Introduction
Mobile phones represent an essential element of our lives by assisting us in several day-to-day activities. Since they are always connected to the Internet, mobile phones represent a unique platform for receiving or fetching information anytime and anywhere. This is leveraged by numerous mobile applications, such as email and instant messenger clients, VoIP (Voice over Internet Protocol) services, and social network platforms to provide their core functionalities.
The key to success of such applications, which essentially provide access to a variety of information channels, is to ensure real-time awareness of users about the delivered information. In order to ensure this, mobile operating systems facilitate the use of notifications (as shown in Figure 1.1) that steer users’ attention toward the delivered information through audio, visual, and haptic signals. This is indeed in contrast with the traditional paradigm of pull-based information retrieval and delivery in which the user has to initiate a request for the transmission of information. Notifications are the cornerstone of push-based information delivery via mobile phones as they allow applications to harness the opportunity of steering users’ attention toward the delivered information in order to maximize its effectiveness. Indeed, mobile notifications are presented in a unified fashion by almost all mobile operating systems. Usually, in the current implementations, notifications from all applications are listed on the phone’s lock screen as well as in a notification bar located at the top of a phone’s screen. In order to provide a brief summary of the delivered information to the users, they present a brief summary including the identity of the sender, a short description of the content of the notifications or the event that trigger them, and time of delivery.
Mobile notifications are triggered by humans as well as machines. The former are triggered by recipient’s social connections generally through chat and email applications for instantiating communication between two or more persons, whereas, in the latter case, messages are generated in an automatic fashion by the system processes or the native applications, such as system monitoring utilities, scheduled reminders and promotional advertisements. Nowadays, since a variety of sophisticated sensors are embedded in phones, the data generated therefrom is exploited by applications for not just improving usability but also for proactively signaling users (usually through notifications) about the occurrence of events that are associated with their context. Some common example of context-based notifications are collocation-based advertisements [1, 34] and context-based suggestions [57, 102].
Figure 1.1: Information delivery through push notifications on mobile phones.
1.2 ISSUES WITH MOBILE NOTIFICATIONS
Push-based mobile notifications were introduced in mobile phones to keep users free from constantly checking for (i.e., pulling) new information, as they signal users on the availability of any newly arrived information. However, people receive numerous notifications arriving autonomously at anytime during the day through their mobile apps [75, 98]. Such services provide a clear benefit to users as they facilitate task switching and keep users aware of a number of information channels in an effortless manner. However, at the same time, these notifications are often triggered at inappropriate times as they do not have any knowledge about recipients’ situation. Psychological studies have found that notifications arriving at unsuitable moments often become a cause of disruption for the on-going task [80, 111]. Notifications delivered at the wrong time can adversely affect the current primary task’s execution [8, 21, 22, 81] and users’ affective state [2, 7].
A previous study has found that in order to not miss any newly available information that is deemed important, people are willing to tolerate mobile interruptions [52]. However, various mobile applications exploit users’ attention by triggering a large number of potentially unwanted notifications [75]. At the same time, some studies have demonstrated that not all notifications are accepted by users as their receptivity is dependent on the type and sender of information being delivered [75, 77]. For this reason, notifications that are uninteresting and irrelevant are mostly dismissed (i.e., swipe away without clicking) by users [30, 98]. Some examples include new game invites, app updates, predictive suggestions by recommendation systems, and marketing messages. Furthermore, continuous trigger of notifications at inappropriate time and context becomes a potential cause of annoyance for users. This might lead them to uninstall the corresponding applications [28, 98]. In order to provide an in-depth overview of these issues, we devote a part of the article to the discussion of issues concerning the cost of interruptions for both users and the service providers.
1.3 SOLUTIONS FOR INTERRUPTIBILITY MANAGEMENT
Interruptibility management has attracted the interest of human-computer interaction researchers well before the advent of mobile devices. However, interruptions received on the desktop have very specific characteristics. In fact, because of their very nature desktops are situated in a constant environment and a user’s physical context (such as surrounding people, location and physical activity) does not always change while they are interacting with desktops, whereas mobile devices are carried by users almost everywhere, which makes the physical context of these devices very dynamic. Therefore, interruptibility management for desktop environments is in a sense less complex.
To address the challenge of understanding and modeling users’ complex behavior in terms of interaction with mobile notifications, studies have focused on exploiting the contextual information (captured through mobile sensors) [77, 98]. Indeed, the interaction of users with notifications is conditional on various contextual dimensions, which might be partially captured by means of the sensors embedded in the phones. However, by monitoring users’ interaction with notifications, researchers have been developing a variety of interruptibility management mechanisms for both desktop and mobile environments. More specifically, these studies have been focused on:
(i) understanding factors associated with users’ interruptibility and receptivity to notifications;
(ii) inferring opportune moments and contextual conditions for notification delivery; and
(iii) identifying and filtering notifications that are deemed uninteresting or irrelevant for users.
Consequently, all studies have focused on various challenges concerning the understanding and learning of users’ behavioral patterns in terms of interactions with notifications. However, the characterization of users interruptibility for delivering mobile notifications is still an open problem. We believe that there is still a considerable scope for improvement, for example by exploiting other physical, social, and cognitive factors for modeling users’ notification interaction behavior. Therefore, in this work we discuss previous efforts and the state of the art in interruptibility management system design for both desktop and mobile environments separately. We then present a critical discussion of these approaches and highlight the current challenges and opportunities in the area that must be investigated to build intelligent mechanisms that could effectively trigger the right information in a given context.
In this book we present an overview of the current state of the art in the area of intelligent notification mechanisms that rely on the awareness of users’ context and preferences. The lecture has been designed for a broad readership. It only assumes a very basic knowledge of human-computer interaction and ubiquitous computing concepts. The chapters cover both theoretical and practical aspects related to the implementation of system for intelligent notification management.