Generative Narrative Networks for Strategic Analysis and Forecasting Conflict
Abstract
All cultures tell stories?or, more accurately, no cultures have yet been discovered that do not tell stories. In common usage, the,word ?story? implies a rather limited episode, for instance grandpa telling a story about his childhood or that big battle in World,War II. But in fact, most stories depend on larger narratives for meaning and context. Thus, stories rarely stand alone, but instead, form narrative systems that both constrain and potentiate possibilities for meaning and are a key tool for understanding the world., Narrative systems embody a unique form of rationality: People understand events and their places in them and ?what is happening? th,rough a narrative lens. Yet to date, no good frameworks exist for the large-scale analysis and modeling of narrative systems. This,project is intended to fill that gap. Previous theoretical work by our team has shown that real-world narrative system can be conce,ptualized as a directed network. Elements in this network consist of locations, events, actions, participants, and things (LEAPT) th,at are connected. For example, in a children?s story like the Three Little Pigs, a wolf is connected to a pig and both are tied to,a location where there is a brick house. In a contemporary factual story Russia and Ukraine are tied to a region called Crimea (loca,tion) and they take actions toward it like trying to seize it or prevent it from being seized. In both stories, there are many othe,r LEAPT elements and connections, and as a whole this network presents a structured picture of a conflict situation. Our previous wo,rk has also shown that simulation techniques can be used to fill-in information that is missing from such networks?a common problem,in intelligence and strategic planning. For example, we might know that there are two connected entities in a narrative about China?,s Belt-and-Road initiative, but we might not know how strong their connection is. The narrative system can be analyzed to help dete,rmine that information. This project is intended to take this capability from its current proof-of-concept to a working prototype,,through the completion of three main tasks. The first is theoretical and conceptual development. This requires determining how LEAP,T factors relate to other important elements of the narrative context like conflicts,,lso an issue of scaling, in terms of both the boundaries of the narrative system being analyzed and the scope of the LEAPT elements.,The second task is to validate and implement the approach. For validation, we will show that narrative systems for a given conflict,correspond with people?s understandings of that issue through use of focus group interviews. For implementation we will develop nat,ural language processing methods to identify narrative elements in texts (for example news stories) and identify appropriate algorit,hms for analyzing large-scale narrative networks.The third task is to apply the techniques fromtask two to generate functionality fo,r analysts and planners. This will entail support for analysis of strengths, weaknesses, opportunities, and threats (SWOT) and analy,sis of what-if scenarios. For instance, who are the most important participants in a conflict in the sense of making the narrative a,bout it cohere? Are there key relationships that could be weakened in a way that would make the narrative less compelling and cohere,nt? Are the locations which, if denied to an adversary, could make the narrative fall apart or take on a different character? We wi,ll develop ways of answering these questions and test their utility for Navy analysts and planners.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 13, 2022
- Source ID
- N000142212596
Entities
People
- Steven R Corman
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy