Citation for published version:
Jones, S & O'Neill, E 2009, Context-aware messaging: how personal, spatial and temporal constraints affect
text-based communication. in Proceedings of the 8th International Conference on Mobile and Ubiquitous
Multimedia , 2009. Association for Computing Machinery, New York, USA.
https://doi.org/10.1145/1658550.1658558
DOI:
10.1145/1658550.1658558
Publication date:
2009
Document Version
Publisher's PDF, also known as Version of record
Link to publication
Publisher Rights
Unspecified
"© ACM, 2009. This is the author's version of the work. It is posted here by permission of ACM for your personal
use. Not for redistribution. The definitive version was published in Jones, S & O'Neill, E 2009, 'Context-aware
messaging: how personal, spatial and temporal constraints affect text-based communication'. in Proceedings of
the 8th International Conference on Mobile and Ubiquitous Multimedia , 2009. ACM, New York, USA.
http://doi.acm.org/10.1145/1658550.1658558
University of Bath
Alternative formats
If you require this document in an alternative format, please contact:
openaccess@bath.ac.uk
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Download date: 10. Aug. 2022
Context-Aware Messaging: How Personal, Spatial and
Temporal Constraints Affect Text-Based Communication
Simon Jones, Eamonn O’Neill
Dept. of Computer Science,
University of Bath,
Bath,
BA2 7AY, UK
{s.jones2, eamonn}@cs.bath.ac.uk
ABSTRACT
There are academic and commercial drivers for context-
awareness to play a prominent role in the future of mobile
services. Implementing a complete model of context
remains an unsolved problem, however, some contextual
elements such as person, time and place are relatively easy
to identify. We develop a simple context model
incorporating personal, temporal and spatial dimensions
and apply it to a context-aware text messaging service. We
report a field study of the service, investigating how
applying these fundamental contextual constraints to
messages can affect the nature of communication between
participants. The results suggest that although contextual
constraints are not appropriate for all types of message
content, they increase opportunities for situated chat in
public spaces, improve group awareness between peers and
facilitate conversations between people, some of whom
would not otherwise communicate with each other.
Author Keywords
Mobile messaging, SMS, context-aware, location-aware,
mobile social software
ACM Classification Keywords
H.4.3 Communications Applications
INTRODUCTION
People routinely rely on voice calls and SMS messages to
communicate with each other. Voice calls offer instant,
synchronous interaction, whereas SMS offers us the ability
to interact with each other asynchronously, at a convenient
time. Time, however, is not the only factor that dictates
whether we choose to initiate, respond to or act upon
communications. People may wait until they are at a
particular location, with a particular person or performing a
particular activity to do so. Although technology currently
allows us ‘anytime, anywhere’ communication [8], we
don’t necessarily wish to communicate or be communicated
with all of the time, everywhere. Furthermore, a
communication may only be relevant in a particular
context. Filtering messages according to contextual
constraints may alleviate information overload, decrease
‘spam’ [4], reduce the burden of interruptions to a user [6]
and overcome problems with forgetting to act upon a
communication by delivering it at the moment when it is
most relevant.
Although SMS continues to thrive, Blom et al. [3] point out
that, in the face of quickly evolving mobile services, it is
unlikely that the nature of SMS will remain unchanged.
Additional features such as multimedia attachments have
already been introduced, and more developed context-
awareness is a potential further stage in the evolution of
text based messaging. At present, most forms of
communication take only the simplest ‘personal’ context
into account, i.e. the identity of the recipient. Thus, the
sender can place constraints on who receives the message.
To develop full context-awareness we would require a
complete model of context; a model which can capture all
of the measurable components of a given situation [9]; for
example not just who should receive a message, but also
where, when, why and how they should receive it.
Although sensor technologies are beginning to allow us to
capture important contextual information there is still a
reliance on inference rules, artificial intelligence or
extensive training data to determine or predict those
elements of context that cannot be accurately measured [2].
Kaasinen [8] suggests that before we can make the giant
leap towards complete context-awareness we must first take
a smaller step towards developing services that use more
easily measurable elements of context, such as location.
RELATED RESEARCH
Location is an important contextual element. Many context-
aware applications being developed focus on location-
awareness. Several free and commercial applications such
as Zhiing, Dodgeball and GeoMe have incorporated
location-awareness and it is becoming increasingly
common for mobile devices to feature location-sensing
technologies. Abowd et al. [1] note that ‘nobody questions
the value of incorporating context into (Ubicomp)
application development, particularly when the context is
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
MUM09, November 22-25, 2009 Cambridge, UK.
Copyright © 2009 ACM 978-1-60558-846-9 09/11 $10.00
the location of individuals’. Locations, particularly in an
urban environment, are often complex, with social
protocols, conventions and values attached to them. These
properties may have strong influences on the behaviours of
those within them, or on the appropriateness of
communication that takes place. For example, a message
promoting items on sale in a particular store may be
considered inappropriate when received at one’s place of
work, but more appropriate when one is standing outside
that store. Perry & Shangar [12] highlight this point by
discussing the importance of spatial communication. It is
common for us to leave notes for each other at different
locations within our environment, for example a note on the
fridge reminding you to buy milk, a message slid under a
door or an advertisement pinned to a board.
