Optimizing Multimodal Alarm Design for Attention Allocation in Discrete Monitoring Tasks

Discrete monitoring tasks are common in scenarios such as flight missions, air traffic control, nuclear power plant monitoring, and clinical healthcare. In these tasks, operators perform sequential routine operations while remaining vigilant for anomalies and responding to alarms. Current alarm systems primarily rely on visual (V) and auditory (A) modalities. However, prolonged monitoring and high cognitive demands during emergencies can overwhelm these channels, jeopardizing system safety. Existing studies suggest that integrating visual, auditory, and tactile (T) modalities in multimodal alarms can improve performance under high perceptual load of visual and auditory modalities. However, the effectiveness of multimodal alarms across different scenarios remains debatable, with redundancy costs observed in some scenarios. Limited research on multimodal alarms in discrete monitoring tasks leaves a gap in understanding their specific effectiveness. A challenging scenario in discrete monitoring tasks involves simultaneous routine operations and alarm responses. Operators often prioritize routine operations due to well-established stimulus-response associations, neglecting urgent alarms. Leveraging sensory modalities can modulate attention control modes (ACMs)—categorized as exogenous or endogenous. For instance, auditory presentation of routine operation information may dominate attention through auditory preemption, overshadowing visual alarms, thus exhibiting exogenous control. In contrast, using VA or VAT alarm could introduce competition between auditory information from routine operations and alarms, prompting operators to shift from a stimulus-response mode to an analytical approach prioritizing alarms. While alarm modality design holds potential for enhancing task prioritization, its impact on ACMs in dual-task scenarios remains unclear. This study developed a generalized discrete monitoring task to investigate six alarm modalities (V, A, VA, VT, AT, VAT) under four workload conditions: Baseline, Low workload, High workload with spaced tasks (HighSpac), and High workload with simultaneous tasks (HighSimu). The HighSimu condition specifically examined ACMs during dual-task competition by presenting routine operation information via auditory channel (with its preemption) to induce exogenous control and requiring operators to prioritize alarms using endogenous control. Twenty-two participants performed simulated emergency tasks in a controlled experiment. Data on hits, errors, misses, choice response times (CRTs), perceived workload (using NASA-TLX), user experience (using UEQ questionaire), and ACMs under HighSimu conditions were collected. Key findings include: 1) Hit rates, errors, and misses: Under Low workload, V alarm performed worst, while other modalities showed no significant differences. 2) CRTs: Under Low workload, VAT and VT alarms performed best, while V alarm performed worst. In HighSimu, V alarm had the longest CRTs, with no significant differences among others. 3) Perceived workload: No differences were found under Baseline, HighSpac, and HighSimu conditions. However, VAT alarm outperformed V and A alarms in perceived "Performance" under Low workload. 4) User experience: VAT alarm consistently ranked highest across workload conditions, while V alarm performed the worst. 5) ACMs: No significant differences were observed under HighSimu, but A alarm showed a relative advantage in fostering endogenous control compared to V alarm. These findings provide insights into enhancing operator performance and system safety, with practical implications for optimizing alarm design and selection for discrete monitoring tasks.