Dynamic Bayesian Sequential Multi-Group and Multi-Agent Binary Decisionmaking

• Activity Number: 441
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Edsel Aldea Pena*+ and Blake Hill and Bereket Kindo and Fazlur Rahman and Ian Wanda and Wensong Wu
• Companies: University of South Carolina and University of South Carolina and University of South Carolina and University of South Carolina and University of South Carolina and Florida International University
• Keywords: binary decision making ; multi-agent groups ; Bayesian decision-making ; decision theory ; risk assessment ; system's approach to decision-making

Abstract

Many real-life decisions are binary in nature: launch/no launch; deploy/do not deploy; approve/disapprove; yes/no; promote/do not promote; tenure/do not tenure; buy/sell; many others. In democratic institutions, the decision is typically made by several multi-agent groups and with the agents having varying information or decision making abilities. In this talk we consider the problem of choosing an optimal decision when G groups, each consisting of several agents, are tasked to make the decision through a sequential process, with the decision of the last group being implemented. Each agent and group will assess their chosen decision through a common loss function, and each agent is allowed to use information arising from his/her own information gathering process as well as the revealed decisions of each of the agents in the preceding group. A Bayesian approach to the problem will be discussed and its properties will be examined and compared with other alternative decision-making schemes. A particular case is a problem where one group is to make a binary decision but with agents allowed to cast initial "straw" votes which are revealed to all agents prior to the final vote.