Mining Conversation Trails for Effective Group Behavior

Abstract

We will study group-decision making processes in order to develop a suite of novel models and tools that can offer specific insights into the group decision-making process. Our key contributions lie in the treatment of human conversation and discussion as trails over a concept graph. With this perspective, an individual s ideas (as expressed through language) can be mapped to explicit entities or concepts, and a single argument or train-of-thought can be treated as a trail/walk over the nodes in a concept graph. Therefore, within a group discussion we will first translate the stated positions, arguments, and stories into a set of distinct paths over a concept graph. Once in graph form, group dynamics can be analyzed as a new type of graph mining problem where agents synchronously traverse concepts, and where existing graph mining methods can be applied to answer many new, interesting questions about the nature of human discussion. Unlike existing network models where nodes represent individuals and information flows over the edges between the individuals, our key idea is to flip this model such that network nodes represent concepts over which individuals walk during a group discussion. In this new mode of thinking important scientific questions will be addressed -- although we expect to raise more questions than we answer. We look at this old problem in a new way by tackling fundamental scientific challenges we will explore a fertile area of future scientific research.

Document Details

Document Type
DoD Grant Award
Publication Date
Oct 18, 2018
Source ID
W911NF1710448

Entities

People

  • Timothy Weninger

Organizations

  • Army Contracting Command
  • United States Army
  • University of Notre Dame

Tags

Fields of Study

  • Computer science

Readers

  • Applied Combinatorial Optimization and Logic Circuit Design.
  • Educational Psychology
  • Theoretical Analysis.