Conference PaperPDF Available

Getting closer: An empirical investigation of the proximity of user to their smart phones

Authors:

Abstract and Figures

Much research in ubiquitous computing assumes that a user's phone will be always on and at-hand, for collecting user context and for communicating with a user. Previous work with the previous generation of mobile phones has shown that such an assumption is false. Here, we investigate whether this assumption about users' proximity to their mobile phones holds for a new generation of mobile phones, smart phones. We conduct a data collection field study of 28 smart phone owners over a period of 4 weeks. We show that in fact this assumption is still false, with the within arm's reach proximity being true close to 50% of the time, similar to the earlier work. However, we also show that smart phone proximity within the same room (arm+room) as the user is true almost 90% of the time. We discuss the reasons for these phone proximities and the implications of this on the development of mobile phone applications, particularly those that collect user and environmental context, and delivering notification to users. We also show that we can accurately predict the proximity at the arm level and arm+room level with 75 and 83% accuracy, respectively, with features simple to collect and model on a mobile phone. Further we show that for several individuals who are almost always within the arm+room level, we can predict this level with over 90% accuracy.
Content may be subject to copyright.
Getting Closer: An Empirical Investigation
of the Proximity of User to Their Smart Phones
Anind K. Dey, Katarzyna Wac*, Denzil Ferreira, Kevin Tassini, Jin-Hyuk Hong & Julian Rojas
Human-Computer Interaction Institute
Carnegie Mellon University
*Institute of Services Science
University of Geneva
ABSTRACT
Much research in ubiquitous computing assumes that a
user’s phone will be always on and at-hand, for collecting
user context and for communicating with a user. Previous
work with the previous generation of mobile phones has
shown that such an assumption is false. Here, we
investigate whether this assumption about users’ proximity
to their mobile phones holds for a new generation of mobile
phones, smart phones. We conduct a data collection field
study of 28 smart phone owners over a period of 4 weeks.
We show that in fact this assumption is still false, with the
within arm’s reach proximity being true close to 50% of the
time, similar to the earlier work. However, we also show
that smart phone proximity within the same room
(arm+room) as the user is true almost 90% of the time. We
discuss the reasons for these phone proximities and the
implications of this on the development of mobile phone
applications, particularly those that collect user and
environmental context, and delivering notification to users.
We also show that we can accurately predict the proximity
at the arm level and arm+room level with 75 and 83%
accuracy, respectively, with features simple to collect and
model on a mobile phone. Further we show that for several
individuals who are almost always within the arm+room
level, we can predict this level with over 90% accuracy.
Author Keywords
Proximity, mobile devices, mobility, smart phone.
ACM Classification Keywords
H.m. Information systems: Miscellaneous.
General Terms
Experimentation, Human Factors
INTRODUCTION
It is without question that we are living in a world where
emerging mobile personal devices and high-capacity
wireless networks are enabling new and innovative
applications that compliment many different aspects of
daily life. Over the past decade, mobile computing has
become ever more present in our society, particularly as
smart phones become more prevalent. By December 2010,
31% of U.S. mobile phone users had smart phones [14] and
this figure is expected to cross 50% by the end of 2011 [7].
This trend holds worldwide, with almost 300 million smart
phones being sold in 2010 [5] and another 500 million
predicted for 2012 [4].
In keeping with this trend, there has been an underlying
assumption in much of the work in ubiquitous computing
that a phone is always with its owner. This assumption, if
true, means that ubiquitous computing systems can use the
phone as a medium for collecting data from users and
communicating information to users, at any time. Further,
this means that the mobile phone is an accurate proxy to
collect contextual information on users’ location and
activity. This assumption was investigated in 2006 [20]
when mobile devices were not as robust and feature-filled
as they are today. At that time Patel et al. found that when
participants had their phones on (81% of the time), they
kept their phone within arm’s reach 58% of that time,
which was less than the participants perceived themselves
as doing, and the phone was within the same room as
participants an extra 20% of the time.
Since this study, five years ago, we have seen the
introduction and mass adoption of Apple’s iPhone and the
Android platform that have redefined the mobile computing
experience, and the operating systems and capabilities of
mobile devices that are available to the average user. We
now live in an era of so-called “smart” phones. These
mobile devices have progressed far beyond a means of
making and receiving phone calls, and for many, have
become an almost indispensable tool to access information,
complete tasks, be entertained and communicate in a wide
variety of ways (e.g., Skype, instant messenger, email). As
a whole, it seems reasonable then to presume that people
rely on their mobile phones far more than they used to and
are thus likely to keep it more accessible as well. Based on
the widespread availability of smart phones, it is important
to re-investigate the assumption (and Patel et al.’s findings)
about users’ proximity to their phones, to determine if the
smart phone is as ubiquitous a device as we believe.
Using a series of surveys and interviews, as well as by
employing an application on Android mobile devices, we
have gathered both quantitative and qualitative data from 28
participants over 4 weeks of real-world behaviors with their
own smart phones. For this study we looked at participants’
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.
UbiComp ‘11, Sep 17Sep 21, 2011, Beijing, China.
Copyright 2011 ACM 978-1-4503-0630-0/11/09...$10.00.
Copyright 2011 ACM 978-1-60558-843-8/10/09...$10.00.
proximity to their smart phone via Bluetooth monitoring. In
addition we automatically collected a wide range of sensor
data from their phones. Supplementing this objective field
data are interviews, surveys and day reconstruction
exercises with users that help to give a sense of context.
Beyond these 28 participants, we also surveyed 367 smart
phone users about their perceived proximity to their smart
phones and their habits surrounding smart phone use.
The contribution of this paper is three-fold. The first is to
present empirical evidence to understand the degree to
which current smart phones are an accurate proxy for their
owners’ location and context. Second, we identify themes
that help explain these results, providing implications for
future application development. Finally, we demonstrate
that we can build accurate predictive models of proximity
using readily available features about user activity. We
begin with a discussion of related work.
RELATED WORK
As mobile phones become smaller and more powerful
computing devices, they have also become more prevalent
in the world at large [5,14]. The evolution of the mobile
phone has led it to move far beyond a simple
communication device and now is a versatile tool for
connecting people to their digital lives on various social
networks and e-mail, providing them with access to the
Internet, and even for entertaining them by supporting the
playing of games, music, and video. This evolution has
even lent itself to referring to these robust devices as smart
phones. The potential for the use of smart phones in
ubiquitous computing has not gone unexplored over the
past decade.
Leveraging the capabilities of smart phones, data can now
be captured that would have previously required carrying
customized hardware or a large number of specialized
devices [22]. Examples of new applications enabled by the
convergence of functionality in smart phones include
healthcare [9,12,21], support for the elderly [18],
augmenting advertising [16], understanding social networks
and behavior [6,17], among many others.
Knowing that a user has her smart phone on and nearby (or
tends to) is useful for ubiquitous computing and mobile
application development, as the smart phone offers the
unique ability to communicate information to a user and
collect information from a user, regardless of location or
time. However, it is not just knowing that the phone is
nearby that can offer a meaningful contribution. There is a
whole field of research and design that has arisen from
being able to leverage the mobile phone as a means to
gather contextual information and build an understanding of
human behavior and the environment at large. Related work
in this area includes activity sensing endeavors such as
UbiFit [1], SenSay [15], and MotionBand [9] that leverage
context-sensing to support people’s day-to-day actions.
