Search   |   Back Issues   |   Author Index   |   Title Index   |   Contents

Conference Report


D-Lib Magazine
July/August 2008

Volume 14 Number 7/8

ISSN 1082-9873

Report on the 1st Collaborative Information Retrieval Workshop

Held in Conjunction with the Joint Conference on Digital Libraries (JCDL) 2008


Jeremy Pickens

Gene Golovchinsky

Meredith Ringel Morris
Microsoft Research

Red Line


The 1st Workshop on Collaborative Information Retrieval was held on June 20, 2008, in conjunction with the Joint Conference on Digital Libraries (JCDL) 2008. The principal organizers of the workshop were: Dr. Jeremy Pickens (FXPAL), Dr. Gene Golovchinsky (FXPAL) and Dr. Meredith Ringel Morris (Microsoft Research).

Today's digital search technologies are designed for a single user working alone, even though prior studies of students and information workers have demonstrated that search is a collaborative process. Nevertheless, such collaborations are difficult to achieve with existing digital tools, often resulting in high overhead such as undesired redundancy of effort. Recently, researchers in the Human Computer Interaction (HCI) and Information Retrieval (IR) communities have begun to develop new tools, algorithms, and systems to facilitate collaborative search. This workshop was particularly timely, considering that the two most common uses of Internet technologies are communication and search; collaborative information retrieval systems exist at the intersection of these two trends. The event brought together researchers from the HCI, IR, and Library Science communities, in both industry and academia, to define and discuss this emerging area.

Due to the relatively new nature of the topic, many sessions throughout the day were infused with discussions of terminology and taxonomy. In social, networked environments, the term "collaboration" has typically been used to refer to "wisdom of the crowd" approaches, such as collaborative filtering and social recommendation technologies. The consensus among workshop participants was that the primary factor that distinguishes these collective intelligence methods from the aims of this workshop is the intent with which one user engages another. In Collaborative Information Retrieval systems, people in a team work together on an explicitly shared information need, whereas in collaborative filtering, individual users receive recommendations based on the correlation of their current behavior with the mass aggregation of historical crowd behavior.

Explicit (active or intentional) collaborative information retrieval has some interesting characteristics that distinguish it from implicit (passive or non-intentional) collaboration systems. There is much more emphasis on interaction, as the system has not only to communicate search results to users, but also must mediate some forms of communication and data sharing among its users in near real-time. New algorithms need to be invented that use inputs from multiple people to produce search results, and new evaluation metrics need to be invented that reflect the collaborative and interactive nature of tasks.

Jeremy Pickens and Gene Golovchinsky kicked off the discussion with a brief presentation of their position paper "A Taxonomy of Collaboration in Online Information Seeking," which set the tone for the workshop by giving an overview of the types of issues currently being explored in this field. The remainder of the workshop was divided into three main sessions: (1) user scenarios, (2) models of collaborative information seeking, and (3) systems and evaluation.

The first session, on user scenarios, began with two presentations of position papers. Madhu Reddy's presentation on "Learning About Potential Users of Collaborative Information Retrieval Systems" reported on his fieldwork studying current collaborative information retrieval practices of healthcare professionals. Brynn Evans' presentation titled "Towards a Model of Understanding Social Search" focused on results of a web-based survey wherein participants described their most recent Web search session, many of which involved social interactions. Following the presentations, workshop attendees discussed other examples of communities that might benefit from collaborative search technologies; suggested examples included students, faculty members, and the military, as well as businesses that conduct research, such as pharmaceutical companies. The group also discussed different roles that collaborators might undertake in different user scenarios – for example, two students researching a topic together in the library may have similar roles, whereas a librarian and student working together to research a topic might require different tools to accommodate their distinct levels of expertise.

In the second session, we focused on models of information seeking behavior. Ed Chi presented the second part of the paper co-written with Brynn Evans; the model mapped the data that Brynn described onto the sense-making model, and identified episodes of collaboration in the overall flow. Chirag Shah presented his model of collaborative information seeking that situates collaboration among other (less explicit) modes of communication (i.e., contribution, coordination, and cooperation) and postulated some necessary pre-requisites for collaboration (diversity of opinion, independence, decentralization, and aggregation). Finally, Max Wilson described his paper (co-authored with m.c. schraefel) "Evaluating Collaborative Search Interfaces with Information Seeking Theory." The paper uses Marcia Bates' Berrypicking formalism to identify opportunities for collaboration that should be supported by the systems that mediate the collaboration.

The third session of the workshop explored existing collaborative retrieval systems. Frank Hopfgartner began by presenting work on collaborative information seeking trails. While traditionally such methods have been developed for asynchronous, implicit collaboration, Hopfgartner made the case that similar techniques are also relevant to explicit scenarios because users are not just sharing the end results of a search, but are storing and sharing the process itself. Colum Foley argued that the two most important factors for the design of explicitly collaborative systems are sharing of knowledge coupled with division of labor. His experiments with collaborative relevance feedback yielded a decrease in result diversity. Complementary relevance feedback may therefore produce better overall group outcomes. In keeping with this diversity theme, Meredith Ringel Morris offered insights into properties of groups that lead to better collaborative performance. Division of labor in a shared task is more effective when it makes use of the differing expertise of each group member. Two major themes of this session were (1) the relative merits of simulation versus full interaction for evaluating collaborative systems, and (2) effectiveness metrics other than precision and recall. Proposed measures included information diversity or uniqueness, discovery rate, total task time, user (team) frustration, engagement and enjoyment, usability, energy or effort expended, opportunity, cost, and user (team) confidence in their coverage and overall satisfaction.

At the end of the day, participants agreed that the area of explicitly collaborative information seeking contains a wealth of interesting and relevant open research problems for multiple communities; information retrieval, seeking and exploration; human-computer interaction; and computer-supported cooperative work (CSCW).

Workshop proceedings may be found at <>. We are looking forward to holding a follow-on workshop next year to learn about the progress that researchers have made in advancing the state of the art of collaborative information retrieval.

Copyright © 2008 Jeremy Pickens, Gene Golovchinsky, and Meredith Ringel Morris

Top | Contents
Search | Author Index | Title Index | Back Issues
Previous Conference Report | In Brief
Home | E-mail the Editor


D-Lib Magazine Access Terms and Conditions