Exploring self-interruptions as a strategy for regaining the attention of distracted users

Abstract

In this paper we present a first exploratory study investigating the effects of a contingently self-interrupting vs nonself-interrupting virtual agent in a smart home environment who transmits information to a human interaction partner. We tested the hypothesis that self-interruptions are a strategy for keeping the user's attention, as measured by postinteraction information recall. Interestingly, our experiment does not allow us to confirm this hypothesis. In fact, users found the self-interruption strategy to be less-likeable. From our observations, we draw suggestions for future implementations of attention-retainment strategies.

Interaction Guidelines for Personal Voice Assistants in Smart Homes

“…Additionally, (3) this additional information allows the monitoring of the interaction partner. In order to have a meaningful interaction, a good deal of different feedback information should be taken into account [25]-whether to verify that the interaction partner understood the current dialogue act or to be able to react on a distracted interaction partner [26]. Feedback signals, such as gazing behaviour, head nods, and verbal backchannels allow the system further insight into the mental states of the user.…”

Do Hesitations Facilitate Processing of Partially Defective System Utterances? An Exploratory Eye Tracking Study

“…[22] found that users prefer a system that deploys fillers, while [20,23] explicitly point out difficulties of filler synthesis compared to other hesitation types. Lengthening and silence fare well in terms of user feedback in offline evaluation studies [23], but embedded in a smart-home scenario, systems that hesitate only by means of silence are perceived as less friendly than non-hesitating systems [24,25]. For the purposes of this study, we focus on lengthening as a marker of hesitation and uncertainty.…”

The Hesitating Robot - Implementation and First Impressions

“…Here we define attention as a state in which the VFoA matches with current focus of discourse (FoD)(cf. [6]). Incremental processing capabilities.…”

“…[4] with a forecasting model of disengagement lead to less costly disengagement management, but no evaluation of the effect on the human interaction partner, e.g., on task performance or subjective rating of the robot are conducted. [6,7] evaluated self-interruptions in an smart-home setting. The authors use silence as an attention-regaining strategy whenever the attention of the human interaction partner moves away.…”

“…Whereas in [7] the attention-regaining strategy consists of a simple pausing of the synthesis (while the interaction partner is not attentive), this strategy has several levels of escalation. To counteract the perceived rudeness observed in [6], we integrate this hesitation strategy into our dialogue system for smoother transitions from fluent delivery to hesitation mode. …”