Invited Paper, Joint White House PARC Conference on Leveraging Cyberspace, October, 1996
Mark S. Ackerman
Computers, Organizations, Policy and Society (CORPS)
Information and Computer Science
University of California, Irvine
ackerman@uci.edu
http://www.ics.uci.edu/CORPS/ackerman.html
Expertise networks are technical augmentations of the ad-hoc social networks that enable people to answer questions, accomplish tasks, and find information. Technologies to create, enhance, and maintain these networks will be important uses of the Net.
In this paper, I briefly survey two interrelated things. First, the paper notes the social aspects of the Net that suggest these augmentations would be useful and interesting to do. Second, it discusses some technical directions to augment expertise networks that may change the ways we work and socialize. Because the technical and social are so interwoven in both the design and use of Net applications, at the end of this paper, I also note how important, although difficult, it is to understand the interplay of the technology and the social change.
One more point of introduction and a caution: I hardly have to mention the rapidity and scale of change that has already occurred through the Internet and the Web. Such change is exhilarating. However, the technological directions, rooted in their socio-historical context, are simultaneously causes and effects of many other simultaneous societal changes. It is important to understand as much as possible about the potential results -- and to be excited about them. Nonetheless, the results will be complex and almost certainly have both positive and negative aspects. Therefore, this paper discusses both the limitations as well as the potential in augmenting expertise networks.
Geertz, the noted anthropologist, noted 30 years ago that more and more intellectual activity was crossing traditional boundaries. The firm intellectual boundaries of the immediate post-War period have continued to melt. The Net and accompanying technologies have continued this confluence of many societal forces and interests. Many projects, including those of my research group, combine information and communication, recently thought to be largely separate concerns; community and information seeking, once thought to be largely separate concerns; and, work, entertainment, and education, once thought to be necessarily separate.
The Net, therefore, creates a milieu in which new technical and social possibilities can arise. Software can construct more flexible expertise networks than society could sustain previously. This new flexibility can change the way we produce, disseminate, and store society's knowledge. Within these networks, moreover, we have the flexibility to include information databases, documents, agents, and people together as resources. And as interesting, the same software allows emergent networks to be used for political action, hobbies, a sense of community, and other forms of social life that we can barely imagine. The standard forms, and even the vocabulary we use to describe the forms, will blur even more.
In summary, the once-distinct societal boundaries and definitions, as well as many others, are being blurred. The small picture is new product categories. More importantly, the larger possibilities include:
Our research group is actively exploring these possibilities. We have many studies going on, but three systems demonstrate some of the possibilities for expertise networks, as well as the power of this approach. The first is called the Do-I-Care agent (DICA), the second is Answer Garden 2, and the third is called the Cafe ConstructionKit.
Figure 1: Cascaded DICA agents
DICA is a semi-autonomous agent that searches for new information from known sources on the Web. Users train it over time to do a better and better job of identifying information that is of use. In this, it is hardly unique, although it has some interesting twists. The point for this paper, however, is that its results can be shared among users for relatively little or no cost. It is therefore possible for experts to give or sell their expertise at finding and understanding information to other people (Figure 1). In this model of interaction, each expert does her activity for herself, since she trains her agents for herself in any case. Others are able to use her efforts easily, since users can make their efforts public with DICA. Since one of the standard findings from the Computer-Supported Cooperative Work (CSCW) field is that people will not share information that may make them lose social face or prove otherwise socially problematic, DICA provides for a number of ways of sharing expertise including sharing the URLs, the model itself (without weights), or the interesting changes; we are also currently exploring how to allow users to be anonymous. As with all of our efforts, DICA's sharing also allows for a range of expertise, rather than forcing a dichotomy between experts and non-experts.
More and more of this type of expertise sharing will most likely exist on the Net over time. As we start to disassemble expertise and sell it, we will also need ways to assemble it in coherent ways for the user. DICA explores this problem area. More information about DICA can be found in [Starr, Ackerman, and Pazzani 1996].
Figure 2: Answer Garden functionality in the Web
The second system is called Answer Garden 2 (AG2). AG2 is designed for an organization or a community of practice. It consists of two parts. Answer Garden 2's front-end is an information-base of commonly-asked questions and their answers (Figure 2). Users browse the database to answer their question. If, however, the answer is not present or the user does not understand, he can invoke the back-end. Answer Garden 2's back-end includes a number of distributed components, but those for getting an answer are of especial interest here. If the user cannot find the answer, the system will route his question to help him get an answer. One possibility is to route the question to an appropriate human expert. The human expert can insert new question and answer into the information-base, thus growing it over time. In AG2, another possibility is to route the question to a synchronous chat system composed of people in the user's work group, a net news group, an intelligent agent, or even a consultant. The question can be slowly escalated among these various methods of obtaining help and expertise; the escalation is under the direction of an escalation agent or the user, depending on how the system is configured. A sample scenario of use is shown in Figure 3.
In AG2, the user has access to a non-brittle information system, one where access to the expertise network is held equivalent to access to computerized materials. As mentioned, AG2 mixes documents, active documents, agents, and people within the same memory and help space. More information about Answer Garden 2 and its predecessor system can be found in [Ackerman and McDonald 1996] and [Ackerman 1994].
(a) The user's first attempt to get an answer goes to a chat channel.
(b) The user's jth attempt to get an answer gets escalated to a help desk.
