Robust influence modeling under structural and parametric uncertainty: An Afghan counternarcotics use case

• Published in: Decision Support Systems 2020-01, Vol.128, p.113161, Article 113161
• Main Authors: Caballero, William N., Lunday, Brian J.
• Format: Article
• Language: English
• Subjects: Behavioral economics; Behavioral OR; Bounded rationality; Crime; Decision analysis; Parameter uncertainty; Parameterization; Persuasion; Prospect theory; Robust decisionmaking; Robustness
• ISSN: 0167-9236; 1873-5797
• DOI: 10.1016/j.dss.2019.113161

An entity often seeks to influence the decisions of others in a system. This dynamic is apparent in a variety of settings including criminal justice, environmental regulation, and marketing applications. However, the central task of the influencing entity is confounded by uncertainty regarding their understanding of the structure and/or parameters of the decisions being made. The research herein sets forth a decision support methodology to identify robust influence strategies under such uncertain conditions. Furthermore, the utility of this framework and its proper parameterization are illustrated via an application to the contemporary, global problem of the Afghan opium trade. Utilizing open source data, we demonstrate how counternarcotic policy can be informed using a quantitative analysis that embraces both the bounded rationality of the economy's decisionmakers and the government's uncertainty regarding the degree of their deviation from perfect rationality. In this manner, we provide a new framework with which robust influence decisions can be identified under realistic information conditions, and we discuss how it can be used to inform real-world policy. •A quantitative decision support framework to design influence policy is developed.•Favorable courses of action are distilled under conditions of deep uncertainty.•Uncertain agent behavior is characterized utilizing Cumulative Prospect Theory.•The method's utility is demonstrated in an Afghan counternarcotics case study.