In 2006 Patel et al. studied people’s proximity to their
mobile phone as well as their perceived proximity to their
phone and found that people had higher expectations of
their availability to their phone then was found to be the
case in actuality [20]. On average they found that the phone
was on 81% of the time. When it was on, it was within
arm’s reach 58% of the time, and within the same room as
their owners an additional 20% of the on time, leaving 23%
when the phone was on and considered “unavailable” to the
user. In actuality, the true unavailable time should include
the time when the phone was off, equaling 38% of the day,
on average. Trends that were noted upon by Patel include a
small, not statistically significant variance on cell phone
proximity between weekday (59%) and weekends (53%),
waking versus sleep (61% awake, 52% asleep, and home
and away (est. 50% to 71%). Furthermore they explored
whether a proximity relationship between users and their
mobile phones could be predicted by applying classifiers to
context data that could be easily acquired from the phones
themselves. With one week of training data, their classifiers
determined whether the phone was within arm’s reach with
86% accuracy.
The findings of this research, and its implications for how
the data gathered on a mobile phone can be leveraged to
give insights into the user’s context, have been cited in
numerous subsequent articles. While some have referenced
it as evidence that mobile phones can serve as a reasonably
accurate proxy for a user’s location and context [8,19,23],
others interpret the paper’s findings as the opposite: the
phone is not a good proxy for a user’s location [1,12,16].
Not surprisingly, the true interpretation depends on the
application, users, and context of use being designed for.
The question we want to explore in this paper is whether
with the evolution of the smart phone (and all the services
and apps now available) over the past five years since the
Patel et al. paper was published, users’ proximity to their
phone has changed. Smart phones have become not only
more commonplace, but more powerful computing devices
used to connect people to their e-mail, social networks, and
entertainment, through a wide variety of available apps. In
investigating user behavior at this stage of mobile
technology, we explore implications for mobile applications
that can inform developers, researchers, and technologists
as they innovate and envision new uses and applications for
mobile phones.
EXPERIMENTAL DESIGN
We now describe the design of our experiment in which we
sought to understand how users’ proximity to their phones
has changed with the availability of smart phones.
Overview
In order to accurately replicate the proximity study from
Patel et al. [20], we contacted the authors of this paper and
requested the experimental instruments they used and asked
for additional details on their experimental setup. By doing
so, we were able to closely replicate the original study. We
briefly describe that study and the instruments used, and
highlight differences in our study.
In keeping with the original experimental setup, we used
mixed methods: surveys to collect information about
subjects’ perceptions about their phone use and phone
proximity, and a 4-week long deployment of data collection
software on subjects’ own Android phones to collect both
proximity information and a wide variety of contextual
information. We also conducted weekly interviews to
provide additional context to this data. Our study differs
from the original study, which provided mobile phones to
subjects, and included 16 participants for 3 weeks.
Surveys
The original survey instrument asked questions about the
respondents’ age and occupation, how they thought they
used their phones and how close they thought they kept
their phones in different situations. We adapted this survey
to collect additional information on mobile applications
they use and their frequency of use, what applications they
would be interested in using, as well as their general
experience with the phone and expectations being met (or
not) with respect to mobile communications. We also
collected detailed socio-economic status information of
respondents, including education level, experience with
mobile technologies (year started using).
We used the responses to the survey to select subjects for
our study of smart phone proximity. Subjects were
randomly selected from those respondents having an
Android phone with an unlimited data plan and being
willing to participate in the 4-week long study.
Mobile Users Data Collection Study
Interview
Participants selected for our study came to our lab, where
we explained our study and what we expected of them in
detail. We also deployed our proximity monitoring and
context data capturing framework on their Android phone,
provided them with a Bluetooth device to collect proximity
information, and collected some proximity calibration data.
We describe our automated data collection process below.
Android AWARE Data Collection Framework
Sensors and Internet connectivity in mobile devices provide
researchers an opportunity to capture real-life context
information from the owners of the mobile devices and to
collect information on the proximity of the phones to their
owners. We developed the AWARE Android framework, to
help gather this information. The framework was developed
using Android SDK 2.1, was tested and then deployed on a
variety of Android devices throughout our user study. In
consideration of space, we only present the sensing
modules in the framework that are relevant to our
exploration of phone proximity. While some of these
modules may not seem relevant to proximity, we erred on
the side of sensing completeness in order to identify as
many factors as possible that impact proximity.
Activity Manager: Every 3 seconds, this module collects
information about the active, inactive and background
processes, current active activity on the screen, CPU and
memory usage by application and system by querying the
Android API Activity Manager.
Battery Manager: This module logs battery-related events
with information such as percentage, temperature, health,
voltage, technology and uptime (amount of time without
charging), when the phone was plugged in and unplugged
to the power supply (USB and AC).
Bluetooth Manager: The framework scans each minute for
Bluetooth devices in the vicinity, keeping a record of their
MAC address, friendly name and Received Signal Strength
Indication (RSSI) value.
Call Manager: This module keeps track of incoming,
missed and outgoing calls, including call duration, time of
call and phone number.
Phone Manager: This module captures the phone’s carrier
information on the device, such as tower location, whether
the phone has an active data plan, IP address, the device ID
(IMEI for GSM devices and MEID for CDMA devices), the
device’s phone number, neighbor cell towers, network
country, network operator, network type (CDMA, GSM,
UMTS, etc.) and roaming status. It also provides
information about the software running on the device,
including version, manufacturer and device model.
Location Manager: This module collects the device’s
location (latitude, longitude, bearing, altitude, speed and
accuracy), using network triangulation and GPS
coordinates. As per the Android’s developer
recommendation, a one-minute interval is used for polling
for GPS and Network-provided coordinates. The Location
Manager first acquires a network-provided location, as it
requires less battery power to quickly get the mobile
device’s location. Once a preliminary location is acquired, a
GPS location is requested, as network location is less
accurate. If the acquisition of GPS location fails for some
reason (e.g., user is indoors, phone failed to acquire satellite
GPS signal, GPS turned off), the network location is
logged. If the user is moving, a new location is requested
for every ten meters of motion on GPS or for every 100
meters through the network location.
Network Manager: This module logs network traffic
(received and sent) on available network interfaces (Wi-Fi,
Carrier network) and assigned IP address for each network
interface, along with network connections/disconnections.
Screen Manager: This module detects when the user turns
On/Off the screen and unlocks/locks the screen.
Sensor Manager: This module logs sensor events from all
the available sensors on the device, such as the
accelerometer, ambient light sensor, magnetometer,
pressure sensor, gyroscope, orientation sensor and
temperature sensor. It also captures the vendor, precision,
power consumption and sensor range values for each.
Messaging Manager: This module logs SMS and MMS
messages received/sent (including time and phone number).
Weather Manager: Each time the device’s location changes,
the weather forecast for the current day and location is
gathered from Weather.com.
Wi-Fi Manager: Every minute, this module logs the current
Wi-Fi state (active, inactive) and access point information
(connected and neighbor access points), such as MAC
addresses, broadcasted SSID, hidden SSID, link speed,
RSSI values, network capabilities, and each channel’s
frequency.
WatchDog: The WatchDog monitors the framework
operation each minute, verifying that all the data logging
modules are running and restarts any that are not. It also
pings our server every 5 minutes to indicate that the
framework is still alive and collecting data. A separate
server process notifies the study participant by email to
check her phone (and eventually reboot it), if it does not
receive this ping for more than 30 minutes.
