CAREER: Modeling the effect of operators' adaptive behavior on system safety

University Of Washington, Seattle WA

Investigators

Abstract

Many technological innovations are designed to increase operator safety by simplifying tasks and user demands in safety-critical situations. However, operators may actually adapt to these systems in unexpected ways that may counteract the intended outcome. In this research, the PI will develop a statistical methodology for predicting operator adaptation to prolonged system use. The test bed application providing the concrete framework for development of the methodology will derive from the driving domain, where a number of safety based systems have already been installed. Driving involves complex interactions between the driver, the vehicle, and the environment, and any breakdowns in these interactions can undermine driving safety. The objective will be achieved through five major components: development of a representative model of operators' adaptive behavior as influenced by technology; quantification of the initiating factors of operator adaptation on system safety by comparing the responses from a field operation test of short term adaptive cruise control (ACC) users with an on-road study of long-term ACC users; acquisition of an understanding how mediating factors and perceptions influence adaptive behavior using surveys distributed to drivers who have used ACC for extended periods; identification of the safety critical situations where adaptive behavior may have negative consequences using a driving simulator study designed to capture collision likely events; and development of a predictive model of adaptation that results from prolonged system use with time-based regression models. The intellectual merit of this project centers on the integration of theory and data to advance our knowledge of adaptive strategies and how that influences system use in unintended ways. The research will extend prior work of the PI and demonstrate why objective and subjective measures are needed to further understand how a person's behavior changes as they interact with intelligent systems. To understand how and why people adapt to information provided by innovative technologies that are continually changing, an understanding of the user's intentions and motivations is needed; the PI plans to fill this gap with time-dependent analyses which have traditionally been limited in human-factors research. Project outcomes will advance our knowledge of the adaptive behavior process, such that those behaviors that counteract the intended benefits of safety systems can be predicted and therefore allow for more robust, effective, and efficient systems to be designed. Broader Impact: Systems that account for the changing strategies of drivers will be more effective in reducing the number of crashes and fatalities in the world. This research will connect the range of applications for analytical techniques used across engineering, econometrics, statistics, and epidemiology. The research activities will be integrated into two graduate courses developed by the PI that focus on designing systems centered on human performance, and analytical methods in human factors engineering, and the PI further plans a variety of outreach activities intended for K-12 students as well as teenage drivers.