Although SMS is primarily used to facilitate one-to-one
communication, location-based messaging can allow
messages to be sent to an environment inhabited by a much
larger number of users. LAMMS [3] is a system that was
developed in order to explore a form of location-based
communication. The system worked alongside SMS,
allowing friends and strangers to communicate with each
other by posting and receiving messages that could be tied
to locations. These types of messages are an important
form of human interaction, with contextual information
influencing and informing our interactions [5]. The
location of these messages gives them further meaning and
places them in context, however, the properties associated
with a particular location may fluctuate over time. LATTE
[11] introduces the concept of temporal constraints applied
to messages as well as the spatial ones used by location-
aware services. LATTE is a location and time triggered e-
mail system in which e-mails are extended to include
dynamic consideration of location and time, to determine
the appropriate recipients for messages. Nakanishi et al.
[10] presented CAMS, a context based dynamic messaging
system which used an amalgamation of schedule and
location information of SMS, voice and e-mail recipients to
redirect communication to an appropriate device or address.
The system required users to register rules about which
means of communication were appropriate in certain
locations or at certain times. Messages and calls were
filtered to the appropriate destinations using these rules.
The Defined Delivery system (DeDe) [7] is a mobile phone
messaging system that also uses contextual parameters such
as time and location to define the context in which a
message will be delivered. This system was trialed with a
socially tight group of 7 individuals, revealing several novel
messaging practices such as anticipatory greetings,
contextually sensitive prompts and relieving mental load. It
also exposed some usage barriers for context based
messaging systems, most notably a sender’s uncertainty of
a message being successfully delivered. The DeDe study
did not investigate how such a system might affect
communication between individuals who are less familiar
with each other, for example by constraining messages to a
location such that they can be received by anybody within
that location. In this paper we report a field trial of a
similar system of our own, which supports both person-to-
person and person-to-location messaging. In addition the
field trial reported here involved a group of participants
with varying degrees of familiarity and communication
with each other, extending the findings of the DeDe study
for less socially tight groups. Our study also differs from
[10] and [11] by allowing authors of messages to specify
unique contextual constraints to each individual message,
rather than the context for delivery being determined by the
system, or pre-defined rules created by the recipient.
CONTEXTUAL CONSTRAINTS
We developed a simple context model incorporating
personal, spatial and temporal dimensions and investigated
how placing constraints on messages using this model
affects communication. Although our model is not a
complete representation of context, we consider the three
contextual dimensions discussed to be fundamental
elements of any context model. Furthermore, they are
measurable and do not rely on inference.
Figure 1 shows the different combinations of constraints
that can be applied to messages using personal, spatial and
temporal dimensions of context. The personal dimension is
divided such that a message can be sent to a single
individual (unicast), a specific group of people (multicast)
or to everyone using the system (broadcast). Similarly, the
spatial dimension is divided such that messages can be sent
to a single location, multiple locations or available
anywhere. The temporal dimension is divided such that a
message is available either at one specific moment in time,
during a period in time or constantly available at any time.
These constraints can be combined, giving many possible
contexts for message delivery.
Figure 1. Contextual Constraints Model
SYSTEM DESIGN
Our messaging system integrates a Bluetooth-based
location detection system, a messaging application installed
on GPRS enabled phones and a messaging server.
Messages sent by a user are transported over their phone’s
GPRS connection to the messaging server, where they are
stored in a database. Users’ phones automatically check for
new messages at this server periodically (by default every
60 seconds). Any messages found for that user (i.e.
matching personal constraints) are downloaded to her
phone, however these messages are not necessarily
displayed immediately. Messages are cached until the
phone detects that the user’s current context matches all of
the contextual constraints placed on the message, at which
point the user is alerted to its delivery.
For our prototype system, Bluetooth beacons within the
environment serve as location markers. These are devices
that simply transmit a uniquely identifiable Bluetooth
address. Although other more accurate location sensing
technologies exist, such as GPS, the 10-20m coverage of
each Bluetooth beacon allows us to mark specific locations,
both indoors and outdoors without having to specify GPS
co-ordinates associated with a particular location or limit
ourselves to locations with a clear GPS signal. The phones
using the messaging application repeatedly scan for other
Bluetooth devices and when a location marker is found the
message cache is checked. Any messages that are
constrained to be delivered at that location and time are
then delivered. Each time an ambient device is detected the
scanning device also requests its user defined name.
Location specification within our system relies on the user
being able to associate a beacon name with a place that they
are familiar with.
In order to choose where to leave a message, the message
sender selects a location from a list of ‘favourites’. The
user can add a location to the list when she is in range of
that location’s Bluetooth beacon, however, to avoid
problems with users forgetting to add locations, the
favourites list is also populated automatically according to
the user’s frequency of visits to a location. Since the
system is able to detect both mobile and static Bluetooth
devices, the favourites list may also contain references to
other Bluetooth devices that are encountered. For example,
the mobile phone of a work colleague might be added to the
list, or a laptop in a bedroom at home. By adding these
devices, users are able to leave messages not only at
locations they frequently visit, but also attached to people
they spend time with or devices they use.