We stored all logged data into an SQLite database on the
mobile’s phone external storage (i.e. miniSD card), rather
than using the phone’s internal memory. Offloading the
data from internal to external memory kept the phone’s
memory available for the proper operation of the device,
reducing the influence of the framework on the device’s
operation.
Figure 1. Bluetooth tags given to participants
Proximity Data Collection
To collect data on the proximity of users to their smart
phones, we provided Bluetooth devices to our subjects, as
in the original Patel paper. However, unlike the original
work, we did not have enough of a single type of Bluetooth
device, so we used a combination of BlueLon Bluetooth
tags (3) and Nokia Bluetooth GPS devices (25) (Figure 1).
We provided lanyards to each participant to wear the device
around their neck. Every morning, participants received an
automated email, reminding them to wear their
device.Similar to the original study, the Bluetooth Manager
in the AWARE framework performs a Bluetooth scan every
60 seconds, and determines the distance of the phone from
the provided Bluetooth device using RSSI measurements.
However, because we did not provide smart phones to our
subjects and because we used a variety of Bluetooth
devices, we collected calibration data for each user (smart
phone-Bluetooth device pair). We collected a few minutes
of RSSI data for each of the following: within arm’s reach
(1-2 meters), within the same room (5-6 meters) and
unavailable (beyond 6 meters). After removing outlier RSSI
values (using a quartile approach), we identified the range
of valid RSSI values for each of our 3 distances. We do
Table 1: Demographic information, percentage data lost to
framework errors. Also, ignoring lost data, percentage phone
off and proximity without (and with) off data
Gender
Profession
%
bad
data
%
arm
%
arm +
room
1
F
Software eng.
4
47 (40)
95 (80)
2
M
Owner, moving
company
25
45 (34)
78 (59)
3
M
Admin. Asst.
21
27 (22)
94 (75)
4
M
Driver
8
76 (52)
89 (61)
5
M
Student
16
76 (60)
99 (79)
6
M
AP coordinator
6
85 (68)
97 (77)
7
F
Designer
19
48 (44)
90 (83)
8
M
Web developer
54
53 (44)
72 (60)
9
M
PC technician
17
25 (19)
87 (68)
10
F
Software eng.
7
87 (82)
100
(94)
11
M
Graphic
designer
14
30 (23)
68 (53)
12
M
Paramedic
15
65 (45)
99 (68)
13
M
Program
manager
53
54 (41)
100
(77)
14
F
Researcher
16
32 (24)
44 (33)
15
M
Salesman
11
26 (22)
80 (68)
16
M
Collections
agent
18
38 (34)
99 (91)
17
F
Librarian
9
74 (68)
96 (88)
18
M
Software eng.
5
47 (44)
100
(93)
19
F
Photo lab tech
22
51 (42)
66 (56)
20
M
Advertising
agent
13
96 (83)
99 (86)
21
M
Software eng.
7
68 (42)
98 (60)
22
M
Systems
analyst
11
43 (34)
86 (68)
23
M
Student
51
67 (45)
99 (66)
24
M
Teacher
18
38 (17)
82 (38)
25
F
Therapist
0
36 (21)
78 (46)
26
M
Social worker
7
38 (32)
93 (78)
27
M
Software
analyst
37
34 (24)
84 (58)
28
F
Manager
9
73 (59)
99 (81)
note that even when using the same model of phone and the
same model of Bluetooth tag, calibration was necessary, as
different combinations of the same models resulted in
different RSSI ranges.
Weekly Interviews
When our deployment subjects returned to our office each
week over the 4-week study, we interviewed them for
ground truth about their proximity to their mobile phone.
This information was compared to the data automatically
logged by the AWARE framework. Similar to the original
study, participants completed a diary of the previous 24-
hour period wherein they record their activity and the
relative location of their mobile phone, as suggested by the
Day Reconstruction Method [13]. This method breaks the
day into episodes described by activities, locations and time
intervals, and the location of the phone during these times.
During the interview, users explained, in more detail, the
factors influencing their proximity. This way causalities and
relations between proximity and user behavior could be
identified, and any inconsistencies in the AWARE
framework data could be clarified. As well, they indicated
when they forgot the phone/tag, or took the tag off.
SUBJECTS
Thirty subjects were recruited using Internet advertisements
in <city removed> using the survey described above as a
screener. We compensated subjects with $250 each for
participating in the entire study. Subjects’ ages ranged from
18 to 45, and included 9 females and 21 males. They came
from a range of ethnic and cultural backgrounds and had
varying income levels and occupations (see Table 1). Two
subjects withdrew from the study: one on the second day of
data collection and the other after completing only the first
2 weeks. In the following section, we present the results of
our analysis of the collected data from the remaining 28
participants, using Droid, Droid X and Nexus One phones.
Figure 2. Distribution of proximity levels for each of the 28
participants, with (upper) and without (lower) off data, with
the last bar representing the average across all participants.
RESULTS
Participants ranged in their participation from 27 to 30
days. On average, our phone failed to collect Bluetooth
proximity data (but collected other data) 18% (std. dev.
15%) of the time. This number is high particularly because
for some subjects we only noticed that there was a data
collection error at the weekly interview, and that no data
was collected that week. The remainder of our results and
analyses will not include this data. Of the remaining time,
users either turned their phone or just our application off for
an average of 22% of the time (Figure 2). One reason this
was so high is that users turned their phones off to conserve
battery when they did not think they would use them or
could not use them. One subject, S24 was a teacher who
turned off her phone during school hours and at night. S17
was pregnant and turned off her phone in the evening and
night. Despite our attempts to keep the AWARE
framework’s energy footprint low by making it mostly
event-driven, some of our subjects complained about the
impact it had on their phone battery and having to recharge
more frequently.
Proximity Results
We acquired between 13430 and 37564 proximity samples
(i.e., number of minutes) from our subjects, with the
average being 26474. Because not all subjects participated
for the same number of days, these results are best viewed
as percentages of participation (not including framework
data acquisition errors, but including time when the phone
was turned off) range from 46 to 94%, averaging 78%.
In contrast to our hypothesis that users of smart phones
carry their phones with them (i.e., within arm’s reach) more
than users of the previous generation of mobile phones, we
found that our participants had their phones within arm’s
reach on average for only 53% of the time when the phone
was on. This is similar to what Patel et al. found in 2006:
58% (see Table 2 for proximity percentages that include the
time the phone was off). However, participants’ perceptions
were that their phone was within arm’s reach 91% of the
time. While most participants grossly overestimated their
proximity, there were 3 subjects (S1, S26, S27) whose
estimation was ~58%, much closer to the actual proximity.
However, while there was not an increase in the amount of
time the phone was within arm’s reach, we did find a
substantial increase in the amount of time that our subject’s
smart phones were outside of arm’s reach but were in the
same room as them: 35% in our study vs. 20% from the
previous study. Combining both within arm’s reach and
within the same room (arm+room) results in a total of 88%
for our smart phone study and 78% for Patel et al.’s mobile
phone study.
Table 2: Comparison of proximity between original study and
ours, not including (and including) off time
Arm’s Reach
Room level
Arm + Room
Original
58% (47)
20% (16)
78% (63)
Our study
53% (42)
35% (28)
88% (69)
Proximity and Contextual Factors
We also examined the impact of different contextual factors
such as day of week, time of day, and location. There were
no differences in the proximity of the phone between
weekdays and weekends: 53% and 52% within arm’s reach
for weekdays and weekends, respectively; and 89% and
87% for within room (and arm’s) reach, respectively. This
matches participants’ perceptions that there was little
difference between weekends and weekdays. The original
paper reported similar results of 59% and 53% for within
arm’s reach for the weekdays and weekends, respectively.
