People can help other people find information in networked information seeking environments. Recently, many such systems and algorithms have proliferated in industry and in academia. Unfortunately, it is difficult to compare the systems in meaningful ways because they often define collaboration in different ways. In this paper, we propose a model of possible kinds of collaboration, and illustrate it with examples from literature. The model contains four dimensions: intent, depth, concurrency and location. This model can be used to classify existing systems and to suggest possible opportunities for design in this space.
Although commonly conceived as a solitary activity, information seeking often involves collaboration with others [2][5] [9]. Collaborative filtering or recommendation systems (e.g., [4]) are common examples of online collaborative search: the behavior of people -paths taken, documents seen, etc. -while looking for information is used to inform the behavior of others searching for similar information later. The people involved do not need to be aware of each other; the system generates appropriate information as dictated by the users' actions. Other approaches (e.g., [1] [10]) assume co-located users who communicate with each other normally, construct queries, and share search results produced by the system.
In this paper, we propose a taxonomy of collaboration in support of information seeking designed to distinguish the various forms of online collaboration. The taxonomy consists of four dimensions (intent, depth, concurrency and location) that can be used to characterize various aspects of collaboration.
Our model is related to that proposed by Hansen and Järvelin [3]: we share the synchronous/asynchronous dimension and we are concerned with computer-mediated communication (although we do not rule out additional human-to-human communication channels). We differ somewhat in our treatment of coupling: CSCW literature uses the terms “loosely coupled” and “tightly coupled” to refer to a variety of phenomena, including organizational structure, software architecture, and degree of collaboration, among others. We focus on technical (rather than on social) issues in this model, and represent some notions of coupling in the “depth of mediation” and “explicit vs. implicit” dimensions.
In the rest of the paper, we first describe the dimensions on the model, and then illustrate them with examples from commercial systems and academic research.
Recommender systems use behavior of a group of people who have engaged with particular content to suggest choices to others searching for similar information. The goal here is to use information previously found by others to inform new search results. This is implicit collaboration: while people may be generally aware that their results are based in part on data obtained from other users, they may not know who those people are or what purpose they had in mind while searching. Thus the collaboration here exists because the search engine uses historical data as a source of evidence for document relevance to a query. In some sense this is not strictly collaboration, but rather a coordination of people’s activities.
We can contrast this with explicit collaboration, in which a small group of people searches for documents to meet a shared information need. The need may evolve over time, but through-out a search session that need is shared by all team members and it motivates their search activities. This is related to Morris and Teevan’s notion of ’task based" collaboration [7].
Mediation of information seeking can occur in the user interface [1][6] [10], and may also be reflected in the underlying search algorithms. We distinguish UI-only mediation from deeper algorithmic mediation that explicitly represents contributions from different people in the algorithms that retrieve information. Examples of this mediation include recommender systems (that use records of individuals’ selections to rank documents for retrieval) and FXPAL’s Cerchiamo [8] that combine relevance feedback from multiple people to rank documents and offer search term suggestions based on team members’ actions.
This dimension reflects the flow of influence among members of a collaborating group. If search activity by more than one person occurs at the same time, it is possible for influence (see the Depth of mediation dimension below) to flow between members during a search session. The asynchronous case describes the condition in which people do not work at the same time; those who search later can benefit from the work of earlier collaborators, but the earlier ones did not benefit from contributions of subsequent collaborators. This does not mean that team members engaged in synchronous collaboration need to operate in lock-step, searching or browsing results simultaneously. Rather it means that they are actively involved in various aspects of information seeking activity at the same time. They may divide their activities in any manner supported by the tools they use; the key is the possibility that each team member’s actions can influence other team members.
Finally, collaboration may be co-located or distributed. Distributed collaboration implies the need for additional channels to coordinate searchers’ activities. Such channels may include chat, voice, or audio conferencing.
In the previous section we introduced the four dimensions of our taxonomy. We now introduce a prioritization or ordering of what we believe to be the most important dimensions for distinguishing between existing and future online,
This content is AI-processed based on open access ArXiv data.