May 1998

Field Work in the Tribal Office

Continued from page 3

By Robert Buderi

Branching Paths

as parc's anthropological studies grew, they branched out along two main paths: workplace studies aimed at designing new office technologies and parallel attempts to strengthen internal Xerox operations such as service groups. Suchman heads the first effort, which really got going in 1989 with the study of two airline control operations, handling gate assignments, meals, luggage, and the like, at San Jose International Airport. In the early 1990s, a second project was initiated at a Silicon Valley law firm.

While the airport project involved detailed observations of the workplace, its successor at the law firm marked an early attempt to design technology based on what the ethnographers observed. (Suchman and her colleagues call this "case-based prototyping.") The two-year effort focused on "M," an attorney whose file cabinet held records that served as templates for drafting other documents.

After pondering videotapes of cabinet-scouring lawyers in action, Suchman and colleagues Jeanette Blomberg and Randy Trigg copied, scanned, and digitized 862 documents, about a quarter of M's cabinet. Computer scientist Trigg then led a collaborative effort with other parc researchers to design and build a prototype search aid capable of retrieving not just text but document images, which could be presented as thumbnail reproductions spread across the electronic "desktop." Although the shrunken images were illegible, searchers could find the right ones by recognizing a letterhead, or even the pattern of words on a page-much the way attorneys might leaf through a pile of papers without reading each one.

Although the search technology was improved several times in consultation with subject "M," it was never intended as anything more than a prototype. In the next field study, however, which started in 1996, the researchers were attuned to commercial possibilities from the start. The "tribe" in question this time is Caltrans (the California Department of Transportation), which is designing a replacement bridge to span the Carquinez Strait at the northeast end of San Francisco Bay.

The parc team on this project, set to last through 1998, includes anthropologists Suchman and Blomberg, joined by computer scientists Trigg and David Levy. Their efforts focus on a contingent of about a half-dozen engineers based at the Caltrans district headquarters in Oakland, across San Francisco Bay about 45 minutes northeast of Palo Alto. To complete its work, however, the Caltrans group must interact with consultants, contractors, and public and government organizations. So the Xerox contingent has dutifully followed them through the urban jungle to a variety of meetings and made extensive visits to the towns bordering the planned construction site, another half-hour north of Oakland.

The Caltrans engineers assemble information from all these venues into project files kept in three-ringed binders. The challenge for Xerox is to move this diverse body of graphical, printed, and hand-written documents (engineering drawings, maps, surveys, letters, memos and more) from the paper world into the digital domain. To that end, parc staffers help scan and digitize these documents, then provide technologies for indexing, accessing, and viewing the data on a web-based interface.

This digitizing may seem easy. Every office has a scanner these days, right? It's not. Scanned documents are typically converted from bit-mapped images to text, allowing users to cut, paste, and perform other word processing tasks just as if they had been created on a computer. However, the Optical Character Recognition process that makes this possible cannot handle drawings, photographs, and handwriting, and the Caltrans records are loaded with all three.

To get around this problem and give Caltrans access to all its records in digital form, the parc team created a hybrid application called the Integrator that allows the engineers to search, retrieve, and peruse both text and image-based documents together. While the Integrator incorporates the novel thumbnail technology prototyped in M's records, what really makes it unique is the ability to search for images based on features such as signatures or letterheads. Suppose, for example, that an engineer needed to know the names of everyone with whom she had corresponded in a given month. The Integrator could retrieve all the letters bearing her signature and also provide a summary that listed all the addressees. With a click of the mouse, the engineer could call up and print any of those letters, or send them over the web to a colleague. In late 1997, the parc team filed patents on the image-based search and summarization features.

For Xerox, beyond helping a major customer (the state of California) maintain records, the Caltrans arrangement provides a real-life test bed for a technology aimed at a variety of products. The workplace study is done in close collaboration with Xerox's Office Document Products Group, which hopes to use technologies like the Integrator in its Document Center Systems line of networked multifunction machines that copy, scan, store, print, and fax all from the same box. Such a strategy marks almost a reversal of normal product development. Typically, notes Suchman, developers start with a general technology and customize it for individual customers. In the Caltrans case, researchers began with a detailed study of a particular job, developed specific tools for helping engineers, and then worked backward to create a powerful general technology applicable to many domains.