The phone was within arm’s reach 56% of the time when
subjects were sleeping (estimated between 11pm and 7am),
and 51% of the time, at other times of the day, whereas the
Table 3: Comparison of proximity at different times of day
Arm
Room
Arm +
Room
Morning (7-9am)
57% (46)
30% (23)
87% (69)
Daytime (9am-6pm)
51% (40)
36% (28)
87% (68)
Evening (6-11pm)
48% (37)
40% (31)
88% (68)
Night (11pm-7am)
56% (46)
33% (26)
89 %(72)
Not Night (7am-11pm)
51% (40)
37% (29)
88% (69)
Patel paper showed a different trend with percentages being
52% and 61%, respectively. There was less distinction at
the arm+room proximity: 89% while sleeping and 88% at
other times. Table 3 shows the distribution of proximity for
different times of the day.
The phone was within arm’s reach 46% of the time when
subjects were home (Patel: 50%), and 54% of the time
(Patel: 71%) when in named locations other than home, and
within room and arm’s reach cumulatively 83% (Patel: 77%
when at home), and 85% (Patel: 82% when not at home),
respectively. Table 4 shows the distribution of proximity
for different categories of locations that our subjects
identified in our initial interview. While proximity at the
arm level varied, proximity at the arm+room level stayed
relatively stable. Most subjects reported that their proximity
to their phone was not different between work and home,
and that perception bears true.
We also checked to see if there was a correlation between
proximity at the arm or arm+room level with either time
spent talking on the phone or the number of SMS/MMS
sent. However, these factors were not correlated.
Table 4: Comparison of proximity in different locations
Arm
Room
Unavailable
Arm + Room
Home
46%
36%
17%
83%
Not Home
54%
31%
15%
85%
Work
48%
33%
18%
82%
Shopping
62%
20%
17%
83%
Leisure
50%
37%
13%
87%
Family
56%
33%
10%
90%
Friends
51%
30%
18%
82%
Gas
74%
3%
22%
78%
Factors Affecting Phone Proximity
We now discuss factors that impact users’ proximity to
their phone, based on qualitative information collected from
the 28 participants in our data collection study, as well as
the 339 additional subjects who responded to our survey but
were not selected. Like the original study, we derived our
factors using affinity clustering to group the self-reported
factors from our interviews and surveys into themes, and
from the features that contained the most information gain
from our predictive models of phone proximity.
During the weekly interviews, as part of the day
reconstruction method, we asked participants to describe
their activities and the proximity of their phone over the
past 24 hours. We asked them for additional detail about
why their phone was in a particular location throughout this
day. We first used this information and the survey data to
identify the themes in the original study that we did and did
not (light grey text below) have evidence for, and then to
identify new themes.
1. Routine: The phone’s proximity was linked to users’
flow of usual activities, e.g., a) at home, leaving phone
in a fixed/central position such as on a coffee table or
shelf or at work, leaving it on their desk; b) at home,
the phone is with the user to support using different
applications substituting for a PC or to call a spouse
inside a big house; c) outside the home, carrying the
phone in a pocket or on a belt clip by a male, and in a
purse or bag by a female (depending on her outfit).
2. Environment: The phone’s proximity is related to the
physical constraints of the space. For example, at
home, users kept their phones in the same rooms as
them, while in the car, the phone was most often within
arm’s reach.
3. Physicality of person/activity: The phone’s proximity
is related to the physicality of the person or the
person’s activity. For example, while playing sports or
exercising, we found that users chose to keep their
phones with them to listen to music. While the theme
matches that of Patel et al., we find the opposite result.
4. Disruption to others: In contrast to the original study,
we have not identified any evidence suggesting that a
user’s phone proximity is based on how it might affect
other people or the environment. This could be because
social norms around cell phone use have evolved over
the last 5 years [1].
5. Disruption of self: The phone’s proximity takes into
account the impact of proximity on the user. For
example, at home, users kept phones in central places
with the idea that it could get their attention regardless
of their location. Others who responded to our survey
reported that they put their phones “away” on
weekends so as not to be bothered.
6. Regulations: We identified a number of situations
where users turned off their phones in certain locations
due to legal or other specific regulations preventing
use. For example, one subject was a teacher who had to
turn off her phone in school. Similarly, others turned
off their phones in churches and hospitals, and in other
locations where camera phones were forbidden.
7. Use of phone by self: Users made choices about the
phone proximity based on their anticipated use of the
phone. Unlike the earlier study, rather than keeping
them close for making a phone call, our subjects did
this for access to data, e.g., carrying the phone inside
the house to be able to check something quickly on the
Internet. In addition, from our survey respondents, we
saw that some people used their mobile phones to
check their private email accounts while at work.
8. Use of phone by others: The phone’s proximity was
affected by the idea that someone else would want to
contact the subject. For example, users would keep
their phone in a fixed/central location in their home to
hear phone calls (similar to the previous study), or be
notified about email/SMS arrivals (new in our study).
9. Use of phone both by self and by others: We primarily
saw evidence of this theme through descriptions of
coordination efforts. For example, some kept the phone
close by to make it easier to coordinate efforts for
going out with friends on the weekend, while the
moving company owner did it to coordinate his staff.
10. Use of handset by others: We did not identify any
evidence suggesting that a user would make a choice
about the phone proximity based on the expectation
that somebody else would physically use the device.
This could be because smart phones and cell phones
have become widely available [4,14] and, therefore, are
becoming more personal devices.
11. No need for use of phone: When users believed they
were not going to use their phones soon, users were
willing to be further away. While only about one-half
of our subjects had a landline phone, for those who did,
the expectation that a caller could reach them using this
line, was evidence of this theme. While at work, having
access to a PC for relevant data/Internet supplanted use
of the phone and affected its proximity.
12. Technical resources: The phone’s proximity is
impacted by technical considerations inherent to
limitations of the phone. We saw evidence of this
theme when users limited the mobility of their phone
when charging (USB or AC adapter). In addition,
survey respondents physically moved with their phone
to acquire improved signal reception in their home.
13. Quick trips: Unlike the previous study, we found that
users did not leave their phones behind when taking
quick trips. Our subjects tended to take their phones
with them when going on a coffee break and when
going to the bathroom. This may be related to theme
#19 about the use of the phone during idle periods.
14. Memory and forgetfulness: We saw multiple instances
where users simply forgot their phone at home or work
or left/forgot it someplace temporarily.
15. Protection of phone from others: Similar to the
previous study, we saw that users made choices about
phone proximity to protect the phone from tampering.
For example, users put their phone out of reach while
playing with children. Similarly, survey respondents
reported leaving their phones behind when going out
for fear their expensive smart phones would be stolen.
In addition to the analysis of the 15 themes identified in the
original paper, we identified 5 new emergent themes:
16. Costs associated with usage: The phone’s proximity is
associated with monetary costs related to phone usage.
While everyone in our study had an unlimited data
plan, they still had to pay for data usage when traveling
abroad. As such, the few subjects that left the U.S.
during our study tended to keep their phone off.
