Data Portraits – background information and information in the background

V. Carpaccio's Portrait of a young woman rendered as a Data Portrait

(Preamble)

QR codes are images of data. They are digital entities that reside on analog materials. Furthermore, you can embed any kind of data you want in them (thank you Denso Wave!) up to 7kb numeric. I wanted to make images with these images of data. Not an image inside one bar code, but an image made up of thousands of bar codes each with discreet data. These images would contain the visual information that we interpret as humans, and they could also store a substantial amount of data that could only be interpreted by machines. The image (or actually, the material the image is printed on) is suddenly a random access floppy disk, a browser, a tracking device, a vehicle for presenting dense layers of intersecting information, and some of this information must be excavated with machines. This is where the idea for Data Portraits starts.

“Data Portraits” is a series of print installations comprised of data, images and 2D bar codes. It’s about images as machine-readable data repositories, installations as an abbreviated form of cultural quantitative analysis, and data re-framed as the surface of an image. A print acts as a browser, a hard-copy piece of tracking software that, over time, emerges as its own dynamic data archive.

As a static repository, a ‘portrait’ is made up of a data result set returned by an automated search using a range of publicly available data sources. I am currently exploring Google’s YouTube gData API and ngram datasets The data is embedded in the image as 2D barcodes.

As viewers begin to scan the codes with custom mobile apps, prints begin to accrue a usage history, generating a dynamic data set. This usage history is comprised of various data points pertaining to the print itself (which portrait was scanned, which code in the portrait was scanned, Time/GPS location of where the viewer was when they scanned a bar code) as well as basic demographic information, should a user choose to share this information. The collected data is also publicly available via a lightweight API (Application Programming Interface).

Data Portraits addresses several ongoing interests of mine.

Search

Cory Arcangel makes valid points in his Art Forum article regarding Web 1.0 aesthetics and openness, most of which I agree with. Unfortunately, we live in a Web 2.0 world as well, where the open web has become a sprawling landscape of pseudo-utopian, power-to-the-people, corporate-sponsored “social currency” sites, marketing schemes and so-called web services. The context wherein one begins to produce net art has completely changed. So how does that influence net/web and data art? 1st wave purists might say it doesn’t (or shouldn’t), or that work that’s susceptible to the tide is somehow suspicious. But this doesn’t get us anywhere. To remain rigidly nostalgic seems to miss out on the current landscape of the W-W-W, which can be frightening and also fantastically banal at the same time. Will Brand has an interesting take on the continuing web 1.0 work over at Art Fag City.

In spite of all the locked down web 2.0 enterprises out there, with its smug design aesthetic and insistence on a user experience fabricated more on marketing schemes than actual people, Search remains a large part of the pie for net-based artwork, largely because it’s so accessible. Along with the Web, though, Search has also evolved. Search, how background data is collected, managed and disseminated (how it is made visible to an end user) is often overlooked. Search could be seen as the underlying entity of the Internet, in terms discovery (‘finding anything you’re looking for’) with many end-points leading back to it. To discuss and engage in Search-based work without touching on the implications of its evolution, and in light of the many open source APIs that make large amounts of data available directly to an artist at the coding level, seems to gloss over many potential applications of Search.

Data Portraits attempts to engage in some of these complexities around Search. First I want to quickly reference “Follow my other Twitter”, by Cory Arcangel in order to highlight some differences. Posted as a Twitter search of twitterati trying to get people to “Follow my other Twitter”, Cory distills the inanity of the twitter platform one swift gesture (the entire work is essentially one hyperlink – href=”http://twitter.com/#!/search/%22follow%20my%20other%20twitter%22″). Works like this frame Search as a kind of manual curatorial process taking place in the foreground. It also calls into question the HOW people use the internet, and “social” sites in particular. And interestingly, Cory leaves ‘traces of the author’s hand on the work’, simultaneously referencing traditional art practices and minimalism while producing purely digital work. In Data Portraits, Search is similarly approached as a form of curation (casting a net and creating a context is how I’m loosely defining “curate” here), but the similarities and there. Search is fully automated with code and subsumed as a background layer by utilizing an API rather than the user-facing search box. While a print is still a curated search at its base, the process and the authorship of that search is obfuscated, even when the search queries are presented with the work. The systems the work is based on and and how they are accessed are rather more baroque in nature than the minimalism Cory suggests.

Because Search is now a background process driven entirely by code (php in this case) and a public API, issues of technical limitations enforced by Google come into play, and so a small amount of hacking is required to achieve the desired data results. For example, the RESTful JS API only lets you grab 100 results at a time spread out across 10 pages. This version of the API is intended for light-weight use on the interface level (meaning accessible without having to use server-side technologies). The more robust gData API allows 100 results per page and up to 10 pages. Since an image contains at least 1200 QR codes, each containing a search result or ID, these two solutions meant a lot of tedious copying and pasting the max results from each page and collating those into one set. No thanks. I ended up using a simple AJAX workaround that returns the full thousand results from each page and collates them for me. After 10 consecutive AJAX calls to the gData API (each returning 100 results), Google actually cuts you off.

AJAX (Asynchronous Javascript and XML) is definitely Web 2.0 territory, along with the gData API.

