Investigation of rank aggregation methods applied to military value of information decision support

Abstract

Valued information can significantly influence decision making processes. On military battlefields, the Value of Information (VoI) assessment methodology prioritizes information importance based on applicability and relevance of information to a specific operation. Complementary pieces of information from multiple reliable sources can always increase the probability of existence for an event or situation. Human decision making is a process of selecting amongst alternative methods of action toward attaining a goal by involving human intelligence. In many cases, a human expert considers mathematical and quantitative information obtained from relationships between systems, then mixes them with personal criteria and constraints grouped as “Human Principle of Choice” or “Prior Experiences and Training”. The problem of aggregating multiple opinions from multiple individuals presents a decision-making challenge. Within the realm of military intelligence analyst decision support, this work considers the matter of aggregating information by first starting with the task of ranking multiple independent judgments from multiple independent individuals. The overarching goal of this research is to investigate the use of heuristic and probabilistic aggregation model as well as preferential voting methods in ranking with respect to the Value of Information (VoI) problem domain. This research presents discussion about ongoing VoI research and an exploratory experiment using the Bayesian inference for ranking data. Additionally, other preferential voting systems in rank aggregation such as Borda count, Condorcet method (Condorcet/Schulze voting), and Instant-runoff voting are explored. The results from this research demonstrate the efficacy of Bayesian Thurstonian and other ranking and aggregation models in comparison with fuzzy logic rules used in prototype VoI systems and clearly highlight the “wisdom of the crowd” effect.