17. Personal Utility applications: The phone’s proximity is
related to use of its applications in a given context. For
example, phones were often used as an alarm in the
bedroom while subjects were sleeping, to support
nutrition or sports training in the gym, and to replace
the use of a PC at home.
18. Data privacy on the phone: The phone’s proximity is
related to access of data applications holding or
accessing private data on the device, e.g., checking
private email on phone while at work and having
access to a corporate network while not at work.
19. Idle time in between activities: The phone’s proximity
is related to time spent on mobile data applications
while waiting for some activity to start. For example,
users checked their email or accessed the web while
waiting for a friend, waiting for a bus, or even while on
the toilet.
20. Applications for planning or scheduling coordinated
tasks: The phone’s proximity is related to tasks
requiring the management of coordination and
cooperation. For example, some users used shared
grocery lists or to-do lists with a partner, used Google
Calendar to add new group events, and then accessed
this information at later times.
21. Protection of phone from environment: The phone’s
proximity was affected by users’ beliefs that the phone
had the potential to be damaged. Survey respondents
reported leaving their phones behind when going
fishing (wary about water damage) and when cooking
in the kitchen (wary about water and heat damage).
PREDICTING USERSPHONE PROXIMITY
Similar to the Patel paper, we also investigated whether we
could predict users’ phone proximity using information that
are already available on the phone, rather than using our
extra Bluetooth tag. For each subject, we use our Bluetooth
tag proximity information as ground truth, and attempt to
predict whether the phone was within arm’s reach, or within
arm+room. using the contextual information we collected
with our AWARE framework. If the predictions are
accurate, application developers can use our prediction
models to determine when they can use the phone to collect
contextual information from phone owners (arm’s reach) or
to collect contextual information about the owner’s
environment and deliver information to them (arm+room).
We created models that could classify phone proximity. We
used a decision tree classifier using the ID3 algorithm so we
could interpret the resulting trees and determine which
features were most important to the classification task.
Features near the root of decision trees usually have high
predictive power and can be treated as important features.
We formulated the model building as two supervised
learning problems, in which the class labels are three (arm
vs. room vs. unavailable) and two (arm+room vs.
unavailable) levels of proximity. Each data instance is a
feature vector extracted with a one minute time window
from the logged contextual data. We used three different
feature sets to build our models. We ranked features for
each subject using the Greedy Stepwise search method with
Consistency Subset evaluation method from Weka [10], and
used the top 3, 5 and all features. Figure 3 shows the 10-
fold cross validation results using all the data from each
subject for our 2 classification problems. We achieved 75
and 83% accuracy for the 3-class and 2-class problems,
respectively, with large variations across our subjects.
Figure 3. Classification accuracy for 3-class (upper) and 2-
class (lower). Blue, red, green represent 3, 5 and all features,
S3 (only 2 weeks of data) is included for completeness.
Figure 4. Analysis of the number of weeks of training required
for accurate 3 class (arm vs. room vs. unavailable) and 2-
class (arm+room vs. unavail) models.
To determine how many weeks of data were needed to
build a reliable model, we trained 3 additional models on
the first one, two, and three weeks of data, testing on the
remaining data using 10-fold cross validation. Figure 4
shows that while 3 weeks of data may not be enough for
producing accurate models in the 3-class problem, 1 week
of training provides reasonably high accuracy in the 2-class
problem. Also, exploiting more features requires more
training data to model their relationship with proximity.
We also analyze the features selected using the search
method (Feature in Table 5) and the features at and near the
root of the decision trees (DT). Similar to the original study,
we found hour of day and time of day to be quite predictive
of proximity. However, location was not very useful. The
other features we found to be useful were very related to
activities that the user performs and interaction with the
phone: acceleration, application used, battery level, battery
temperature and screen status. Battery temperature is
particularly interesting as high values are correlated with
close proximity: carried in a pocket next to a warm body, or
being used for CPU-intensive applications.
Table 5: Predictive features for 3-class and 2-class prediction
problems. Number of participants using each feature from the
search method (Feature) and decision trees (DT).
arm+room vs. other
arm vs. room vs. other
Feature
DT
Feature
DT
mean acceleration (acc)
8
2
15
2
std deviation of acc.
5
0
12
3
application used
4
8
4
8
battery level
18
20
15
22
mean battery temp.
24
19
24
17
tower ID for CSDMA
0
3
0
0
day of the week
29
27
26
29
tower ID for GSM
1
0
1
0
hour of the day
28
24
28
26
screen status (on, off)
21
4
12
2
ringer status (on, off)
2
6
0
2
weather
5
1
3
3
DISCUSSION
We now discuss our hypothesis about the proximity of
smart phones, the results of our study, and our analysis.
Actual Phone Proximity
We were very surprised to see that there was no increase in
the proximity of users to their mobile phones, with the
availability and widespread use of smart phones. Our
intuition led us to believe that access to the Internet, the use
of smart phones as entertainment devices, and the huge
uptake in apps would increase proximity. In fact, we saw a
slight decline at the arm’s reach from 58% to 53%.
However, we did see a considerable increase in proximity at
the room level from 20% to 35%, resulting in an overall
increase at the arm+room level from 78% to 88%. We first
describe the implications of this increase and then discuss
the possible reasons for it.
Many ubicomp systems make the assumption that users
always have their smart phones with them. If that
assumption were true, it would allow these systems to:
Collect user context (e.g., motion, activity)
Collect user’s environment context (e.g., sound)
Get the user’s attention at any time and present
information (e.g., notifications)
Provide an always-available service for the user
However, our work and Patel’s earlier work showed that
users don’t have their phones with them at all times.
However, our work does demonstrate that users often are in
the same room (or very close by) to their phones. While
only a little more than half the time can the phone be used
as a proxy for the user’s physical context, almost 90% of
the time, it can be used as a proxy for environmental
context, a mechanism for getting the user’s attention, and a
medium for delivering always-available services.
From our interviews, surveys and our analysis of themes,
we have a better understanding of why this change occurs
in users’ proximity to their phones. We heard several
examples from each of our participants of placing their
phone down on a table or desk, to keep the phone close by
and easily reachable, but not immediately at-hand. It was
enough to have the phone easily reachable in the case of
notifications or phone calls, to look something up on the
Internet or to use an app for a short period of time. None of
these require the phone to be within arm’s reach. Almost all
the themes discussed (Patel’s orginal and our new ones)
help provide evidence for why people keep their phones
nearby, but not necessarily within arm’s reach: routine,
environment, disruption of self, use of phone by self, use of
phone by others, use of phone by self and others, use of
handset by others, no need for use of phone, technical
resources, quick trips, memory, personal utility
applications, data privacy on the phone, idle time in
between activities, applications for planning, and protection
of phone from environment.
Perception of Phone Proximity and Individual Difference
On average, most of our participants believed they were in
close proximity to their smart phones almost 22 hours a
day! Half of our participants said they were always next to
their phones. As we have shown, this number is closer to 10
hours a day, when taking off time into account. Users
clearly tend to greatly overestimate their proximity.
However, users are within arm+room level almost 16.5
hours per day, when considering off time. We do see
considerable individual differences. From Table 1, 16 of
our participants keep their phone at the arm+room level
over 90% of the time the phone is on, with 13 of those at
95% or above. The remaining participants have arm+room
levels ranging from 44 to 89% and average 76%.
