Maximizing the Collective Intelligence of a Network Using Novel Measures of Socio-Cognitive Diversity
Abstract
We seek to explore the degree to which it is possible to augment Òwisdom of crowdÓ effects by developing novel, theory-based measures of socio-cognitive diversity and using them to select smaller, smarter sub- crowds. Socio-cognitive diversity refers to differences in individualsÕ prior beliefs and information sources, including information acquired through social interaction. We will consider the case of networked crowds in particular Ð that is, groups of communicating individuals who share information in the process of arriving at a judgment (for example, a group of military intelligence analysts who work together to predict the location of a high value target). We focus on networks because (a) real-world analysis and decision-making typically involve some degree of collaboration; (b) communications among members (i.e., who said what to whom) constitute a rich data source from which measures of diversity can potentially be extracted using automated methods. To accomplish the above goals, we will develop and compare several different methods for characterizing crowd diversity, including: a social method, which infers diversity through analysis of social network data (with the idea being that individuals from disparate regions of a network are more likely to possess diverse information); a cognitive method, which infers diversity based on the overlap of topics and information sources cited by individuals in their communications with one another; and a hybrid socio-cognitive method, which combines the two. In developing these various methods, we will use publically available social media data sets (e.g., Twitter) and model problem domains (e.g., fantasy sports) that allow us both to compute diversity and to measure its relationship to real-world judgmental accuracy. As the project progresses, we will broaden our approach to explore additional data sets and problem domains (e.g., geopolitical events).
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 04, 2018
- Source ID
- W911NF1610300
Entities
People
- Brandon Minnery
Organizations
- Army Contracting Command
- United States Army
- Wright State University