Human-Computer Interaction Institute
School of Computer Science, Carnegie Mellon University
Sensemaking with Shared Visualizations:
Aruna D. Balakrishnan
Increasingly, collaborators are separated geographically and are also
faced with large quantities of information that can complicate collaboration.
Visualizing information can help collaborators sort through large quantities
of data, but visualizations help only when they promote effective
problem-solving behaviors such as division of labor and open communication.
This dissertation explores the impact of network visualizations on
collaborative problem solving by examining three laboratory studies.
Using the "detective mystery" as an experimental paradigm, remote collaborators worked synchronously via instant messenger to identify a serial killer who was hidden within multiple crime reports. In the first study, the evidence was divided between the pair of collaborators. When collaborators were given a network visualization tool that showed them how the evidence was linked, they performed better than those without a visualization. The visualization also fostered discussion between partners.
The second study looked at whether the visualization would help collaborators if they already had full access to all the evidence. Whereas the visualization tool improved performance for collaborators with half the evidence, the same visualization tool did not improve performance when each collaborator had access to all the evidence. Collaborators seemed to be overwhelmed and did not approach the task systematically. Unlike their counterparts, who each had half the evidence, collaborators with all the evidence talked less, discussed fewer hypotheses, and did not divide the labor.
The final study asked whether interpersonal and task-oriented discussion-prompt interventions encourage collaborators to adopt problem-solving strategies necessary for success. Discussion-prompt interventions helped pairs improve their search and analysis process.
This dissertation suggests that visualization tools may prompt collaborations to be more systematic, but this depends on collaborators effectively using the visualization, finding relevant patterns within the visualization, and ultimately using these findings to direct their analysis.