and our preference, I think—is to show everything, there are indeed data sets simply too large or complex to be visualized usefully in their entirety. One of the things that I think we do best is create interfaces that allow the user to filter data down into subsets that are manageable. It's important to build tools that can be played with and manipulated easily and in realtime. Those interactions are what help people discover new and exciting things at their own pace.

Lillis: I think we all have assumptions that we make about what "the crowd" is doing or thinking. Were you surprised by some of the patterns you saw emerging through your projects?

Allen: Absolutely. When I was working on the Digg Labs pieces I was constantly surprised at all of the weird stuff that people were submitting and digging. The dog pile effect of particularly big stories—which, in the world of tech, means events like the iPhone announcement and the AACS encryption key controversy—was pretty shocking, and a lot of fun to watch. Some stories broke on Digg before they broke on major news outlets, and it was fascinating to watch the conversation around them develop in this totally organic environment. It felt especially voyeuristic before the tools launched, too, because nobody knew that they were being watched like that. For a brief period we toyed with the idea of building versions of the visualizations that would help Digg find bots and track other abuses. But the public visualizations ended up being much more interesting and buzz-worthy.

Lillis: In the projects where you're collecting live data, have you witnessed people "playing to the room," where they seem to have changed their behavior because they know they're being monitored?

Allen: No doubt. Some Digg users dugg so many stories that their dots on Swarm turned into giant yellow orbs bigger than the stories themselves. A couple of people even posted videos on YouTube of their activity making the visualizations do weird stuff. Some stories blew up so quickly that they took over the screen, as was also the case during the MTV VMAs last September, when Kanye  West stormed the stage and interrupted Taylor Swift's acceptance speech. At the height of that controversy there were thousands of mentions of Kanye on Twitter every minute, and a significant portion of them also happened to use the word "asshole":

Lillis: In some instances you're pulling in multiple data sets, have you had occasion to combine these to create or offer something you weren't expecting?

Allen: We've done some pretty cool stuff with a group called MySociety in the UK that cross references multiple data sets. Tom Carden created the first of our interactive travel time maps, which overlaid the shape representing how far you can get via public transit within a given time period with the cost of homes in the same area. The thresholds for each variable were adjustable individually and in realtime, and the map showed you areas where the data overlapped—that is, where you could buy a house for less than £500k *and* get to work from in less than an hour. We later developed this into a slippy map which you could pan and zoom, and introduced a third variable: "scenicness" scores culled from a site that MySociety set up to crowdsource Flickr photo ratings that could help you filter out less visually appealing regions.

Lillis: On your site, you refer to your clients as collaborators, how important is this distinction in your work?

Allen: I would say that it's paramount. The tighter our connection with the client, the faster things happen. We appreciate that some clients are going to defer to us on every design-related decision, but the smart ones who can call us out and involve themselves in the process are typically more fun to work with. We thrive on fast-paced projects, rapid iteration, and constructive feedback. If we haven't spoken to our client in a week something's broken.