Identifying and Overcoming Attention Limitations in the Detection and Identification of Alarms in Close Temporal Proximity

Abstract

In many high-risk domains, such as aviation, driving, and space operations, safety depends greatly on the timely detection and correct identification of alarms. However, due to high levels of system complexity and coupling in those environments, a large number of alarms is sometimes triggered within a short period of time, a so-called alarm flood. Alarm floods pose a challenge for operators at various stages of information processing, including detection, identification and response selection. The detection of alarms during an alarm flood can be difficult due to masking effects. Masking occurs when one signal is obscured by the presence of another simultaneous or asynchronous stimulus. To date, various forms of simultaneous and asynchronous masking, such as change blindness and attentional blink, have been studied almost exclusively for only two stimuli and in single-task conditions. The effects of masking in the presence of a large number of signals and in multi-task environments are not well understood. Therefore, the goals of this dissertation were to (1) establish the relative contributions of simultaneous and asynchronous masking to missed and misdiagnosed visual and auditory alarms during routine operations and in an alarm flood, (2) identify the stimulus onset asynchronies (SOAs) at which masking effects are observed in demanding multitask environments, (3) investigate the impact of the number and temporal distribution of alarms on alarm detection and identification and (4) develop and test countermeasures to overcome observed performance breakdowns. To this end, a series of four experiments were conducted using a simulation of a drone-based package delivery service. Both simultaneous and asynchronous masking were observed, intramodally and crossmodally, for visual and auditory alarms. The SOA at which asynchronous masking was observed was longer (around 800ms) than reported in basic studies of the phenomenon. The effects of both simultaneous and asynchronous masking were stronger with an increase in the number of alarms; this increase was more pronounced for simultaneous masking. Simultaneous masking also had a stronger effect on auditory alarms, compared to visual alarms. In clusters of 4 or 6 sequential alarms, the first and last alarms were more likely to be detected and correctly identified. During alarm floods, a speed-accuracy tradeoff was observed, i.e., the response time to alarms was shorter but the identification accuracy was lower, compared to routine operations. Finally, when the criticality of alarms was mapped to preattentive features of the alarm signal (color and pitch), the masking effects were alleviated, especially for clusters of 4 or 6 concurrent visual alarms. The findings from this dissertation contribute to a better understanding of the effects of two types of masking on the detection and identification of large numbers of critical signals in high workload multi-task environments. They help inform the development and expansion of models and theories of human perception and cognition and also highlight that results from basic research do not necessarily generalize to applied settings. From an applied perspective, the findings will provide guidance for the design and evaluation of alarms and alarm systems. Ultimately, the present research helps prevent catastrophic outcomes due to missed or misinterpreted alarms, and thus improves safety in many real-world environments.