Empirical Investigations of Trust-Related Systems Vulnerabilities in Aided, Adversarial Decision Making

摘要

In modern military environments, command-and-control decisions are increasingly supported by information systems which collect, analyze, and display information from multiple sources and sensors, to give decision-makers real time information about an evolving tactical situation. Aided-adversarial decision-making (AADM) refers to military command and control decision in such environments, in which computerized aids are available to groups of co-located and distributed decision-makers, and in which there is a potential for adversarial forces to tamper with and disrupt such aids. In aided, adversarial, decision-making environments, various threats from and offensive opportunities for Information Warfare (IW) activities may exist. In these situations, it is crucial to understand the effect of degraded or altered information on human decision-makers, particularly when that information may be intentionally manipulated. The research described in this report continues two prior phases of research which focused on defining, characterizing, and (where possible) modeling the dependencies and vulnerabilities of AADM on components of information, and considered the role of the human decision maker in AADM, developing a theoretical framework to investigate issues of trust in AADM, and a scale to measure human-automation trust. This report presents research which describes and further develops the theoretical approach begun earlier, describes the completed trust scale, and describes an experimental test bed and an initial experiment which tested the theoretical framework developed. Additionally, an initial description of how cultural issues in AADM can be represented by formalisms in different decision-making models is presented and experimentation in the area of graphical data presentation and trust in AADM is described.