Background

Users are being embraced as partners in developing computer services in many current computer supported cooperative work systems. So called Web 2.0 applications and collaborative authoring reflect this tendency. For example, wikis promote user collaboration, information sharing, and communication by giving users permission to add, edit, and remove contents [11]. Such tools mean more collaborations are moving from being between people who know one another to potentially anonymous collaborators with little shared understanding. Therefore, for successful collaboration, individual participants must identify and understand others’ contributions.

A difficulty in such remote collaborations is that insufficient understanding between contributors reduces the value and efficiency of collaborative work. For example, individuals may make contributions duplicating work already done by others or may unknowingly undo or delete others’ work while instantiating his/her work. Replicated effort and destructive editing are appropriate in many situations, but a lack of understanding between remote collaborators increases the likelihood of accidental and unnecessary occurrences.

For effective collaboration, users need to understand the efforts and motivations of others. One approach is through records of user history. Following the steps of other collaborators’ works can help the users increase their understanding of collaboration by resolving the ambiguities caused from (1) the diverse backgrounds of users, (2) the complexity of their work, and (3) interaction barriers [7].

However, problems with the use of history in collaborative settings include difficulty in locating activity of interest in large tasks, the problem of history records being at a system activity level rather than a human activity level, and difficulty in supporting navigation and comprehension in the branching histories used to represent alternative directions.

These problems are generally independent of the domain of the collaboration and motivate the research on representations and interfaces for user history. Numerous studies have been conducted on pertinent visualization techniques to exhibit the history information, because the techniques play an important role in human cognition [3] (e.g. history visualization [2, 5, 12], awareness [1, 17], and visualization for navigation/manipulation [6, 13, 16], and branching history [4, 8, 10, 15]).

Overview

This research examines support for providing user location and comprehension via an application-independent mechanism for automatically interpreting and visualizing user history.

Firstly, in order users to locate and comprehend periods of activity relevant to their work, the mechanism provides diverse visualization and navigation interfaces of collaboration history. By displaying lists of times or lists of work states, users get a sense for the flow of activity from state to state. Branching history, where users can try alternative solutions by branching out from the collaboration history, can be represented via tree structure to indicate the relation between different branches in histories. To alleviate user disorientation in large history, overview + detail visualizations [3] are employed. By presenting both specific points in the history and highly abstract summary of the entire history, those visualizations prevent users from losing context during history navigation. Additionally, the users can define the scope of visualization and navigation via history filters, as well as replay history via playback interface.

Figure 1. Visualizations of collaboration history
(Click here for a larger view of the image)

Secondly, to support comprehension of the history, the mechanism employs automatic history interpretation to group low-level system records into a more human-level representation of activity. The interpretation mechanism performs Hierarchical Agglomerative Clustering (HAC) using a dissimilarity function to decide grouping of clusters. Initially, HAC regards each history record as a single cluster, and clusters are supposed to be grouped gradually. Likewise, the two clusters with the smallest dissimilarity are grouped, and this process continues until all events are in one single cluster. As a default function, HAC uses the time interval between consecutive events. However, a system developer can redefine the dissimilarity function reflecting the characteristics of a target system to obtain better grouping result. This result is used to update visualization interfaces to provide better representation of collaboration history.

Figure 2. Automatic Grouping of History Records
(Click here for a larger view of the image)

The history mechanism in this research is under development as a JAVA library, and integrated into two software applications for evaluation. The applications are the next release of the VKB (Visual Knowledge Builder) [14] and the web-based DE (Design Exploration) [9]. Although these are currently under development, their main frameworks are complete enough to integrate the mechanism. After the integration, the mechanism needs to be evaluated regarding their ability to improve human location and comprehension of specific sub-tasks in a history. User studies are also necessary to evaluate the role of the provided history information in human problem-solving processes.

This research evaluates the potential for user history to help in a wide range of collaborative applications. Unfortunately, each application has to implement its own history mechanism, limiting the features actually available. The history mechanism presented here provides a variety of visualization and navigation interfaces for locating and comprehending activity, and automatically groups system-level events into higher-level activities. The mechanism has the potential to be integrated into a variety of applications with little adjustment for compatibility.

References