EMPIRICAL STUDY
The system was installed on 9 mobile phones and each
phone was given to a person who had agreed to participate
in the study. The participants (referred to in this paper as
participants A-I) consisted of 6 males and 3 females with
ages ranging from 19 to 24, all of whom were University
students. All participants were accustomed to using text-
messaging services on a daily basis. The relationships
between individuals in the group of participants included a
couple, housemates, friends, course mates and
acquaintances. By including participants with varying
levels of friendship and familiarity, we had a means of
assessing the service’s impact on communication across
these relationships.
Figure 2 illustrates that the Bluetooth signals from each
beacon can be detected within a certain radius. Five
beacons were positioned at locations that were identified as
being frequently visited by all of the participants within the
study: the university entrance, library, bar, shop and
computer science department. Each participant was also
given a beacon to place at home.
Before the participants began to use the system they
completed a pre-study questionnaire/semi-structured
interview to assess the nature and extent of their existing
interactions with the other participants. They were asked to
describe their relationship with each of the other
participants and give details of their use of SMS, providing
a basis for comparison when analysing changes in factors
such as message frequency and content.
The study ran for 20 days, during which the participants
divided their time between their homes in the city and the
university campus. On the first day users were provided
with the messaging service software and given an
introductory handout and briefing. Each participant was
given a description of the system, along with a list of
example scenarios in which it could be used. Participants
were not instructed to complete particular tasks or told how
often to use the system. They were asked to use the
context-aware system exclusively instead of SMS during
the study, even if they did not apply contextual constraints
to messages, so that we could log their communications.
Only two of the participants (the couple) used SMS at all
during the study. Participants were also discouraged from
placing unnecessary constraints on their messages,
providing us with greater insight into how and when they
chose to use context in their communication.
Participants were also interviewed post-study, including
Figure 2. Bluetooth beacons act as location markers
questions on how they felt their communication with others
had been affected by their ability to place contextual
constraints on messages.
RESULTS
Figure 3 shows the number of context-aware messages sent
on each day of the study. During this period all participants
sent a total of 98 messages. The majority (66%) of
messages were sent within the first week of the study. The
number of messages sent decreased over the following two
weeks, with occasional ‘spikes’ in messaging activity.
Message traffic peaked on day 5 with 17 messages,
approximately 20% of all messages being sent on that day.
Users reported that the more intense use of the system
within the first week of the trial was due to factors such as
the ‘novelty’ of the application and their desire to test it and
all of its functions. Users admitted that not all of the
communication during this initial period was entirely
necessary and would probably not have occurred without
the introduction of the service. This behaviour is typical of
many new services and applications in the early stages of
adoption. One of the users described the ability to leave
messages attached to locations or objects and to schedule
their delivery as ‘introducing a sense of discovery’ to their
messaging services. Most users reported that they found
the use of the service a fun experience at first, because of
the mystery of not knowing where or when they might
receive a message, or what messages were out there waiting
for them, although they stated that they did not go actively
looking for them. These users also stated, however, that the
novelty of using the service had begun to wear off in the
second and third weeks of the study. They commented that
the experience offered by the service and the motivation for
using it had changed from fun to convenience.
Figure 3 shows that use of the system dropped dramatically
on days 7 and 8, a Saturday and Sunday. A similar
decrease in the number of messages sent using the system
was also observed during the following weekend, on days
14 and 15. The post-study interviews revealed that users
were less inclined to use the system at the weekends
because the nature of their activities changed. They
explained that the weekends were used to perform different
activities to those on weekdays and that the people they
spent time with and communicated with were different. In
many cases these were people who were not taking part in
the study.
There was a second spike in messaging activity at the
beginning of the second week, on day 9. The majority of
the messages on this day were left by participants C and H,
at locations on campus. They started conversations about
the food on offer in the Student Union shop and the lack of
people in the Student Union bar. After the initial burst of
messages on day 9, there was a gradual decline in the
number of responses throughout the week. Users felt that
this was because conversations left at locations were very
context-specific and did not often produce opportunities for
the conversation to change topic. They felt that because
they were posting a message that would be received within
a specific context, it would be inappropriate for them to
send the conversation in another direction. So, as
conversation became strained, the number of messages sent
gradually decreased.
COMMUNICATION BETWEEN PARTICIPANTS
The pre-study questionnaires and semi-structured
interviews were used to collect estimates of how frequently
the participants communicated with each other using SMS
before the study. Since they were instructed to use context-
aware messaging and to avoid using SMS during the study,
by comparing their estimates of SMS use to their recorded
use of the context-aware system, changes in the frequency
of communication could be estimated.
For each participant, Figure 4 shows the average number of
messages sent per week, using both self-reported SMS and
logged context-aware messages. For the couple
(participants D and F), who were the most frequent SMS
users before the study, use of context-aware messaging was
much lower than their previous reported SMS use. They
were, however, the only participants to use SMS during the
study. The qualitative data collected pre- and post-study
provided intriguing insight into the reasons.
Figure 3. Number of Messages Sent Each Day
Figure 4. Comparing SMS and Context Aware Use
