Modeling User Response Performance to Vibrotactile Alerts

Vibrotactile alerts are widely studied and implemented in today’s user interfaces, especially in environments saturated with visual and auditory information. However, designing vibrotactile alerting systems is challenging due to the vibrotactile research literature being siloed and inconsistent, as well as the lack of available research tools for predicting user response performance before developing physical prototypes. This dissertation describes steps taken to address these two research gaps. A scoping literature review was performed to synthesize the literature on individual aspects of alert design, including signal frequency. Additionally, the predictive performance of the noticing-saliency-effort-expectancy-value (N-SEEV) model with a novel tactile module was evaluated using data from a dedicated human subjects experiment in a simulated driving environment. From the scoping literature review, ground transportation was the most common application domain, and strong correlations between signal characteristics and performance metrics were not found. A minimum reporting standard or guideline, application-specific systematic reviews, and additional research on vibrotactile signal designs are recommended to resolve observed problems with the reproducibility and comparability among studies and signals. As for the model validation study, after comparing the extended N-SEEV model’s predictions with observed metrics from 30 participants, predicted percent dwell time and miss rate were each highly correlated with the respective observed values. Predicted noticing time did not correlate with or share a similar distribution to observed noticing time. Based on the results, the base N-SEEV model is recommended for additional evaluation in visual-only driving environments and tasks. Accurately predicting noticing time to vibrotactile alerts will require additional parameters in the tactile module, a new model structure/framework, and/or a new behavioral or physiological marker for noticing. Researchers and practitioners can use the insights from both studies in this dissertation to guide the development of future vibrotactile alerting systems and human performance models to predict user response time and accuracy to those alerts.