For the 16 participants that keep their phones nearby almost
all the time, it is unclear whether a prediction system is
necessary for determining when the phone is nearby. With
proximity rates of more than 90%, assuming the phone is
nearby (arm+room) could be more accurate than many
learned models of proximity. There are 7 users whose
proximity at the arm level is 75% or greater, and the same
statement can be made for them.
Phone Proximity By Context
Just as we were surprised by the lack of increase in
proximity at the arm level with smart phones, we were also
surprised that there was no difference in proximity between
weekdays and weekends. There were some however,
interesting differences in proximity at the arm level by
location. Users were less likely to keep their phones within
arm’s reach at home, work, places of leisure, and at friends’
residences. In contrast, they were more likely to keep their
phones within arm’s reach while shopping, buying gas, and
visiting with family. The difference between proximity
from home to other locations defined and geo-located by
our participants was 46 to 54%. From our interviews with
our participants, we found that in places where they were
most comfortable and familiar, they tended to leave their
phones further away from them: close by so they could use
them, but not within arm’s reach. As well, the 3 locations
where phones were closer, tend to involve activities that
allow for smart phone interaction during short idle periods.
Similarly, we saw differences in proximity by time of day.
While proximity at the arm+room level was independent of
this factor, proximity at the arm level was not. Participants
tended to have their phones closer during sleeping hours
than during other hours (56 vs. 51%). This can partly be
explained by the use of smart phones as an alarm clock.
Users were also closer to their phones during the morning
hours (7-9am). Participants explained that they used their
phones to check email and other Internet resources shortly
after waking up, and during their commute to work. The
times with the lowest proximity are the work hours and the
time at home after work and before sleeping. Combined
with the location results just described, the lower proximity
values are not surprising. Both differences in location and
time of day offer opportunities for designers of mobile
phone applications to use the phone as a proxy for the user
and their environment, and for communicating with users.
Predicting Phone Proximity
We produced models of proximity that used contextual
factors from phone sensors that could predict when the
phone was within arm’s reach with 75% accuracy, and
within arm+room with 83% accuracy. These models are not
computationally intensive and can be executed quite easily
on today’s smart phones. There was large variability in
accuracy by users, and in the features that were most
predictive. This implies that a single population model or a
small number of population models (as suggested by Patel
et al.) may not be possible. It is interesting that all the most
predictive features are directly related to user activities,
both those in the real world (e.g., movement) and those on
the phone (e.g., phone usage as measured by battery level
and temperature). Perhaps by collecting information on the
active applications on the phone and using more
sophisticated features related to user activity we can
improve the accuracy of our models. Certainly by taking
into account those that were almost always within the
arm+room level, and combining with the predictive models
(taking the maximum accuracy for each subject), we can
reach an overall predictive accuracy of 92%.
Limitations
As with the original study, our study has some limitations.
We studied a limited population (28) of smart phone users
for a limited period of time (4 weeks). As such, it may be
difficult to generalize these findings to other populations of
smart phone users. We provide this detailed exploration of
how smart phone users use their phones and their proximity
to their phones to illustrate the challenges and opportunities
for designers of phone applications looking to use the
mobile phone as a proxy for user context and attention.
One issue with our study was the amount of data we lost
data due to issues with our data collection framework. We
discovered that some of our participants had an automated
task killer app on their phone. A task killer app removes
processes (in our case, modules from our framework) that
use up significant memory or CPU. This caused us to lose
significant amounts of data (up to 54%, although most had
data losses ~10% or lower) from some of our participants,
until we identified the app and asked them to remove it.
Another issue we did not account for was the amount of
time the phone was off. While both we and the original
paper report reasonably high proximity numbers from our
studies, these numbers drop significantly when time when
the phone is off is considered. Participants in our study had
their phones off or our framework off 22% of the time. This
results in our arm+room proximity to drop from 88% to
69%, when taken into account. As off time can be
considerable, especially for some users (e.g., paramedic,
teacher, pregnant participant, foreign travel), off time needs
to be considered when designing mobile applications that
make assumptions about phone availability.
CONCLUSION AND FUTURE WORK
We have presented a field data collection-based study of 28
smart phone users to understand whether their phones can
serve as proxies for their context and availability. We found
that when their phones are on, they are only within arm’s
reach 53% of the time, but within arm+room 88% of the
time. Based on these results, we build on the work of Patel
et al., and show how mobile application designers can
leverage smart phones as proxies for users’ environmental
context, availability for delivering information and
availability for accessing information. We demonstrate that
we can reasonably predict user proximity with easily
collected features about user activity, and when combined
with knowledge about individual users and their normal
proximity, we are very accurate (greater than 90%). In the
future, we intend to collect and leverage additional features
regarding activity and user context to further improve our
predictive ability. We also will build this ability into mobile
applications as a demonstration of its effectiveness.
ACKNOWLEDGMENTS
This work was supported in part by the National Science
Foundation under grants 0746428 and 0910754.
REFERENCES
1. Banjo, O., Hu, Y. and Sundar, S.S. Cell phone usage and
social interaction with proximate others: Ringing in theoretical
model. The Open Communication Journal, 2, 2008: 127-135.
2. Brandt, J., Weiss, N., and Klemmer, S. R. Lowering the
burden for diary studies under mobile conditions. CHI '07
extended abstracts. 2007. 2303-2308.
3. Consolvo, S., et al. Activity sensing in the wild: a field trial of
ubifit garden. CHI '08. 2008. 1797-1806.
4. Cozza, R. Forecast: Mobile Communications Devices by Open
Operating System, Worldwide. Gartner Report. 2011.
5. Cozza, R., et al. A. Market Share Analysis: Mobile Devices,
Worldwide, 4Q10 and 2010. Gartner Report. 2011.
6. Eagle, N. and Pentland, A.S. Reality mining: sensing complex
social systems. Personal Ubiquitous Computing, 10 (March
2006): 255-268.
7. Entner, R. Smartphones to Overtake Feature Phones in U.S. by
2011, Nielsen Wire,
http://blog.nielsen.com/nielsenwire/consumer/smartphones-to-
overtake-feature-phones-in-u-s-by-2011/, March 26, 2010.
8. Garzonis, S. Mobile Service Awareness Via Auditory
Notifications. Doctoral thesis, University of Bath, June 2010.
9. Györb'ıró, N., Fábián, Á. and Hományi, G. An activity
recognition system for mobile phones. Mobile Networks and
Applications, 14 (February 2009): 82-91.
10. Hall, M., et al. The WEKA Data Mining Software: An Update.
SIGKDD Explorations, 11(1), 2009: 10-18
11. Helal, S., et al. Smart phone based cognitive assistant.
UbiHealth 2003, Workshop on Ubiquitous Computing for
Pervasive Healthcare Applications. 2003.
12. Henderson-Summet, V. Facilitating Communication for Deaf
Individuals with Mobile Technologies. Doctoral thesis,
Georgia Institute of Technology, 2010.
13. Kahneman, D., et al. A Survey Method for Characterizing
Daily Life Experience: The Day Reconstruction Method.
Science, 306 (December 2004): 1776-1780.
14. Kellogg, D. Among Mobile Phone Ussrs, Hispanics, Asians
are Most-Likely Smartphone Owners in the U.S. Nielsen Wire,
http://blog.nielsen.com/nielsenwire/consumer/among-mobile-
phone-users-hispanics-asians-are-most-likely-smartphone-
owners-in-the-u-s, February 2, 2011.
