lessons learned from the original experience are painstakingly transformed into products that can most effectively deliver these lessons to their target audience.
Unfortunately, all of this mediation comes at a price, both in terms of time and money. These projects, each of which required millions of U.S. dollars and years of effort by large development teams, cannot be endlessly replicated to process all of the stories told by members of an organization, even wealthy organizations that are only a fraction of the size of the U.S. Army. In accordance with the original intent of their development, these systems are extremely effective at delivering a modest number of lessons-learned for a modest number of training objectives. However, they do not represent a general knowledge management solution for story-based organizational learning, where the time between real-world experience and the training of others in the organization needs to be nearly instantaneous, with negligible costs.
In 2004 we began the Story Representation and Management project, a new research effort at the Institute for Creative Technologies at the University of Southern California to develop intelligent technologies for automating the labor required to create multimedia learning environments based on the real-world stories told by members of an organization. We focused specifically on three areas that required the greatest amount of time and effort in the development of the U.S. Army training applications mentioned above. First, story capture technologies were developed to automatically identify stories in discourse, obviating the need to collect these stories from practitioners through directed interviews. Second, story retrieval technologies were developed to automatically locate stories in a corpus that are directly relevant to particular learning objectives. Third, new designs for interactive story-based learning environments were explored, aimed at minimizing the costs of developing effective computer based training. Collectively, these technologies constitute a story management pipeline, and represent the first steps toward the realization of the vision of story-based organizational learning described in section 1. Our efforts in each of these areas are described in the next three sections.
Technologies for Story Capture
The primary methods that are currently used to gather stories for use in organizational knowledge management and training applications involve face-to-face interviews with subject-matter experts, e.g. as part of a cognitive task analysis [Clark, 07], in small group meetings [Snowden, 00], or with directed story-elicitation interviews [Gordon, 05a]. The facilitator/interviewer that participates in these methods is typically not a member of the organization, and their primary role is to ensure that a large number of stories are communicated and recorded that are relevant to specific knowledge management or training needs of the organization during the course of the interview. However, studies of casual storytelling in workplace environments have shown that stories relevant to the organization are common in everyday, non- facilitated conversations among co-workers [Coopman, 98]. If these stories could be captured directly from conversations among co-workers in an automated manner, then the costs associated with collecting stories relevant to organizational practices could be dramatically reduced. Although there are substantial privacy and user-interface challenges associated with extracting stories from conversations among co-workers, our research efforts in this area were specifically focused on technical feasibility.