by Victor Kaptelinin (Editor), Mary Czerwinski (Editor)
The desktop metaphor has been around for quite a while, ever since Xerox PARC came up with it for the Star workstation in 1981. Apple borrowed the concept and made it famous and successful with the Macintosh in 1984; since then, it has become a standard interface feature in most graphical user interfaces.
Of course, it's a matter of some debate whether the collection of interface features we call the "desktop metaphor" actually works as a metaphor anymore. I've argued in print that these elements are no longer perceived as metaphorical by many or most users, and Johndan Johnson-Eilola has similarly examined interface elements as knowledge spaces rather than metaphors. But let's bracket that discussion, since the authors in this collection are less concerned about whether the desktop metaphor is still operant, and more concerned about how this collection of interface elements limits the sorts of things we do with computers. As the editors argue in their introduction, the collection is "a collective exploration of the design space of new-generation digital work environments" (p.11).
The editors summarize the authors' contributions as identifying and addressing the following limitations of the desktop metaphor:
- Information access vs. information display
- Multiple information hierarchies
- Communication and collaboration
- Coordinated use of multiple technologies (pp.9-10).
These are certainly limitations of the desktop metaphor, which was configured largely for standalone machines that addressed neophytes working in office environments and using machines with very limited processing and graphics capabilities compared with today's machines. To address these limitations, the authors come up with very different answers, and some of them argue with each other quite directly.
The first contributed chapter is Freeman and Gelertner's "Beyond Lifestreams: The Inevitable Demise of the Desktop Metaphor," in which the authors revisit their 1994 Lifestreams project. Lifestreams threw out unique file names and the spatial placement of interface elements, instead supplying "a time-ordered stream of documents combined with a few powerful operators for locating, organizing, summarizing, and monitoring information" (p.19). One of the most important, and hardest-to-implement, features was summarizing: the automatic abbreviation of documents or document collections (p.23). Freeman and Gelertner argue that in addition to these features, a Lifestreams unified model should address these issues:
- Storage should be transparent
- Folders and directories are inadequate organizing techniques
- Archiving should be automatic
- Computers should make "reminding" an integral part of the desktop experience
- Personal data should be available from anywhere, to any device, and compatibility should be automatic
- The system should provide a means of summarizing a set of documents into a concise overview (pp.24-26)
My first thought was that Google's products, particularly GMail and GDocs, seem to be reading from this playbook. But we also see some of these characteristics in the Nintendo Wii (which summarizes gameplay) and Facebook.
In fact, as I read the other contributions, what really struck me was that none of these contributors seemed to be addressing Facebook, MySpace, GDocs, Gmail, Bascamp, and other relatively recent web-based innovations that have chucked the desktop metaphor for dashboards and tagging. When David Karger talks about Haystack, which provides an "inbox" for email and contacts, he seems oblivious to the Web 2.0 revolution that has been in full swing for quite a while now:
Should users then abandon applications and move all their data to the web? Hypothetically, a user could create a separate web page for each email message, each directory, each music file, each calendar appointment, each individual in their address book, and so on. Editing these pages, the user could indicate arbitrary relationships between their information objects. Feeding these web pages to a tool like Google would give users powerful search capabilities ...
Of course, such an approach could never work: it requires far too much effort on the part of the user. (p.62)
Just a reminder: this collection was published in 2007. User-generated web content was well established by 2001, and even web tyros were already exploiting its potential for organizing diverse data types by then.
Indeed, many of the contributions seem a bit quaint here in January 2008, since they fail to anticipate the enormous changes that came in the last couple of years due to distributed interacting services and the boost that comes from making one's resources social. (It's unclear how much time passed between the time the chapters were written and the time they were published.) For instance, Plaisant et al. argue in "Personal Role Management" for a single, omnibus system that stores email, calendar, documents, and other information, integrated with the university system; this tightly controlled set of information can be separated into roles (e.g., school, work, general). But this centralized, "walled garden" approach has generally fared poorly against more distributed open systems that allow users to choose their own web apps and organize connections between them as they see fit.
But things get interesting again as we move to the second half of the book, in which sociocultural approaches such as activity theory are brought to bear. In "Solyent and ContactMap," Fisher and Nardi talk about these two projects as ways to support knowledge work. "Little of what knowledge workers do is done alone," they argue (p.171), so they concentrate on mapping connections in individuals' social networks by mining emails. This approach is admittedly less rich than a context-aware approach might be, they say, but it is still powerful -- and they are able to automate it (p.188).
Voida, Mynatt, and MacIntyre go in the same theoretical direction with "Supporting activity in desktop and ubiquitous computing," in which they argue that activities are multifaceted, dynamic, collaborative, multileveled, and distributed across spaces (pp.196-197). They describe Kimura, an activity-centered work environment that supports knowledge workers by separating their digital workspace into a focal region (a monitor) and two peripheral displays (projected on the office walls). Work activities are modeled as document clusters named "working contexts" (pp.199-200), displayed as "a montage of images garnered from system activity logs" (p.200). Toward the end of the article, the authors argue that such an environment can allow users and researchers to conceptualize levels of activity systems (e.g., activity system of the industry, the organization, and specific groups within that organization) (p.216).
Jakob Bardram takes a somewhat different tack, describing "activity-based computing" or ABC, in which "the basic computational unit is not the file ... or the application ... but the activity of the user" (p.224). He contrasts this approach with the traditional design ideal of supporting "tools and materials," a design ideal that meshes well with the assumptions of traditional activity theory; that traditional approach includes a deep skepticism toward workflow, which has been seen as implying human activity as a mental construct rather than arising from material conditions in dialectic with human objectives (p.226). Bardram carefully delineates that approach from ABC, which does not support workflow but a system in which "the human activity defines the computational activity" (p.227). ABC supports the cooperative nature of human activity via computer-based collaborative artifacts (p.228). Bardram supports this point with an empirical case study of ABC in a hospital environment.
We get back in the weeds again with Ravasio and Tscherter's "Users' theories of the desktop metaphor, or why we should seek metaphor-free interfaces." Here, the authors argue that problems stem from "an overly concrete metaphor that no longer applies" (p.265).
Kaptelinin and Boardman are up next in "Toward integrated work environments," which revisits the most successful CSCW application (email), explores why it is so successful, and chronicles Katelinin's work with the UMEA system. The authors point out that email is not a tool but a habitat (p.298) and that users use email for dedicated task management rather than turning to task management software (p.300). They argue that cross-application coordination is needed (but they don't mention Google) (p.304) and describe how UMEA creates project contexts as by-products of work on projects (p.314).
Finally, Moran and Zhai argue in "Beyond the desktop metaphor in seven dimensions" what I have been thinking all along: That a strict concept of the desktop metaphor is a straw man (p.335). We are on our way to a new metaphor, they say (wrongly, I think) (p.335). This is in some ways the only chapter that sounds as if it has been written after 2002 -- Moran and Zhai recognize the existence and repercussions of cloud computing, predict the movement from applications to services, discuss the interaction between collaboration and tagging, emphasize that personal information is becoming public, and talk about how activities are now providing records of experience (p.349). There's not much in this chapter that you won't get from reading TechCrunch, but it's stated in academic terms and interconnected appropriately with research. That is, it's a strong finish to a book that too often seems stuck in 2002.