Ambient Notification Environments

Summary

Reports indicate that many users interact with their smartphone and wearable devices more than 100 times per day. Oftentimes, these interactions result from direct notifications, presented on the screen or via sounds. New communication applications (e.g. from email to WhatsApp) has increased the frequency of notifications, while social media applications are inherently motivated to entice repeat visits and interactions. The end result is that more and more systems as well as applications compete for the user’s attention using notifications. Projecting this into the future, it is apparent that current implementation schemes that rely on direct notifications will not scale. A simple extrapolation of the rate in notifications suggests that a near future whereby users will only attend to notifications with no time leftover for productive work. Therefore, a radical restructure of notification delivery is necessary - specifically, one that keeps the user in-the-loop without consuming all of the user’s attention. This is a key challenge. If no alternatives to direct notifications can be realized, current visions of ubiquitous computing and smart environments are likely to be unrealistic, given their anticipated undesirability to the end-user. In the Dagstuhl Seminar 17161 "Ambient Notification Environments", we looked at how novel approaches to notification delivery can address the above issue. We brought together researchers and experts that understand the technical, psychological, and social aspects of notification systems. This facilitated a broad discussion that was fuelled by the discussants’ joint expertise in mobile and smart home technologies, in ambient sensing and presentation, and psychological models of human attention, to name a few. This discussion brought to the foreground, many of the underlying challenges that deserves further research. On the hand, a technological push in novel communication and smart devices drives an immediacy in user interaction. On the other hand, this will result in a higher demand on human attention. In the current report, we document the diverse approaches that were proposed as an alternative means towards notifications that are more personalized and contextualized, with the express purpose of reducing user effort. The central aim of the seminar was to understand the challenges and questions that we face when designing future interactive systems and to overcome the mounting notification problem. In addition to the ideas raised in the individual presentations and the group sessions, we jointly identified the following research questions and challenges:

1. How can notifications be designed to be simultaneously non-intrusive and yet noticeable?
2. How can artificial intelligence be designed to provide effective context-aware notifications?
3. What is a suitable notification architecture for integrating user devices, smart (shared) environments, and personal data without compromising personal privacy?
4. What is a suitable conceptualization for notification systems and how will a taxonomy for classifying notification delivery that is centered on user preferences and privacy look like?

Overall, the seminar contributed to a common vision of how interactions between human and systems can and ought to progress - one where technological progress does not necessitate an ever growing burden on the user’s attention. In this report, we balance an analysis of the underlying drivers and problems, ideas for novel conceptual and technical approach, and most importantly a set of questions and research challenges in this domain that merit the attention of like-minded researchers.