This project will bring together a multi-disciplinary research team from the University of Washington, Boeing, Microsoft and from the Center for Human-Machine Interaction in Denmark to explore the concept and practice of collaborative information retrieval. The research will address two central effects of the dramatic spread of intranets and Internet-based information and tools: the tremendous amount of information that is available to individuals, groups, and organizations, and the increased potential for work to be carried out in distributed groups.

Most information retrieval research and development focuses on individuals, an individual querying a database or archive, an individual delegating retrieval to an agent in a multi-agent system, and so forth. Most of these individuals are of course acting in group, organizational, and societal contexts, and retrieval is often undertaken to address collective concerns. Our research begins with the premise that a better understanding of the social context of information retrieval will contribute to better design and use of computational tools to support it.

We understand collaborative information retrieval (CIR) to be any activity that collectively resolves an information problem taken by members of a work-team, regardless of the nature of the actual retrieval of information. Information retrieval is construed in the broadest sense and includes processes such as problem identification, analysis of information need, query formulation, retrieval interactions, evaluation and presentation of results, and applying results to resolve an information problem. Members of a work-team are those who were brought together to solve a problem which is the goal of the team. The processes of CIR may lead a member or members of the work-team through various IR behaviors or paths. For example:

• Two or more team members may follow a path together to find information for a specific problem.
• Team members may use different paths to find information for a specific problem, either in parallel or sequentially.
• Team members may be guided by suggestions from other team members who had found the same or similar information earlier, and so on.

The project is directed toward collaborative teams, that do not have information retrieval as their primary task. The information retrieval is serving the team in their joint effort to perform a work function, such as product design. The information retrieval needs and the roles of the team members in work performance and retrieval tasks will be structured by the characteristics of their joint task and the professions and competence of the individual team members.

An analysis of the collaborative retrieval strategies during a work session therefore have to be related to the particular task - and competence- characteristics in order to be able to generalize and to distinguish between the requirements of different work scenarios. In short, the project will apply a work-centered methodological perspective.

The aims of this study are as follows:

• to analyze the structure of collaboration in selected information intensive, actual work scenarios and to describe the context that will shape collaborative information retrieval patterns;
• to explore the nature and occurrence of collaborative information retrieval in work settings and to propose technological innovation and organizational change that could support, facilitate, or improve it;
• to construct a work-centered conceptual framework that includes collaborative information retrieval.

These aims will be achieved by applying the framework to teamwork situations, and by addressing the following research questions:

• How and when do work teams establish a CIR path? How much is made explicit?
• How are the roles of team members in information retrieval shaped by the work content?
• What factors determine a work-team’s choice of a CIR path? How do collocated and distributed teams differ? Homogeneous and heterogeneous teams? New teams and teams with common history? How does the organizational context or culture affect practice?
• What formal and informal processes do teams use to coordinate their IR activities?
• What difficulties do work-team members encounter in retrieving information collaboratively?
• What is the role of CIR in a team’s work?
• What technologies do work-team members use for CIR?

• What factors motivate work-teams to use CIR?
• What factors dissuade work-teams from retrieving information interactively?
• Can we identify situations in which CIR would be effective and others in which it would not be desirable?
• What technological innovations can improve, support, and facilitate CIR?
• What organizational changes can improve, support, and facilitate work-group?

This is an ambitious research agenda, requiring several areas of expertise. With that in mind, we have assembled a multi-disciplinary team each of who finds this within their central focus of research. Raya Fidel and Harry Bruce of the University of Washington have prominent records in information retrieval and qualitative and quantitative studies of distributed technology use. Annelise Mark Pejtersen from the Center of Human-Machine Interaction in Denmark provides both conceptual strength and many years of experience in qualitative and quantitative study of information retrieval in work situations. Sue Dumais of Microsoft Research has also been a leader in IR research. Jonathan Grudin of Microsoft Research and Steven Poltrock of Boeing have worked together and independently on collaboration and technology, recently co-chairing the ACM Conference on Computer Supported Cooperative Work.