15. Krause, A., Smailagic, A., Siewiorek, D.P. Context-Aware
Mobile Computing: Learning Context-Dependent Personal
Preferences from Wearable Sensor Array. IEEE Transactions
on Mobile Computing, 5(2) Feb. 2006 pp. 113-127
16. Krumm, J. Ubiquitous Advertising: The Killer Application for
the 21st Century. IEEE Pervasive Computing, 10 (January
2011): 66-73.
17. Laasonen, K. Clustering and prediction of mobile user routes
from cellular data, 2005. Knowledge Discovery in Databases.
2005. 569-576.
18. Massimi, M., Baecker, R.M., and Wu, M. Using participatory
activities with seniors to critique, build, and evaluate mobile
phones. ASSETS 2007. 2007, 155-162.
19. Partridge, K. and Golle, P. On using existing time-use study
data for ubiquitous computing applications. In UbiComp'08,
2008, 144-153.
20. Patel, S.N., et al. Farther Than You May Think: An Empirical
Investigation of the Proximity of Users to Their Mobile
Phones. UbiComp ‘06. 2006, 123-140.
21. Preuveneers, D. and Berbers, Y. Mobile phones assisting with
health self-care: a diabetes case study. MobileHCI '08. 2008,
177-186.
22. Satyanarayanan, M. Swiss army knife or wallet? IEEE
Pervasive 4(2), Jan.-Mar., 2005, 2-3.
23. Varshavsky, A. and Patel, S. Location in Ubiquitous
Computing. In Krumm, J. (ed.), Ubiquitous Computing
Fundamentals, Taylor and Francis Group, Florida, (2010)
Chapter 8, 285-319.
... Text messaging interventions also have to compete with other everyday activities events, making it difficult for people to engage with interventions in a timely manner [100]. For example, a user may not notice an incoming text message at first because their phone is charging elsewhere or because they silenced notifications to avoid distractions at work [40,114]. ...
Preprint
Full-text available
One-way text messaging services have the potential to support psychological wellbeing at scale without conversational partners. However, there is limited understanding of what challenges are faced in mapping interactions typically done face-to-face or via online interactive resources into a text messaging medium. To explore this design space, we developed seven text messages inspired by cognitive behavioral therapy. We then conducted an open-ended survey with 788 undergraduate students and follow-up interviews with students and clinical psychologists to understand how people perceived these messages and the factors they anticipated would drive their engagement. We leveraged those insights to revise our messages, after which we deployed our messages via a technology probe to 11 students for two weeks. Through our mixed-methods approach, we highlight challenges and opportunities for future text messaging services, such as the importance of concrete suggestions and flexible pre-scheduled message timing.
Preprint
Daily life unfolds in a sequence of situational contexts, which are pivotal for explaining people’s thoughts, feelings, and behaviors. While situational data were previously difficult to collect, the ubiquity of smartphones now opens up new opportunities for assessing situations in situ, that is, while they occur. Seizing this development, the present study demonstrates how smartphones can help establish associations between the psychological perception and the physical reality of situations. We employed an intensive longitudinal sampling design and investigated 9,790 situational snapshots experienced by 455 participants for 14 consecutive days. These snapshots combined self-reported situation characteristics from experience samplings with their corresponding objective situation cues obtained via smartphone sensing. To account for the complexity of real-world situations, we extracted a total of 1,356 granular situation cues from different sensing modalities. We applied linear and nonlinear machine learning algorithms to examine how well these cues predicted the perceived characteristics in terms of the Situational Eight DIAMONDS, finding significant out-of-sample predictions for the five dimensions capturing the situations’ Duty, Intellect, Mating, pOsitivity, and Sociality. Analyses of (grouped) feature importance revealed that these predictions relied on complex constellations of cues representing various situational information about the Persons/Interactions and Objects present, the Events/Activities happening, and the current Location and Time. Furthermore, a nomological network analysis provided evidence for the construct validity of our cue-based DIAMONDS predictions. We conclude by discussing how smartphone-based situational snapshots, in general, and our prediction models, in particular, advance psychological research on situations.
Article
Automatically recognizing a broad spectrum of human activities is key to realizing many compelling applications in health, personal assistance, human-computer interaction and smart environments. However, in real-world settings, approaches to human action perception have been largely constrained to detecting mobility states, e.g., walking, running, standing. In this work, we explore the use of inertial-acoustic sensing provided by off-the-shelf commodity smartwatches for detecting activities of daily living (ADLs). We conduct a semi-naturalistic study with a diverse set of 15 participants in their own homes and show that acoustic and inertial sensor data can be combined to recognize 23 activities such as writing, cooking, and cleaning with high accuracy. We further conduct a completely in-the-wild study with 5 participants to better evaluate the feasibility of our system in practical unconstrained scenarios. We comprehensively studied various baseline machine learning and deep learning models with three different fusion strategies, demonstrating the benefit of combining inertial and acoustic data for ADL recognition. Our analysis underscores the feasibility of high-performing recognition of daily activities using inertial-acoustic data from practical off-the-shelf wrist-worn devices while also uncovering challenges faced in unconstrained settings. We encourage researchers to use our public dataset to further push the boundary of ADL recognition in-the-wild.
Book
Full-text available
Describes technological methods and tools for objective and quantitative assessment of quality of life (QoL) Appraises technology-enabled methods for incorporating QoL measurements in medicine Highlights the success factors for adoption and scaling of technology-enabled methods
Chapter
Daily behaviors influence an individual’s health and, in turn, all the domains of their quality of life (QoL). Accurately quantifying these behaviors may allow individuals to improve their overall awareness of these behaviors, make necessary habit changes, and receive more individualized treatment approaches. Currently, self-reported patient-reported outcomes (PROs) are the most common means of assessing daily behaviors. However, this method has multiple limitations, including the infrequency of collection, its subjective nature, its reliance on memory recall, and the influence of social norms. In comparison with PROs, using personalized and miniaturized technological innovations, including smartphones, mobile applications, and wearables, can enable the continuous assessment of daily life behaviors that contribute to or result from an individual’s QoL in a more accurate and timely manner. These technologies have the potential to transform the current state of quantifying QoL, allowing for improved research and the implementation of more individualized approaches to prevention and treatment. This chapter thus presents potential areas of future research and development opened by the use of these technologies in the field of QoL.
Chapter
Personal interactions are an important element of an individual’s health and life quality in the long term. As the site of many interpersonal interactions has been moved to the digital domain, human society has never been more intertwined. The digital footprints of interpersonal interactions can be quantified and measured via smartphones and wearables, providing more objective, quantitative, and accurate measurements. This chapter focuses on quantifying personal relationships in the context of quality of life, specifically focusing on novel technology-based quantification solutions. It first analyzes traditional qualitative quality of life measures based on subjective self-reporting that include measures of personal relationships, specifically the WHOQOL-BREF, WHOQOL-100, RAND-36, KIDSCREEN-27, SWLS, and Beach Center FQOL, as well as other non-validated measures. The chapter then proposes novel technological solutions for data gathering and analysis by introducing the concept of digital item representation, a process that leverages personal datasets originating from smartphones and wearables. The chapter also discusses issues relating to users’ privacy that influence the acceptance of such everyday technologies as well as the quality of data collected in the long term.
