Collaboration through virtual organizations is crucial for analyzing and tackling problems such as preventing terrorists, understanding business risks, tracing outbreaks of disease, and conducting criminal investigations. Nonetheless, these collaborations can fail to ?connect the dots? for lack of a technological and organizational framework for leveraging large amounts of dynamic shared information collaboratively. A key research question is how to support the ability of collaborating organizations to attend to and follow up on scattered and often unexpected information from different sources. This research will gain a fundamental understanding of inter-organizational collaborative investigative analysis. The investigators are conducting interviews with members of virtual organizations and controlled laboratory studies to examine interventions that might increase information use across organizational boundaries, and developing and evaluating tools to accomplish these interventions. The research contributes in three areas: (a) understanding how analytical processes across organizational boundaries are evidenced in how people make use of shared but unexpected information, (b) understanding the benefits and costs of different methods for managing collaborative analysis in an environment of large amounts of dynamic information, and (c) designing and testing the effectiveness of interfaces that provide information to support emergent attention and responsibility taking in a multi-organizational collaboration.
This project will inform collaborative investigative analysis in fields such as criminal justice, intelligence, science, and epidemiology. The results could lead to new visualization tools for analysts in these areas, recommendations for organizational practices to improve the quality of collaborative analysis, new methods for training professional analysts to solve complex, interconnected problems, and learning tools for graduate programs in fields such as intelligence and epidemiological analysis and criminal justice. The research will also involve undergraduate and graduate students, and will result in their further training and education in interdisciplinary research.
Collaborative analysis of information plays an import role in detective work, epidemiology, and national intelligence. The goals of this project were to improve collaborative analysis by identifying the kinds of problems that arise when collaboration occurs across multiple organizations and designing and evaluating new computer tools that could make it easier for teams to work together to analyze data. To achieve these goals, we conducted laboratory studies that looked in detail at how people shared, or failed to share, information during a collaborative crime analysis task. The participants in these studies sifted through evidence of crimes in a database that mirrored real data that detectives work with. The entire task was modeled on the way detectives work. In all of these studies, the detectives were separated in time and/or space, including at different labs. The data they worked with contained evidence of a serial killer who had committed a series of murders but who had not been caught due to the complexity of the evidence. This problem is very much like a real situation faced in the U.S. We discovered that detectives who belonged to different organizations had a difficult time in the process called sensemaking, that is, organizing data in a way that made sense to find a serial killer. We saw problems with detectives working together. For example, an individual may fail to consider all the available information, come to an erroneous conclusion, and then share that conclusion with partners. Or, an individual analyst may come to the correct conclusion but for organizational reasons, fail to share it with others who need to know. The biggest problem we identified was that detectives who did sloppy investigations and gave bad advice had too much influence on others. Other detectives took the bad advice and failed to correctly identify the serial killer. We developed some tools aimed at helping collaborative analysis. One tool helped detectives (in the lab) visualize the data and link it to a timeline. Another set of tools aimed to improve collaborative processes by encouraging detectives to carefully review advice instead of pre-judging it as helpful. Some of the tools helped but the most helpful (in preventing bad advice from affecting others) was to be sure to select at least one very effective team member. That is, a bad advisor was generally ignored as long as there was at least one articulate, correct detective on the team. Broader Impacts: The project provided training in STEM disciplines to eight undergraduate students, masters students in computer science, two doctoral students, and a faculty member at a local school of management. It also contributed new findings and methods to information science, computer science and management science. The data, materials, and computer code from the research have been made freely available to other researchers. We hope that through our dissemination efforts the results will help improve collaborative analysis in areas of societal concern, such as crime and epidemiology.
