Pitfalls of Information Access with Visualizations in Remote Collaborative Analysis

In a world of widespread information access, information can overwhelm collaborators, even with visualizations to help. We extend prior work to study the effect of shared information on collaboration. We analyzed the success and discussion process of remote pairs trying to identify a serial killer in multiple crime cases. Each partner had half of the evidence, or each partner had all the available evidence. Pairs also used one of three tools: spreadsheet only (control condition), unshared visualizations, or shared visualization. Visualizations improved analysis over the control condition but this improvement depended on how much evidence each partner had. When each partner possessed all the evidence with visualizations, discussion flagged and pairs showed evidence of more confirmation bias. They discussed fewer hypotheses and persisted on the wrong hypothesis. We discuss the possible reasons for this phenomenon and implications for design of remote collaboration systems to incorporate awareness of intermediate processes important to collaborative success.