Introduction and Overview
The Intelligence Community Studies Board (ICSB) of the National Academies of Sciences, Engineering, and Medicine convened a 2-day virtual workshop on December 17 and 21, 2021, to explore insights from world-class experts and technologists familiar with the extensive range of issues associated with anticipating rare events—those characterized by a very low probability of occurring—of major significance.
Intelligence preparation of the battlefield includes information about, and warnings related to, significant rare events.
Intelligence preparation for high-risk environments catalyzed by rare events requires some of the most difficult types of time-sensitive data collection, situational interpretation, and analysis to establish priorities using multiple criteria. Developing decision advantage through the analysis of rare events requires the consideration of both qualitative and quantitative factors, such as the possibility of the event, the elements and circumstances of the event as it occurs, and the consequences of the event.
Because rare event analysis is often plagued by data sparsity, noise, ambiguity, and contextual bias from data sources, it is not always possible to properly align opportunistic sets of data about a rare event gathered separately and independently across various contexts. Additionally, because the data often are collected opportunistically and not in a focused way, it can be incomplete or of poor quality for this purpose. Taken together, these factors can result in reliance on interpretive operations that push the limits of purely data-driven analysis and lead to vulnerability to deception or bias-related misinterpretation. Conversely, the use of highly focused, specific, rare event–directed sensing and data collection resources intended to increase sensitivity can increase the probability for missed signals and false alarms. In the case of operational environments, a rare event would be characterized by any extreme disruption of routine objectives, context, and other characteristics of those operations and environment.
In summary, data collected about an unpredicted event can range from being opportunistic, contextually constrained, and relatively uncontrolled to highly focused and specific. In either case, some data about the event, possibly more generally useful, could easily go unmeasured or remain unknown, given the particular operational context. Thus it might be difficult to align specific operational data about a rare event with other data collected for other purposes about that particular or similar event. Correlations can become tenuous, entirely illusory, or even become lost.
The U.S. Defense Threat Reduction Agency (DTRA)-sponsored workshop (see Appendix A for the original statement of task) described in the upcoming chapters was forged out of the need for greater focus on rare events. In turn, this workshop also called for greater institutional consideration and incorporation of the way rare event data are both collected and analyzed.
Over the course of the 2-day workshop, the speakers discussed analytical methods, computational advances, data sources, and risk assessment approaches for anticipating rare events, including natural disasters, pandemics, anthropogenic threats, and widespread technological change.
Christopher Barrett, chair of the planning committee, opened the workshop by noting that the idea of anticipating rare events of major significance immediately raises a variety of complicated issues, such as how to know something is major, how to look for it when it has never happened before, or how to know which of the many minor events that occur will emerge and become significant. Each of these vexing issues is begging for better solutions. He also mentioned the challenge of assigning responsibility to individuals, organizations, and even artificial intelligence systems for making judgments and interpreting situations as they pertain to anticipating and preparing for a rare event of major significance. He then introduced Theodore Plasse, chief of the Analysis and Plans Division at DTRA, to provide background and perspective from the sponsor.
Plasse explained that the Department of Defense (DoD) created DTRA in 1998 to provide better predictive analysis that would enable actions to counter weapons of mass destruction (WMDs). His team started with a predictive analysis that forecasts enemy courses of action based on space and time, geographic positions, where things move around, how long they take to move, and when they will come together. With some reworking of this general model, the team was successful at satisfying the task it was given, and it has since worked with the Intelligence Advanced Research Projects Activity (IARPA) to further refine this model for conducting predictive analysis for WMD. This workshop came about, said Plasse, after he attended ICSB meetings and spoke with its members about the possibility of expanding DTRA’s work beyond WMD.
To conduct this workshop, ICSB appointed a workshop planning committee to identify potential speakers and design the workshop agenda. Planning committee members and National Academies staff worked with the sponsor in advance of the meeting to refine the workshop’s topics for discussion. Approximately 85 participants, including the planning committee, invited panelists, national security community staff and officials, and National Academies staff, participated in the virtual workshop.
The workshop rapporteurs prepared this proceedings as a factual summary of what occurred at the workshop. The planning committee’s role was limited to planning the workshop. Statements, recommendations, and opinions expressed are those of individual presenters and participants, and are not necessarily endorsed or verified by the National Academies, and they should not be construed as reflecting any group consensus.