ARTstor: the fine line between transformative resource and expensive doodad

This is the incomplete tale of good people whose hard work could fail to accomplish much. (Anybody who has served on a committee will recognize some elements of THAT archetype.) This version of the retelling is particularly sad, because the work that's being done is terribly important. Like a farmer with a bumper crop of wheat who cannot deliver his surplus to starving regions to the south because he only knows how to ship his grain via a westbound train, so now are the well-intentioned people at ARTstor managing to miss their mark within the academic community.

Background
ARTstor is a non-profit organization funded by the Andrew W. Mellon Foundation. Its core activity since it was created in 2001 has been the creation of the Digital Library Collection. In the intervening span, they've managed to collect several hundred thousand digital images of art objects with their corresponding meta-data. That's an amazing accomplishment by itself.

Why would they bother to do such a thing? The advent of viable digital formats for all sorts of media is a boon for educational institutions. The analog/physical world of traditional libraries has a lot of attendant problems; most notable among these are access and preservation challenges. The Mellon Foundation had come to the rescue of libraries before with the wildly successful JSTOR project. By digitizing the back issues of hordes of scholarly journals (431 as of July, 2004), academic libraries were suddenly able to subscribe to a gigantic reservoir of full-text articles. Their patrons gained greater access, while the libraries were now able to make selective decisions about which journals to preserve in bound form on their shelves.

A similar problem exists in the area of images. Most schools that teach art history have assembled enormous collections of images to support the teaching needs of their faculty. These images are usually in the form of 35mm slides. While slides are a high-resolution medium, they have problems. The most common problem relates to preservation—they slowly turn magenta after a number of years. They also have the inherent problem that their physical form curtails access by multiple simultaneous users. (You know, like students.) Furthermore, the slides are kept separately from their meta-data, which hampers browsing, finding, and research.

Even a small College like Swarthmore has 200,000 slide images, with an annual accession rate that averages around 10,000 images per year. Building the digital collection of our future is not a small problem, even if it is one that mostly affects a small department. Consider also that many of our other programs would benefit from a greater ability to build or access image collections that connect to their own teaching and research.

The academic community has largely acknowledged that our future with respect to collecting images will be done predominantly in digital form. The hardware and software tools necessary to scan, photograph, manipulate, catalogue, and display digital images have collectively reached a level of quality that is appropriate for the academy. The bottleneck is the re-creation and re-description of the images themselves. If Swarthmore only has the resources to manage the accession of 10,000 new images a year, how will it ever keep up with new demand while converting its previous twenty years of collected images?

This is the gap that we all hoped ARTstor would fill. By offering colleges and universities a collection of several hundred thousand images of canonical (and semi-canonical) works, we could be given the head start we so desperately need. Then, rather than collect everything from scratch, we could focus locally on converting and acquiring those images that were needed to address the curricular specialties of our programs. The alternative is a dreadfully inefficient process in which every institution scans or buys the same image of Renoir's bathers and Michelangelo's David, piece-by-piece.

This is the information age, so we shouldn't have to resort to medieval methods for collecting.

The Hitch
The education community needs an ARTstor collection that can serve as the core of our collection development strategy, whether or not that is what ARTstor intended to create. Of course, ARTstor can't be a replacement for institution-specific collections. No matter how broad the ARTstor collection, it cannot provide adequate depth to support all courses. If you're teaching an upper level course on Buddhist temple architecture or Islamic painting, you're not going to find sufficient materials in ARTstor for your lectures and student study sets. ARTstor also has a conspicuous absence of 20th century artworks due to intellectual property constraints.

For this single reason, we must insist that ARTstor offer interoperability with other digital asset management systems designed for image collections, such as Luna Insight and MDID. These are the kinds of applications that visual resource organizations use to construct their own teaching collections.

Teachers who teach with images have a variety of requirements for the asset management systems they use to organize and display images. (If you're interested, you can download a checklist of requirements we drew up for our product search at the Tri-College Consortium.) Among these, there's one overriding deal breaker: all the images for a lecture have to be sequenced and displayed seamlessly from a single source. Unusual exceptions can always be made, but no professor wants to toggle back and forth between multiple sources during a lecture. Teachers of art history move through slides on a slide carousel with practiced fluidity; their digital substitutes won't be used if they create frequent distractions for the instructor and students. The focus must be on the art, not on distractive technology used to show it.

At long last, this is the tragic flaw in our hero, ARTstor. ARTstor's database is not interoperable with other digital image management systems. They have created a database on our behalf, but have inexplicably locked it up in an information silo that makes it almost useless for its most important consumers. You can do everything with ARTstor images that you need to do for most teaching purposes, with the single, depressing exception that you can't include images from other image collections. Nor can you use your MDID or Insight client to perform a federated search across your local collections and ARTstor images. (The latter of these approaches is greatly preferable. Decisions about tools and content should be kept as separate as possible.) Consequently, it is still impossible for an instructor to sequence a simple slideshow of images from these multiple sources.

The situation calls for an open API that would allow any ARTstor subscriber to use its digital asset management system of choice to query the ARTstor image set and display resulting images side-by-side with locally collected images. Unfortunately, that is not the strategy that ARTstor appears to be pursuing.

The shame of the situation is not that ARTstor can't deliver their images through third party systems. Luna Imaging had a small pool of testers who were accessing the ARTstor collections through Insight, but I have been advised that this access was suspended recently. Other developers are willing and able to build the necessary enhancements into their software. The explanation I've been given is that ARTstor is concerned about legal issues that could arise by allowing this type of remote access. To this I can only respond that if they've just trying to figure out that problem now, they've waited years too long to start. It is a total failure to understand higher education's needs not to have built interoperable access into the specifications for the product at the outset. It's as if the organization has spent years trying to figure out how to deliver a gazillion images to its consumers without giving any thought to how the consumers actually use images. If that's really the case, ARTstor is a (non-profit) business with a bad business model.

Conclusion
Colleges like ours have been in a holding pattern for years because we were waiting for ARTstor, waiting for ARTstor, and still waiting for ARTstor. (Apologies to Samuel Beckett.) We can't afford to wait another year or two for this problem to be solved. Everybody I've talked to in the visual resources and academic computing communities has expressed a similar perception that ARTstor has a lot of promise, but is painfully slow to deliver on it.

This interoperability issue is the whole ballgame. Sadly, ARTstor must resolve it or be irrelevant. Other application and business model developments are wasted efforts in the meantime until this barrier is removed. I sincerely hope that ARTstor can be persuaded to make this the first priority of the organization. Although I have been critical of their pace and direction here, I especially want to emphasize that the entire visual resources community is rooting for the success of their effort. How could we not? They're our reluctant champions in waiting.