Chapter
Mobile network connectivity enables individuals to use various Internet-based applications and is nowadays an integral part of the physical environment. More specifically, this connectivity shapes individuals’ modes of gathering information and their communication capabilities. In turn, this impacts the individual’s decision-making and, in the long term, may influence their health and quality of life (QoL). This chapter focuses on longitudinal modeling of the availability of mobile connectivity such as Wi-Fi and 3G or 4G for individuals living in the Geneva area (Switzerland). We analyze connectivity over 5 years (2015–2020) based on data collected from 110 mQoL (mobile QoL) Living Lab participants. The participants are from three different cohorts corresponding to distinct data collection periods (2015–2017, 2018–2019, 2020). We derive four features that quantify an individual’s connectivity level: the network access technology (Wi-Fi or cellular), signal strength, the overall data consumption (upload and download), and the participants’ mobility patterns while connected. We also compare the connectivity levels of the three cohorts over time. Our findings reflect the relations between mobile connectivity and the smartphone network activity of the mQoL study cohorts during their daily activities, which may impact their QoL. We summarize the results and conclude this chapter by exploring the different QoL technologies and services enabled by mobile connectivity. However, the effects of connectivity on specific QoL domains, such as psychological aspects (i.e., positive/negative feelings) or social relationships, should be investigated further.
Chapter
This chapter will discuss the usage of more objective and unobtrusive ways technology can be used to assess leisure activities. It is well known that leisure activities are positively correlated with measures of quality of life and subjective well-being. How we spend our free time has a great deal of influence on how we subjectively assess the quality of our lives. One aspect of our leisure time, which is gaining more and more interest, is the use of smartphones and wearables. According to global statistics, almost half of the global population spends more than 5 h a day using their smartphones. The use of technology has a profound effect on the way we spend our lives, socialize and entertain. Because our use of technology leaves a massive amount of digital data, we are now able to search for patterns of digital behaviour and use them as proxies or predictors for real life behaviours, bypassing or complementing self-reports and subjective measures. Our discussion revolves around several aspects of technology and leisure time. First, how technology use relates to leisure activities and what alternative unobtrusive measures could be developed to measure or predict leisure activities. Second, we will discuss the positive and negative aspects of technology use.
Chapter
Quality of life (QoL) is a subjective term often determined by various aspects of living, such as personal well-being, health, family, and safety. QoL is challenging to capture objectively but can be anticipated through a person’s emotional state; especially positive emotions indicate an increased QoL and may be a potential indicator for other QoL aspects (such as health, safety). Affective computing is the study of technologies that can quantitatively assess human emotions from external clues. It can leverage different modalities including facial expression, physiological responses, or smartphone usage patterns and correlate them with the person’s life quality assessments. Smartphones are emerging as a main modality, mostly because of their ubiquitous availability and use throughout daily life activities. They include a plethora of onboard sensors (e.g., accelerometer, gyroscope, GPS) and can sense different user activities passively (e.g., mobility, app usage history). This chapter presents a research study (here referred to as the TapSense study) that focuses on assessing the individual’s emotional state from the smartphone usage patterns. In this TapSense study, the keyboard interaction of n = 22 participants was unobtrusively monitored for 3 weeks to determine the users’ emotional state (i.e., happy, sad, stressed, relaxed) using a personalized machine learning model. TapSense can assess emotions with an average AUCROC of 78%(±7% std). We summarize the findings and reflect upon these in the context of the potential developments within affective computing at large, in the long term, indicating a person’s quality of life.
Article
Ecological momentary assessment (EMA) is used to gather in-situ self-report on behaviors using mobile devices. Microinteraction EMA (μEMA), is a type of EMA where each survey is only one single question that can be answered with a glanceable microinteraction on a smartwatch. Prior work shows that even when μEMA interrupts far more frequently than smartphone-EMA, μEMA yields higher response rates with lower burden. We examined the contextual biases associated with non-response of μEMA prompts on a smartwatch. Based on prior work on EMA non-response and smartwatch use, we identified 10 potential contextual biases from three categories: temporal (time of the day, parts of waking day, day of the week, and days in study), device use (screen state, charging status, battery mode, and phone usage), and activity (wrist motion and location). We used data from a longitudinal study where 131 participants (Mean age 22.9 years, SD = 3.0) responded to μEMA surveys on a smartwatch for at least six months. Using mixed-effects logistic regression, we found that all temporal, activity/mobility, and device use variables had a statistically significant (p<0.001) association with momentary μEMA non-response. We discuss the implication of these results for future use of context-aware μEMA methodology.
Article
Full-text available
Observational research on the social impact of cell phone usage in public places suggests that the mere presence of cell phones in public conflicts the private and public spheres and inhibits social interaction with proximate others (strangers or known persons). The purpose of this paper is to develop a theoretical model for which social effects of cell phone usage in public places documented in observational studies can be empirically tested. In this paper, we discuss various variables to consider in the study of cell phone usage (CPU) and social interaction with proximate others (SIPO). We offer a modest experiment of CPU in the context of social participation, a form of social interaction. Focusing on helping behavior in particular, results indicate that while on the cell phone, users are less likely to offer help. Findings imply that CPU in public places can distract users from social responsibilities, as they neglect the environment surrounding them.
Article
Full-text available
This issue marks the end of Satya's second two-year term as editor in chief. He introduces his successor, Roy Want of Intel Research, and reviews the progress that has occurred in the last four years.
Article
Full-text available
Today, the cell phone is the dominant mobile device. What comes next? As circuit density improves, we see a convergence of mobile device functionality—cell phones combined with PDAs or with email capability or both. Where do we go from here?
Article
The use of pervasive computing technology such as java smart phones and multi-modal sensors in smart homes of the future can poten-tially enhance elders' independence and quality of life. We present ongoing research projects whose goal is to reduce the demand on elder's attention and effort while performing daily tasks. We present three applications: a Mobile Patient Care-Giving Assistant (mPCA), a General Reminder System (GRS), and an Augmented Awareness System (AAS). The mPCA applica-tion is a cognitive assistant designed to improve the independence of live-at-home for Alzheimer Disease (AD) patients. GRS is a reminder applica-tion targetted to elders with dementia. AAS is a notification application that boosts the elder awareness about certain events in the surrounding (mail de-livery, water leak, etc.) We present these applications and discuss the OSGi-based framework on which these applications are built.
Article
We present a novel system that recognizes and records the motional activities of a person using a mobile phone. Wireless sensors measuring the intensity of motions are attached to body parts of the user. Sensory data is collected by a mobile application that recognizes prelearnt activities in real-time. For efficient motion pattern recognition of gestures and postures, feed-forward backpropagation neural networks are adopted. The design and implementation of the system are presented along with the records of our experiences. Results show high recognition rates for distinguishing among six different motion patterns. The recognized activity can be used as an additional retrieval key in an extensive mobile memory recording and sharing project. Power consumption measurements of the wireless communication and the recognition algorithm are provided to characterize the resource requirements of the system.
Article
This survey article proposes the idea that advertising is the next major application for ubiquitous computing. Part of the support for this idea is that ubiquitous computing applications will very likely be supported by advertising, continuing the success of advertising on the Worldwide Web and paralleling the predicted growth of mobile advertising. More interesting, however, is that ubiquitous computing will eventually support advertising in several ways. The author explains how advertisers are already adopting certain ubiquitous computing technologies, and shows how ubiquitous computing research can help advertisers in the areas of ad targeting, ad feedback, customer awareness, and privacy. The article concludes with a scenario illustrating ubiquitous advertising along with suggestions for ubiquitous computing researchers in light of the author's predictions.