Search

What happens to deterrence and escalation when decisions can be made at machine speeds and when states can put fewer human lives at risk? What are potential areas for miscalculation and unintended consequences, and unwanted escalation in particular? This exploratory report provides an initial examination of how artificial intelligence and autonomous systems could affect deterrence and escalation in conventional crises and conflicts.

The Army Quadrennial Defense Review (QDR) Office asked the RAND Arroyo Center to conduct research on the close-support capabilities available in recent conflicts to help inform future planning decisions. Initially, the research the office requested was to include a broad set of close-support capabilities; however, this research subsequently refocused specifically on close air support (CAS) capabilities in recent conflict and the possible implications for the future. The research provides an overview of the requirements process that led to existing CAS capabilities, contains information from warfighters, and includes the results of an assessment using state-of-the-art machine learning methods for coding to extract information on CAS capabilities that were provided to the Army over the past decade or so. This work should be of interest to a broad range of technologists, concept developers, materiel developers, and many others in the defense acquisition community.

Congress requested a study to assess the challenges in fielding U.S. ground combat and tactical wheeled vehicle fleets, including discussion of requirements; capability gaps; technology, operational, and business risks; and recommended actions.

The rationale, design, and operation of the Perceptronics' Group Decision Aid, are described, and some experimental studies of its effectiveness are reported. The decision aid is supported by a PDP 11/45 minicomputer using the UNIX operating system. It features simple individual data entry terminals and a large screen (7 ft) color video display for feedback of computer-generated information. The system's purpose is to guide the group decisionmaking process by selective elicitation of a decision tree which incorporates value and probability inputs from all group members. Conflicts in value judgments are identified and discussions are initiated through use of multiattribute utility analysis. A specially trained system operator, called an intermediator, facilitates group interaction with the aiding program so that group members need have no prior familiarity with computers or decision analysis. Experimental studies are described comparing the performance, behavior, and responses of aided and nonaided groups in a complex international crisis decision problem. Improvements with aiding were noted in the completeness of the decision process and in the satisfaction and confidence of the group participants.

Management of communications between a remote or automated system and a supervisory human operator can be a formidable task. The selection of information for transmission is a recurrent, subjective decision involving many factors – machine state, operator capabilities, communications costs, and channel limitations, among others. An adaptive computer program has been developed which incorporates these various factors into a decision model resident in the remote element. The program is designed to capture the supervisory operator's decision policy by using a training algorithm based on pattern recognition techniques. Some exploratory studies of the technique are described.

This report describes progress on work centered on the demonstration of an interactive computer aid for group decision making. The report includes; (1) a new sensitivity algorithm for decision tree analysis that incorporates information about individual differences in judgmental knowledge, (2) procedures for identifying and resolving utility conflicts, (3) a refined version of the group/machine and intermediator/machine interface and displays, (4) a preliminary problem scenario booklet to be used in the system's empirical evaluations, and (5) a report on the project's software development efforts. The next phase of the program will concentrate on the system's implementation, operational testing, and initial evaluation. (Author)

Distributed Artificial Intelligence is concerned with problem solving in which groups solve tasks. In this paper we describe strategies of cooperation that groups require to solve shared tasks effectively. We discuss such strategies in the context of a specific group problem solving application: collision avoidance in air traffic control. Experimental findings with four distinct air-traffic control systems, each implementing a different cooperative strategy, are mentioned.

Shared decision making between man and intelligent machine is becoming an important part of advanced systems. The research described in this paper is directed toward developing human factors criteria for the man/machine interaction. The research effort includes evaluations of task allocation techniques, information feedback, operator training, and decision analysis methods.

Distributed planning and control techniques provide potential advantages over centralized porcessing in speed, cost, reliability, flexibility, and minimization of long-distance communications for a variety of tasks, including military threat assessment, command and control decisionmaking, disaster relief coordination, and civilian air traffic control. Six different architectures for distribution of control among multiple processors are presented, and the influences of different task environments on each are discussed. The work focuses primarily on the use of distributed planning and control for civilian air traffic control. One architecture, in which each aircraft is controlled by a separate processor, is described in detail in an illustrative scenario. An initial system design is presented in which cooperating 'experts' comprise a processor. These cooperating experts sense or infer aircraft intentions, generate and evaluate plans, control and monitor plan execution, and communicate with other processors.