The industry partners in this project are in research divisions and are committed to working and publishing in this area. One project will not answer all of the questions posed above, but the project described below represents a strong step, and this team of investigators is ideally constituted to make substantial progress.

Most of the existing work on technology support for information retrieval has focused on individual behavior. This helps us understand information gathering by groups up to a point, but group work is generally more complex than the sum of the behavior of the individuals involved. Specific needs arise in group interaction. Information that is acquired is shaped and communicated within the organization and these factors can all affect subsequent information seeking. These factors and behaviors are still not well understood.

At the same time, there is a growing focus on collaborative teamwork as an essential skill in today’s business environment. Universities and schools are responding to requests to provide experience and training in performing work collaboratively that is quite at odds with traditional approaches. Successful businesses can profit from having employees, suppliers, and partners distributed across regions and nations, and the benefit is of course greater if they work together effectively. Technology can enable these collaborations and provide the means to increase their effectiveness. We need to know how teams collectively obtain and integrate the information they need, when collocated or when distributed, and how to design technology to support this.

Computational support for activity at the group level is particularly important at this time for several reasons i) the breadth of information available over intranets and the Internet is growing with a rapidity that is difficult to comprehend, much less manage; ii) in response, support for individual information retrieval has shifted from being an enhancement to being a necessity, and is the focus of extensive research and development; iii) for many collocated and especially for distributed groups, an increasing percentage of activity is digitally mediated, providing greater opportunity for computational support; iv) one goal of multi-agent systems, finding and negotiating over information on behalf of individuals, will benefit from and perhaps be extended by a better understanding of information retrieval in group contexts. We propose to examine the information retrieval needs of collocated and distributed workgroups in organizational settings in order to understand and contribute to the design of computational support.

There is a real need to understand the social practices that govern collaborative information retrieval. Repeatedly, in recent years, research has indicated that our understanding of group behavior is insufficient to design successfully for group support. Early efforts at “expertise location” within organizations relied on technical systems to identify relevant individuals. Answer Garden, for example (Ackerman, 1994), represented a hybrid sociotechnical system. This system, if unsuccessful in answering a question, directed the query to an expert responsible for the topic area. A subsequent field study of expertise-seeking (McDonald and Ackerman, 1998) showed that behavior is much more complicated. People balance many factors in deciding whom to consult and often ‘the leading expert’ is at most the choice of last resort. Based on this more detailed understanding of expertise identification, the authors are building a new support system. Another example is the result of decades of work on group support that focused on formal meetings (reviewed by Kraemer and King, 1988). This research demonstrated the key role of unplanned, informal interaction in many workplaces (e.g., Kraut et al., 1986) and led to extensive technology development. These developments have, in turn, drawn attention to the complexity and difficulty of supporting informal group processes and demonstrated the need for more detailed study.

Although this has not been studied systematically, we expect to find that teams collaborate in acquiring information in different ways. Many teams need information to perform their work. Product development teams, for example, acquire information about new technologies and market segments. Investment teams acquire and evaluate information about investment opportunities. These are examples of knowledge work, which is often performed by collaborative teams composed of people with diverse backgrounds and areas of expertise. Each person’s expertise includes how and where to find domain information, how to evaluate it with respect to the team’s goals, and how to communicate it to the team members. This process is potentially ongoing and not confined to a single pass or phase of a project.

There are many ways that a team may collaborate in acquiring information. The team may formally plan to seek information, allocating different search tasks to different people. For example, a team preparing a recommendation regarding the launch of a new airplane program may assign people to approach different airlines regarding their potential interest in the airplane capabilities. Other members of the team may evaluate the readiness of new materials that could reduce the weight of the plane, increasing its range and fuel efficiency. Teams often track progress on such information acquisition tasks.