Public v. Private spaces

Much of the current post web 2.0 landscape is glutted with subscriber-based “Social Networking” services. Facebook and Twitter requires would-be users to sign up before accessing the real Facebook and/or Twitter experience. In fact, it’s difficult to get around the web these days without giving up some personal information in order to get past the gates. It’s sadly true that large tracts of the web now resemble gated communities, so the reality of an “open web” is elusive.

Data Portraits emulates this gated presentation with users by exploiting a public and private version of the total experience. Anyone may view a print, walk away, and be done with it. Anyone may scan any of the codes with any scanning software they choose. However, mobile bar code scanners at large are unable to make use of the bar code data because of how the data is formatted (a cryptic pipe delimited string as opposed to the ubiquitous URL string). And this is where the mobile app comes in and essentially forces viewers from the public Data Portrait to the private one. The mobile app is free and available on the Android Marketplace (iPhone/iPod and BlackBerry apps are in the works). It streamlines the scanning process by automatically parsing the data and retrieving the result in the app itself, so the user never needs to leave the app. It also provides a history feature so you can keep track of what you’ve scanned and see those results again without having to find and re-scan bar codes. The “full” Data Portraits “engagement”, to use a popular corporate-speak term, is predicated on a gated model of interaction. Viewers must acquire additional mobile software and approve data gathering permissions presented in that software in order to fully interact with the prints. Only when a user succumbs to this acquisition and agrees to give up certain rights can they come into the fold. The latter part, giving the app permission to collect and share data, brings me to the Analytics and Tracking aspects of Data Portraits.

Analytics & Tracking

In spite of the prevalent gated nature of the “open web”, it’s impossible to get around without leaving behind a transactional footprint, a trail of behavioral information. The highly tactical and often anonymous gathering of information by so-called “white label” entities is both fascinating and problematic. Corporations would have you believe that we are in an age where consumers have influence over the brands they use, and that we define and influence these brands and an evolving collective intelligence by engaging in EPR campaigns and sharing our information and ideas as a series of fragmented communities and posts.

This reminds me of a quote by Mark Twain. “…If man could be crossed with the cat it would improve man, but it would deteriorate the cat.” Swap out “man” with “The Man” here. As we feed our subjective likes and dislikes, our perceptions of our surroundings and cultural awareness into the “computing cloud”, we are continually reduced to an exercise in business analytics and “success metrics”, ultimately generating a culture of Brand.

I think this is the crux of the 1st wave data artists’ argument against web 2.0 concepts, and this I can get behind. So, of course, Data Portraits, and the mobile app that comes with it, collects user data and scans and spits it back out as part of the physical installation, and also provides a lightweight RESTful API so other artists may use the scan data (not the personal data) in their own works. That last link returns a JSON string of 1 scan result, sorted by series. Results can also be returned in XML or HTML format.

So really, this projects fully resides in the post 2.0 world, and behaves accordingly. I say there IS a place for exploiting some of this new technology.

Transcription

In earlier work I often attempted to transcribe some form of data from one medium to another. With oscilloscope graphics, I transcribed audio-based information into real-time Cartesian graphs. In the video feedback work at Experimental Television Center, audio was again mapped to various parameters in the analog video circuit. With Data Portraits, there are several layers of transcription. There’s search query to code, where queries are transformed to the white noise aesthetic of the barcode. There’s also an exploration of the dynamics of the figure-to-ground relationship. The entire pixellated image becomes the figure to the embedded data as the ground. The ground leaves its mark on the surface of the image, but the data content remains unseen until activated by a viewer’s scanning device. And, because the viewer’s scanning interactions, or transactional behaviors, are tracked and stored, viewing trends are mapped to an image’s surface as well.

Degradation and generational loss

Most obviously, there’s the visual loss of information via pixellation. Details are obliterated in favor of data. Subjectively, the search query points to a devaluation of the subject in the image. Images and data do not always align in a predictable manner, creating a tension between what’s initially seen and what’s uncovered by viewers. From a software perspective, there’s the idea of deprecation, another term for the gradual progression towards obsolescence. As soon as an image is printed, it begins to deteriorate towards a potentially dead media, technologically speaking. As Data Sets shift, the original terms used to generate the ground become increasingly irrelevant. With the gData API, a YouTube video that was returned for, say, “McAwesome” may be replaced with a video that would be returned for a completely different search. The relationship between figure and ground shifts over time, and sometimes “dead pixels” form in the print when a video id is simply no longer being used. Furthermore, the life-span of the QR code is currently unknown. It may be around forever, or it may disappear in 5 years, or Denso Wave may suddenly decide to exercise its TM rights and start demanding payment from anyone using the QR technology. The point is, at some point down the road a Data Portraits print may very well be a piece of software without an operating system. In the end, the static object (the print) embodies the forced mutability of Image, Data, and Content

—————————

This work differs from the work of Web 1.0 enthusiasts in that it does not rely a frozen landscape, a contextual code freeze. It attempts to fully engage in all the new “advances” that have come about in the past 5 years. When the underlying software that the works rely on changes, these data portraits will only be a shell of what they once were and the bulbs that have been burned out won’t be replaced. Viewers may experience a search process leading to null pointers and empty interspace. And this is how it refers 1st wave net art, it’s still very much about exposing weaknesses in the system and hacking into the fringes of what is now the monolithic “open web”.