Scatterbox:
Context-Aware Message Management
Stephen Knox — Ross Shannon — Lorcan Coyle — Adrian K. Clear
— Simon Dobson — Aaron J. Quigley — Paddy Nixon
Systems Research Group
School of Computer Science and Informatics
UCD Dublin, IE
stephen.knox@ucd.ie
ABSTRACT. Applications that rely on mobile devices for user interaction must be mindful of the
user’s limited attention, which will typically be split between several competing tasks. Con-
tent delivery in such systems must be adapted closely to users’ evolving situations and shifting
priorities, in a way that cannot be accomplished using static filtering determined a priori. We
propose a more dynamic context-driven approach to content delivery. We demonstrate our ap-
proach using Scatterbox, a pervasive computing application we have developed which performs
sensor fusion to derive a user’s current situation. Based on the user’s level of interruptibility,
Scatterbox prioritises and forwards relevant messages to their mobile phone. We draw conclu-
sions from a preliminary evaluation of the system.
RÉSUMÉ. Les applications embarquées dans les appareils portables tels que téléphones ou PDAs
et qui interagissent avec leur utilisateur, doivent prendre en compte la disponibilité de celui-
ci. L’utilisateur exécute en effet en général des tâches n’ayant pas de lien avec celles pour
lesquelles il reçoit une notification de son appareil. Pour de tels systèmes, les priorités de
l’utilisateur doivent être prises en compte pour lui livrer des messages pertinents á un instant
donné. En général, ceci ne peut pas être effectué á partir de filtres calculés de manière statique.
Une approche de livraison de messages, dynamique et dépendante du contexte de l’utilisateur,
est ici proposée. Cette approche est illustrée en utilisant Scatterbox, une application perva-
sive qui permet la fusion d’information afin de déterminer le contexte dans lequel se trouve
l’utilisateur. En fonction du niveau de disponibilité de celui-ci, Scatterbox rend prioritaires cer-
tains messages et les envoie sur son téléphone portable. Nous tirons ces conclusions suite á une
évaluation préliminaire du systéme
KEYWORDS: situation-awareness, context, pervasive computing, sensor fusion
MOTS-CLÉS : situation consciente, contexte, application pervasive, fusion d’information
Revue d’Intelligence Artificiel, pages 1 à 20

2 Revue d’Intelligence Artificiel
1. Introduction
Rich sources of context data are an integral part of building intelligent pervasive
computing applications. A central concept in pervasive computing is that “technol-
ogy recedes into the background of our lives” (Weiser, 1991). To accomplish this, the
classical view of human-computer interaction needs to be extended to include a cloud
of interoperating heterogeneous electronic devices that offer services to the user. This
scenario requires that disparate devices are able to interoperate easily, breaking away
from traditional distribution channels and reaching the user through whichever device
they currently have available. As small mobile devices with built-in wireless capabili-
ties such as mobile phones and PDAs become more widespread, they present an ideal
opportunity to afford ubiquitous communication services to the user.
Despite all of the recent advances in the mobile device space, the majority of of-
fice workers are still tied to their desktop email clients throughout the day. Email
has been overloaded as a means to communicate information to the user, and is now
used to pass along personal messages, work notices, mailing list communications,
calendar reminders, notifications of messages on social networking websites, calls for
papers, news alerts, and more. The constant influx of information from these disparate
sources, with different (and sometimes shifting) priorities, have left many users feel-
ing overburdened and out of control (Whittaker et al., 2006). Many find themselves
constantly tending their email rather than doing meaningful work. Deciding which
emails are the most important ones to read first can be a challenge in itself.
Mobile devices are now capable of performing as fully-featured email clients, but
encounter problems due to their limited screen size, and the “drastically short term”
limited attention span of mobile users (Oulasvirta, 2005). By notifying mobile users
of every new email, people are being increasingly distracted in situations where they
are even less likely to be able to attend to every incoming communication. Because of
this, there is obvious utility to a method of filtering the messages soliciting the user’s
attention. Static email filtering is limiting, as it does not change over time (Zhou et
al., 2005), and thus cannot keep up with the user’s constantly changing situations and
tasks. A more dynamic approach is required.
Context-aware systems can determine many facts from the environment that can
inform their behaviour, for example, who is present, or what task they are performing.
An example of this is a smart meeting room that manages shared projectors (Chen et
al., 2004), or a smart room for elderly people which detects incidents such as falls and
reacts to them quickly (Rialle et al., 1998). Further related research in context-aware
systems is discussed in Section 2.
Decisions in these systems can only be made if there are enough data, of suffi-
cient fidelity, to support them. Therefore there must be a network of sensory devices,
each providing differing inputs, that contribute to the knowledge of the system as a
whole. These devices will often go unnoticed by the user, but they provide valuable
information about the user’s surroundings and the context of their activities.