Teams may acquire information in an ad hoc, unplanned way. For example, a team developing a new software product may have extended discussions about the product’s intended behavior and market. Such discussions often raise questions that cannot be immediately answered by anyone on the team. Often people will seek missing information on their own initiative, then bring this information back to the team. This new information may raise new questions that they or others will investigate later.

People may work alone or together at various stages of a group project. Team discussions are a common means of identifying goals and the information needed to achieve the goals, but a single individual may be assigned to prepare an analysis of information requirements. The team may also discuss where information may be found, identifying other people who have critical experience or knowledge. Two or more people may work together to locate and extract information. Two people may conduct a joint interview with an airline executive, for example, with the expectation that two people will understand and remember better than one. Teams often work together to integrate the resulting information into an expanding knowledge of their shared problem domain and its solutions.

These various manifestations of information behavior in teams could be described as collaborative information retrieval. To date, the social and organizational processes that characterize collaborative information retrieval have not been studied systematically. By better understanding how people collaborate in a range of information retrieval and management activities, we can begin to articulate design specifications for more effective technologies and multi-agent systems to support and augment collaborative information management activities.

Most of today’s information retrieval tools are designed to serve individual searchers rather than people working in groups. OPACs and Web browsers allow individual users to type in queries and view results, but it is difficult for individuals to share search strategies, search queries, or the analysis of results with others except by looking over someone’s shoulder. In contrast, informal observations suggest that sharing of information retrieval strategies and results takes place in a wide variety of work settings. In physical libraries individual patrons seek help from librarians, librarians interact informally with each other to share tips and keep up to date about new resources, and patrons interact with each other (Twidale and Nichols, 1998; Proctor et al., 1998). Nardi and Miller (1991) have described similar spontaneous and casual collaborations among co-workers to develop, debug, and use spreadsheets. And, as described above, corporate teams also engage in a wide variety of collaborative efforts to acquire and exchange information.

Previous research on understanding and supporting the interaction between end-users and trained search intermediaries in the evolving networked environment comes close to our notions of collaborative information retrieval (Fowell and Levy, 1995; Billings, 1994). The focus of this work, however, is much narrower in scope than our proposal, focusing on a single relationship (patron-librarian) in one setting (library) over short periods of time.

Collaborative filtering is a popular approach to allowing users to share information with each other. In collaborative filtering, the behaviors of similar individuals are used to recommend items that a current user may like (CACM, special issue on recommender systems, 1997). More generally, new users can benefit from the experiences of others by using previous interaction histories to enrich digital objects (Hill and Hollan, 1994). Hill et al. (1992) explored the notion of edit ware and read ware. Their system graphically depicted the history of author or reader interactions with electronic documents, thus guiding new users through a large information space using well-worn paths. Wittenburg et al. (1998) developed a system to support asynchronous group browsing of Web favorites. Collaborative filtering systems have been successful in narrow domains where subjective preferences are important and where a small number of objects are used by a large number of users (e.g., movies, books, or restaurants), but less successful in the large heterogeneous environments that are of interest to us.

Several research groups have recently developed systems for capturing and visualizing search histories (e.g., Hightower et al., 1998; Twidale and Nichols; 1998). These systems are typically aimed at helping individual searchers quickly access their own previous searches, although presumably others might benefit from these histories as well.

Finally, research on organizational memory focuses on methods for effectively creating and sharing information within large organizations (e.g., Ackerman, 1998; Berlin et al., 1993; Fischer and Reeves, 1992). While this work shares some goals with our proposal, we address a much wider range of information seeking, creation, and use activities by members of a work team.

In short, there are several gaps in groupwork and information retrieval research that our proposal aims to address. We need a detailed understanding of the interaction dynamics of group information behaviors. We need to understand these activities and the role that they play in work and organizations so that we can recommend ways to support, facilitate, and enhance the effectiveness of those team-based, workplace activities which we are calling collaborative information retrieval (CIR). Once we understand CIR we can articulate design guidelines for new technologies to support it.