Towards an in-car voice assistant for frustration mitigation

Abstract

Background: Frustration is an emotion that can turn into anger and (road) rage. These negative emotions and the aggressive driving behaviors associated with them are in turn relevant causes of accidents and road fatalities. To address this, the project F-RELACS (Frustration Real-Time Recognition for an Adaptive In-Car System) funded by the German government aims to demonstrate a frustration-aware assistant that estimates the user’s momentary level of frustration and offers tailored support via a voice-user interface to mitigate frustration. Speech interventions seem particularly promising, because verbal communication with human passengers is familiar. Additionally, the cognitive strain burdened by spoken language and auditory interaction is quite little while executing at the same time the primary driving task. So far, little is known about the exact nature of speech interaction with voice user interface assistants in order to reduce the driver’s frustration. Method: Theoretical basis of our experimental study is Gross' process model of emotion regulation. In the context of F-RELACS, two use cases were developed serving in this study for testing effects of specific emotion regulation strategies against a control condition without intervention. In the first use case, the effect of situation modulation is investigated in automated vehicles (SAE level 4), where users experience frustration dealing with the navigation system. The second use case examines the effect of cognitive change in a manual driving context with frustrating experiences due to traffic. In our between-subject design study not only the regulatory strategy is manipulated (intervention vs. no-intervention) but also the time of intervention (early vs. late start of intervention). Results: Results show that speech interventions reduce the level of frustration: The frustration level at the end of the experiment is higher in the control than in the intervention conditions. In both use cases the frustration level at the end of the experiment is lowest for persons with an early start of intervention. Conclusions: Results of this study show the potential of in-vehicle voice assistants for frustration mitigation both in manual and automated driving. Implications for a refined design as well as for further research will be pointed out.