Figure 3: Two possible escalations for a question
The third system is the underlying component system for Answer Garden 2. It is called the Cafe ConstructionKit (CafeCK, named before Java came along), and is a component toolkit and application environment for information retrieval and computer-mediated communication applications. CafeCK provides a set of reusable objects that include message transport for asynchronous and synchronous communication (including a Zephyr-like system, NetNews, and email), parsing for a variety of semi-structured protocols, private and public channels for narrowcast communication, message filters, and message retrieval by a variety of semi-structured methods. By selecting from the set of available components (or by extending it) and by writing a simple Tcl program, an application writer can create a set of distributed processes to handle information retrieval, information access, or electronic communications. Figure 4 shows the CafeCK architecture for a synchronous chat application called Cafe/Espresso.
Figure 4: The Cafe/Espresso synchronous chat application - components and architecture
Of interest here is that CafeCK makes no distinctions among information access, distance learning and "mere" socializing. At the technical level, these activities, while needing different types of user support, are fundamentally equivalent. In fact, one of the research goals of CafeCK is to experiment with applications of mixed heritage that play with our familiar boundaries of work, information, entertainment, education, and community. More information about CafeCK can be found in [Ackerman and Starr 1996].
These are some of the potential enabling technologies for the future. They are not only enabling technologies; however, they are also responses to some unique conditions of the Net, both technical and social. Indeed, the Net is a particularly interesting setting for both social research and engineering development because the technical possibilities and constraints are so interwoven with the social possibilities and constraints. In a critical sense, we are doing research on both the social world and the technical systems simultaneously. We are inventing both the upcoming social world and the technical systems that will enable and support that new social world.
There are many positive aspects to expertise networks, but there are many troubling aspects as well. Care in steering design is paramount if we wish to avoid the above Yogi Berra malapropism.
Below are just two troubling aspects, again in the context of these research projects:
However, these systems do tend to bracket off the sociable from task-related information seeking. A considerable amount of knowledge transfer occurs in informal social interaction, and losing the flexibility for users to fluidly move between sociable and task-related interactions makes the systems (as well as the users' lives) more brittle and restricted. More work is required to know whether these systems can achieve their limited purposes or whether the social (and sociable) world needs to be more intrinsically incorporated.
One's ability to technically foresee problematic design issues is limited [Petroski 1992]. Moreover, systems are socio-historically placed, and any designer is limited in his or her ability to counter general societal forces (as in Cowan's [1983] description of household technologies in the context of other historical changes). Still, cyberspace technologies push in novel directions, as the Web has done, and designers should consider the interplay of social and technical as much as possible.
Expertise networks -- the technical augmentation of ad-hoc social networks by which people seek information, answer questions, and accomplish tasks -- offer an intriguing prospect for cyberlife. We have new possibilities for creating and maintaining very flexible networks of people. Moreover, combining documents, active documents, and agents with people can provide new knowledge capabilities as well as novel forms of social life.
Nonetheless, many challenges remain. Our ability to design technically far exceeds our understanding of what socially needs to be incorporated into a design. Accordingly, future work on expertise networks must include two streams of research. First, we need more empirical research to determine how these ad-hoc groups form, maintain themselves, and disband; i.e., what makes them temporary yet stable collectivities. Second, we need to continue to design, to build systems, and to challenge our assumptions about how the social world can and will function.
All of the quotations are from Yogi Berra. I want to thank Nick Arnett, whose paper pointed me towards the Yogi Berra page on the Web.
These projects have benefited greatly from many conversations with too many people to acknowledge here. However, two debts must be mentioned. Tom Malone started me on this entire path. And the members of my research group, Brian Starr, Dave McDonald, Jack Muramatsu, and Wayne Lutters, contributed a great deal to this understanding.
The work described here has been funded, in part, by grants from NASA (NRA-93-OSSA-09), Interval Research Corporation, the UCI Committee on Research, and the California Department of Transportation.
Ackerman, Mark S. 1994. Augmenting the Organizational Memory: A Field Study of Answer Garden. Proceedings of the Conference on Computer-Supported Cooperative Work (CSCW'94) : 243-252. Also available under http://www.ics.uci.edu/CORPS/ackerman.html.
Ackerman, Mark S., and David W. McDonald. 1996. Answer Garden 2: Merging Organizational Memory with Collective Help. Proceedings of the ACM Conference on Computer-Supported Cooperative Work (CSCW'96) : forthcoming. Also available under http://www.ics.uci.edu/ CORPS/ackerman.html.
Ackerman, Mark S., and Brian Starr. 1996. Social Activity Indicators for Groupware. IEEE Computer, 29 (6) : 37-44. Also available under http://www.ics.uci.edu/CORPS/ackerman.html.
Allen, Thomas. 1977. Managing the Flow of Technology. Cambridge, MA: MIT Press.
Cowan, Ruth Schwartz. 1983. More work for mother: the ironies of household technology from the open hearth to the mircowave. New York: Basic Books.
Kuokka, Daniel, and Larry Harada. 1995. Supporting Information Retrieval via Matchmaking. Proceedings of the AAAI Spring Symposium Series on Information Gathering from Heterogeneous, Distributed Environments : 111-115.
Petroski, Henry. 1992. The evolution of useful things. New York: Knopf.
Starr, Brian, Mark S. Ackerman, and Michael Pazzani. 1996. Do-I-Care: A Collaborative Web Agent. Proceedings of the ACM Conference on Human Factors in Computing Systems (CHI'96) : short papers, 268-275. Also available under http://www.ics.uci.edu/CORPS/ackerman.html.
